AI-ready questions.

821 ai-ready questions from the bank, ready to practice out loud.

Learn the ideas first · free with an account

Catching AI's confident mistakes Describing your own practice Prompt craft for everyday work STAR and CARL: structuring your story Which tasks to hand to AI Delegating to AI agents safely What data can go into which AI tool

All 821 questions

AI tool choices last weekWalk me through the last week of your work. Where did you actually reach for an AI tool, and where did you choose not to?Tool UsageEntry–leadership

Catching AI mistakesTell me about a time you caught an AI tool getting something wrong. How did you notice?JudgmentEntry–leadership

Workflow redesigned around AIWhat's a workflow you've redesigned around AI in the last six months, and what did the old version look like?Tool UsageMid–leadership

Communicating AI-assisted work to othersWhen you share AI-assisted work with a teammate or manager, what do you make sure they know?Stakeholder CommunicationEntry–leadership

AI tasks you stopped usingWhat's a task you used to use AI for that you've since pulled back from doing with it?JudgmentMid–leadership

AI tools that stuckWhich AI tools have actually stuck in your workflow, and which ones did you try and abandon?Tool UsageEntry–leadership

Catching AI mistakes before sharingHow do you set yourself up to catch an AI mistake before it goes downstream to someone else?Error HandlingMid–leadership

Keeping up with AI toolsHow are you keeping up with the pace of AI tools in your field without drowning in it?UpskillingMid–leadership

Human effort behind AI outputsWhen is the human still doing the work, even if the AI is producing the output?JudgmentSenior–leadership

Defending AI work to skepticsHave you had to explain an AI-assisted output to a skeptical colleague? How did you handle it?Stakeholder CommunicationMid–leadership

AI in one regular taskPick one task you do regularly. Walk me through how AI fits into that task today, end to end.Tool UsageEntry–leadership

First check on AI outputWhen you've handed something to an AI tool, what's the first thing you look at when it gives the answer back?JudgmentEntry–leadership

Cleaning up confident AI errorsDescribe a time an AI tool was confidently wrong and you had to clean up after it.Error HandlingMid–leadership

Work AI should never touchWhat's a part of your job that you'd never want AI involved in, and why not?Tool UsageMid–leadership

How your prompting has evolvedHow have your prompts changed over the last year? What were you doing wrong at the start?UpskillingMid–leadership

Building shared AI workflowsHave you built or modified a tool, prompt template, or workflow others on your team use? Walk me through it.Tool UsageMid–leadership

Signals a task fits AIWhat signals tell you a task is a good fit for AI, before you've started it?JudgmentEntry–leadership

Talking AI with resistant colleaguesHow do you talk about AI use with people on your team who are skeptical or resistant?Stakeholder CommunicationMid–leadership

When AI made you slowerTell me about a time using AI made you slower instead of faster. What happened?Error HandlingEntry–leadership

Your default writing toolWhat's your default tool for writing right now, and what made it the default?Tool UsageEntry–leadership

Deciding on AI workflow complexityHow do you decide when to invest in a more elaborate AI workflow versus just doing the thing yourself?JudgmentMid–leadership

AI advice that didn't workTell me about a piece of AI advice or pattern you copied that didn't work for you, and what you did instead.UpskillingMid–leadership

AI for messy early workWhat does your AI usage look like for the messy, half-formed parts of your work — not the polished outputs?Tool UsageEntry–leadership

Checking external AI outputWhen you share AI-assisted output externally — to a client, customer, or auditor — what changes about how you check it?Stakeholder CommunicationMid–leadership

Your team AI usage opinionsWhat's a strong opinion you have about how people on your team should and shouldn't use AI?JudgmentSenior–leadership

Handling AI-generated content errorsHow would you triage a situation where someone on your team shipped AI-generated content with a factual error?Error HandlingSenior–leadership

Onboarding teammates to AI toolsWalk me through how you'd onboard a brand-new teammate to the AI tools your team relies on.Tool UsageMid–leadership

Evaluating AI fit for othersIf a teammate said 'AI just doesn't work for what I do', what would you actually want to know before agreeing or disagreeing?JudgmentMid–leadership

New AI capability you've adoptedWhat's an AI capability that didn't exist a year ago that you've integrated into how you work?UpskillingEntry–leadership

Disclosing nuanced AI involvementHow do you handle the question 'did AI write this?' when you've used AI as part of the work but the answer is more nuanced than yes or no?Stakeholder CommunicationEntry–leadership

Great versus frustrating AI daysWhat does a great AI-assisted day look like for you? What does a frustrating one look like?Tool UsageEntry–leadership

Human in the loopWhat's a specific case where you'd want a human in the loop even if the AI's accuracy is already higher than the human's?JudgmentSenior–leadership

Fixing versus redoing AI outputHow do you decide whether to fix an AI's output, redo it, or throw it away and start from scratch?Error HandlingMid–leadership

Skills developed through AI workWhat's a non-obvious skill you've found yourself developing because of how often you work with AI?UpskillingMid–leadership

Time savings versus workflow changeHow do you tell the difference between an AI tool that's saving you time and one that's just making your work feel different?JudgmentMid–leadership

Evaluating new AI toolsWhen a new AI tool launches in your space, what's your usual evaluation routine?Tool UsageMid–leadership

Teaching others your AI practicesIf your manager asked you to teach a 30-minute session on how you use AI today, what would you cover and what would you leave out?Stakeholder CommunicationMid–leadership

AI use case you'd avoidWhat's a use of AI you've seen elsewhere that you think is a bad idea, and what would you do instead?JudgmentSenior–leadership

Custom solution versus AI toolDescribe a situation where you had to decide whether to build a custom solution or use an AI tool for a recurring problem. What factors drove your decision?JudgmentMid–leadership

AI slowing you downTell me about a time when using an AI tool actually slowed you down or added friction to your process. What did you learn?Tool UsageEntry–senior

Documenting AI-assisted workHow do you decide what level of detail to include in documentation when AI helped you produce the work?Stakeholder CommunicationEntry–senior

Essential versus optional AI trainingWalk me through how you would onboard a new team member on the AI tools your team relies on. What's essential versus optional?UpskillingMid–leadership

Validating AI output before productionDescribe a time when you had to validate AI-generated code or analysis before putting it into production. What was your process?Error HandlingMid–leadership

Disclosing AI use to stakeholdersTell me about a time when a client or stakeholder asked if you used AI on their project. How did you respond?Stakeholder CommunicationMid–leadership

Skills deepened by AIWhat's one skill you've deliberately invested in learning more deeply because AI now handles the basics?UpskillingEntry–senior

Choosing manual work over AIDescribe a situation where you chose to do something manually even though an AI tool could have done it faster. Why?JudgmentEntry–staff+

Troubleshooting inconsistent AI resultsTell me about a time when you had to troubleshoot why an AI tool was giving you inconsistent or unexpected results.Error HandlingEntry–senior

Balancing experimentation and stabilityHow do you balance experimenting with new AI tools against the risk of disrupting a workflow that already works?JudgmentMid–leadership

Customizing AI prompts or settingsWalk me through a time when you customized or fine-tuned an AI tool's prompts or settings to get better results for your specific use case.Tool UsageEntry–senior

Explaining AI limitations to stakeholdersTell me about a time when you had to explain to a non-technical stakeholder how AI was being used in a project and what its limitations were.Stakeholder CommunicationMid–leadership

Delegating tasks to AIDescribe your process for deciding which parts of a complex task to delegate to AI versus handle yourself.JudgmentMid–staff+

Closing an AI literacy gapTell me about a time when you identified a gap in your AI literacy and took action to close it. What prompted that realization?UpskillingEntry–senior

Managing different AI comfort levelsHow do you handle situations where team members have very different comfort levels or philosophies about using AI tools?Stakeholder CommunicationSenior–leadership

Over-relying on AI toolsDescribe a time when you caught yourself over-relying on an AI tool. What made you realize it, and what did you change?Error HandlingEntry–senior

Tracking AI iterationsWhat's your approach to version control or tracking when AI is iterating on your work? How do you stay oriented?Tool UsageMid–staff+

AI ethics and complianceTell me about a project where the use of AI tools raised ethical or compliance questions. How did you navigate that?JudgmentSenior–leadership

Testing AI output qualityHow do you test whether an AI-generated output is actually meeting the standard you need before you use it?Error HandlingEntry–senior

Distinguishing hype from capabilityWalk me through how you stay current on which AI capabilities are hype versus genuinely useful for your role.UpskillingMid–leadership

Rolling back AI-assisted workDescribe a time when you had to roll back or undo work that was done with AI assistance because it didn't meet requirements.Error HandlingMid–staff+

Setting AI expectations with managersTell me about a conversation where you had to set expectations with your manager about what AI could and couldn't do for your team.Stakeholder CommunicationMid–leadership

Choosing alternatives to AIWhat's a type of work where you've seen colleagues successfully use AI, but you've chosen a different approach? Why?JudgmentEntry–senior

Deciding good enough for AIHow do you decide when an AI tool's output is 'good enough' versus when it needs more human refinement?JudgmentEntry–staff+

Integrating AI without losing prioritiesDescribe how you've integrated AI tools into your daily routine without letting them dictate your priorities or attention.Tool UsageEntry–senior

Trusting autonomous AI agentsTell me about the most autonomous task you've handed to an AI agent — something that ran several steps on its own without you in between. How did you decide it was safe to let it go?Agentic WorkflowsMid–leadership

Placing checkpoints in agent tasksWhen you let an AI agent run a multi-step task, where do you put your checkpoints — and why there and not elsewhere?Agentic WorkflowsMid–leadership

Agent failures and cleanupDescribe a time an AI agent went off the rails partway through a task. How far did it get before you caught it, and what did the cleanup look like?Agentic WorkflowsMid–leadership

End-to-end agent tasksWhat kinds of tasks do you now hand to an agent end to end that, a year ago, you'd have done one step at a time yourself?Agentic WorkflowsEntry–leadership

Prompts versus agentic workflowsHow do you decide between writing one careful prompt and setting up an agent that can use tools and iterate on its own?Agentic WorkflowsMid–leadership

Scoping agent permissionsWhen an agent can act on real systems — send messages, edit files, run code, move money — how do you scope what it's actually allowed to touch?Agentic WorkflowsSenior–leadership

Reviewing agent execution tracesWhat's your routine for reviewing the trace of what an agent actually did, versus just looking at whether the final output is right?Agentic WorkflowsMid–leadership

Connecting AI to custom systemsTell me about a time you connected an AI tool to your own data or systems — a custom project, a knowledge base, an integration, an MCP server. What did you have to get right?Agentic WorkflowsMid–leadership

Correcting agent assumptionsHow do you keep an agent on track when the task is ambiguous and it keeps making reasonable-but-wrong assumptions?Agentic WorkflowsMid–leadership

Guarding against agent failuresWhat's the failure mode you worry about most when an agent can take actions on its own, and how do you actually guard against it?Agentic WorkflowsSenior–leadership

Limiting agent resource spendingHow do you decide how much time, money, or tokens to let an agent burn attempting something before you step in and do it yourself?Agentic WorkflowsMid–leadership

Setting up recurring agent tasksWalk me through how you'd set up an agent to handle a recurring task for your team. What guardrails would you build in from day one?Agentic WorkflowsSenior–leadership

Agents as wrong toolWhen have you decided an agent was the wrong tool for something, even though it technically could have done it?Agentic WorkflowsMid–leadership

Reviewing agent workHow has working with agents changed what 'reviewing the work' means for you day to day?Agentic WorkflowsMid–leadership

Working without visible agent stepsWhat do you do differently when you can't see an agent's intermediate steps versus when the whole trace is visible to you?Agentic WorkflowsMid–leadership

Tracing an agent's decision pathTell me about a time you had to figure out why an agent took the particular path it did. How did you trace it back?Agentic WorkflowsMid–staff+

Handling unpredictable agent failuresHow do you handle an agent or tool that succeeds most of the time but fails unpredictably on a small slice of cases?Agentic WorkflowsMid–leadership

Most complex AI workflow builtWhat's the most complex AI workflow you've built or relied on, and what actually held it together?Agentic WorkflowsMid–leadership

Building trust for unsupervised agentsHow do you build enough trust in an agent's output to eventually let it run unsupervised on something that matters?Agentic WorkflowsMid–leadership

Deciding on agent confirmation requestsWhen an agent pauses to ask you for confirmation mid-task, how do you decide whether to approve, redirect, or take over entirely?Agentic WorkflowsMid–leadership

Task breakdown with agentic stepsHow has the way you break down a task changed now that an agent can carry several steps on its own?Agentic WorkflowsEntry–leadership

Handing over an agentic workflowIf you were handing an important agentic workflow to a teammate to own, what would you insist they understand before they trust it?Agentic WorkflowsSenior–leadership

Avoiding over-engineered automationHow do you avoid the trap of building an elaborate agent for something a five-minute manual task would have handled fine?Agentic WorkflowsMid–leadership

Monitoring agents without micromanagingDescribe how you keep yourself in the loop on an agent's progress without babysitting every step.Agentic WorkflowsMid–staff+

Automation that didn't work outWhat's a task you tried to automate with an agent that just didn't stick? Why not?Agentic WorkflowsEntry–senior

Accountability with agentsAs agents take on more, how do you keep accountability clear — who owns the outcome when an agent did most of the work?Agentic WorkflowsSenior–leadership

Data sensitivity boundariesWhere do you draw the line on what company or client information you'll put into an AI tool, and how did you arrive at that line?Governance RiskEntry–leadership

Stopping for sensitive dataTell me about a time you wanted to use AI for something but the sensitivity of the data made you stop. What did you do instead?Governance RiskMid–leadership

Regulated data in AI workflowsHow do you handle confidential, regulated, or customer data when an AI tool would genuinely help but the rules are strict?Governance RiskMid–leadership

Understanding where prompts goWhat's your actual understanding of where your prompts and data go when you use the AI tools in your workflow?Governance RiskEntry–leadership

Policy clashing with toolsHas your organization's AI policy ever clashed with a tool you wanted to use? How did you handle it?Governance RiskMid–leadership

Teammate pasting sensitive dataWhat would you do if you found a teammate routinely pasting sensitive data into a consumer AI tool?Governance RiskMid–leadership

AI data retention risksHow do you think about the risk that an AI tool leaks, retains, or trains on what you put into it?Governance RiskMid–leadership

Shadow AI on teamsWhat is 'shadow AI' on a team, and how would you handle it if you discovered it was widespread?Governance RiskSenior–leadership

Balancing productivity and data governanceHow do you balance the productivity AI tools give you against the data-governance rules you're required to follow?Governance RiskMid–leadership

Guarding against prompt injectionIf untrusted text — a customer email, a web page, a document — feeds into an AI step in your workflow, how do you guard against it manipulating the output?Governance RiskSenior–leadership

Writing team AI usage policyIf you were asked to write your team's AI usage policy, what are the first three rules you'd put in, and what would you deliberately leave out?Governance RiskSenior–leadership

Balancing AI compliance with experimentationHow do you keep a team compliant with AI rules without killing the experimentation that makes the tools worth having?Governance RiskSenior–leadership

AI output and IP ownershipWhat's your stance on AI output and intellectual property — who owns it, and what can go wrong?Governance RiskMid–leadership

Tempting but risky AI usesIn your field, what's a use of AI that's tempting and effective but you'd consider it a compliance or liability risk to actually do?Governance RiskMid–leadership

Vetting AI tools before useHow do you check whether a given AI tool is approved and safe to use before you start relying on it?Governance RiskEntry–senior

Blocking AI on risk groundsTell me about a time you had to say no to an AI use that others were excited about, on risk or governance grounds.Governance RiskMid–leadership

Audit trails with AI-assisted decisionsHow do you handle the audit trail when AI is part of producing a decision or a record that might be reviewed later?Governance RiskMid–leadership

Pushing back on unsafe deploymentHow would you respond if leadership wanted to deploy AI in a decision that affects people's outcomes — credit, hiring, care, benefits — faster than you thought was safe?Governance RiskSenior–leadership

AI hygiene practices worth spreadingWhat's a piece of AI hygiene — a habit or a check — that you wish more people on your team practiced?Governance RiskMid–leadership

Preventing copyright leakage from AIHow do you keep AI from quietly introducing someone else's copyrighted or licensed material into your work?Governance RiskEntry–senior

Client confidentiality and AI toolsWhen client confidentiality is on the line, how do you decide whether an AI tool can touch the engagement at all?Governance RiskMid–leadership

Bias and fairnessHow do you think about bias or fairness in the AI tools you use to make or support decisions?Governance RiskMid–leadership

Allowed versus wise AI useWhat's your read on the gap between what's technically allowed with AI in your org and what's actually wise? Where do they diverge?Governance RiskSenior–leadership

Norms for disclosing AI useWhat norms does your team have about disclosing AI use, and do you actually think they're the right ones?Collaboration NormsMid–leadership

Crediting AI-assisted workHow should AI-assisted work be credited or attributed when a team ships it together?Collaboration NormsMid–leadership

AI output in performance reviewsHave you been in a review where AI-generated work clearly changed the dynamic — what to praise, what to nitpick? What happened?Collaboration NormsMid–leadership

Feedback on unrefined AI workHow do you give feedback on work a colleague obviously generated with AI but didn't bother to refine?Collaboration NormsSenior–leadership

Addressing misattributed AI outputHow do you handle it when someone presents AI output as their own careful, original work?Collaboration NormsMid–leadership

Using AI live in meetingsWhat's the right etiquette for using AI live in a meeting — drafting, summarizing, or fact-checking in real time?Collaboration NormsMid–leadership

Developing junior staff with AIHow do you mentor or develop junior people when AI can already do much of what they used to learn by doing?Collaboration NormsSenior–leadership

Early-career reliance on AIWhat worries you, if anything, about early-career people leaning on AI before they've built the underlying skill?Collaboration NormsSenior–leadership

Maintaining quality across different toolsHow do you keep a shared standard of quality when everyone on the team is using different AI tools in different ways?Collaboration NormsMid–leadership

Productivity comparisons with AIHow do you handle the productivity-comparison problem — when AI makes it look like one person is shipping far more than another?Collaboration NormsSenior–leadership

Handing off AI-assisted workWhen you hand off AI-assisted work to someone downstream, what do you make sure travels with it?Collaboration NormsMid–leadership

Building honesty about AI mistakesHow do you build a team culture where people are honest about AI mistakes instead of hiding them?Collaboration NormsSenior–leadership

Collaboration with AI assistantsHow has pair-working or collaborating changed for you when one or both people have an AI assistant in the loop?Collaboration NormsMid–staff+

Resolving disagreement on AI trustIf two teammates disagree on how much to trust an AI tool for shared work, how do you get them to a workable norm?Collaboration NormsSenior–leadership

Reviewing AI-assisted workWhen you review something a teammate made with AI, what do you spend your attention on now versus two years ago?Collaboration NormsEntry–senior

Preserving individual voice with AIHow do you keep AI from flattening everyone's work into the same voice or the same default approach?Collaboration NormsMid–leadership

Building healthy team AI habitsWhat's a healthy team habit around AI you've either started or wish you could start?Collaboration NormsMid–leadership

Raising the floor for everyoneHow do you make sure AI tools raise the floor for the whole team rather than just widening the gap between power users and everyone else?Collaboration NormsSenior–leadership

Choosing between AI toolsHow do you decide which AI model or tool to reach for now that several exist with genuinely different strengths?JudgmentMid–leadership

Trading speed for capabilityWhen is a slower, more expensive, more capable model worth it, and when is the cheap fast one clearly fine?JudgmentMid–leadership

Deciding without chasing trendsThe 'best' tool for your work keeps changing every few months. How do you make decisions without constantly chasing the newest thing?JudgmentMid–leadership

Real productivity versus feeling modernHow do you separate real AI-driven productivity from just feeling busy and modern with AI?JudgmentMid–leadership

Responding to do-more-with-less pressureLeadership says 'use AI to do more with less.' How do you respond honestly without either over-promising or stonewalling?JudgmentSenior–leadership

When AI use is irresponsibleWhere in your field do you think it's genuinely irresponsible to use AI right now, even though it's allowed and would save time?JudgmentSenior–leadership

Almost-right answers that are dangerousWhat's a task where AI gives you a fast answer that's almost right, and why is 'almost right' actually dangerous there?JudgmentEntry–staff+

Trusting your judgment over AIHow do you avoid letting an AI's confident, fluent answer talk you out of your own correct judgment?JudgmentMid–leadership

Deciding among many AI optionsWhen AI makes it cheap to generate ten options, how do you avoid drowning in choices and actually decide?JudgmentMid–leadership

Understanding versus trusting AI workHow do you tell whether you genuinely understand AI-assisted work well enough to defend it, versus just trusting it looks right?JudgmentMid–leadership

Choosing human accountability over speedWhat's a place where you've deliberately kept a human slower-but-accountable instead of an AI faster-but-opaque, and why?JudgmentSenior–leadership

Trusting AI in unfamiliar domainsHow do you decide whether to trust an AI's answer in a domain where you can't easily check it yourself?JudgmentMid–leadership

Focusing on the hard partWhen AI handles the easy 80% of a task instantly, how do you make sure the hard 20% still gets your full attention?JudgmentMid–leadership

Hidden costs of popular shortcutsWhat's an AI shortcut that's popular in your field that you think is actually a bad trade? What's the hidden cost?JudgmentEntry–senior

Custom AI versus off-the-shelf toolsHow do you decide when a problem deserves a custom AI setup versus an off-the-shelf tool versus no AI at all?JudgmentSenior–leadership

Keeping skills sharp with delegationHow do you keep your own skills sharp on tasks you now mostly delegate to AI, so your judgment doesn't atrophy?JudgmentMid–leadership

Overriding AI recommendationsWhat's a decision you made recently where AI gave you input, and you went against it? How did you weigh that?JudgmentMid–leadership

Knowing when to stop iteratingHow do you know when you've used AI enough on a task and adding more would just be polishing or procrastinating?JudgmentEntry–senior

Balancing trust and skepticismAs AI capability jumps every few months, how do you avoid both over-trusting it today and under-using it out of old habit?JudgmentSenior–leadership

Surprising AI use casesWhat's a task you assumed AI couldn't help with that turned out to be a great fit — what changed your mind?JudgmentMid–leadership

Trusting AI over human expertsWhen an AI tool and a trusted human expert disagree, how do you decide whom to believe?JudgmentMid–staff+

Speed without directionHow do you guard against AI making you faster at producing the wrong thing — speed without direction?JudgmentMid–leadership

Your current AI setupWhat does your AI setup actually look like right now — which tools, for which parts of your day?Tool UsageEntry–leadership

Recent automation winsWhat's something you automated in the last month that you used to do entirely by hand?Tool UsageEntry–leadership

Using AI for thinking workHow do you use AI for the thinking parts of your work — framing, deciding, sense-making — not just the producing parts?Tool UsageMid–leadership

Your most-used promptShow me your most-used prompt, template, or workflow and explain why it's earned a permanent place.Tool UsageMid–leadership

Using voice, images, screenshotsHow do you use voice, screenshots, or images with AI tools, if at all? What did that unlock?Tool UsageEntry–senior

Small AI habitWhat's a small AI habit that's quietly made a big difference in your work?Tool UsageEntry–leadership

Organizing reusable promptsHow do you keep a library of prompts, contexts, or reusable setups so you're not rebuilding them every time?Tool UsageMid–leadership

Giving AI useful contextWalk me through how you give an AI tool the context it needs to be useful on your specific work, not generic.Tool UsageMid–leadership

Turning tasks into routinesWhat's a task you do so often that you've turned your AI approach to it into a repeatable routine?Tool UsageEntry–senior

General versus specialized toolsHow do you decide whether to use a general AI assistant or a specialized tool built for your exact task?Tool UsageMid–leadership

Surprising AI use casesWhat's a way you use AI that a colleague was surprised by — something not in the obvious playbook?Tool UsageEntry–leadership

Working around context limitsHow do you handle the context limits of AI tools when your real task is bigger than what fits in one go?Tool UsageMid–leadership

Tracking versions and changesHow do you keep track of versions and changes when AI is rapidly iterating on a piece of work with you?Tool UsageMid–staff+

Matching style and formatWhat's your approach to getting an AI tool to match a specific style, standard, or format you need repeatedly?Tool UsageMid–leadership

Stuck versus preparedHow do you use AI when you're stuck or facing a blank page, versus when you already know what you want?Tool UsageEntry–senior

Chaining multiple AI toolsTell me about a time you chained two or more AI tools together to get something done. What was the seam between them?Tool UsageMid–leadership

Building reusable AI workflowsWhat's a task where you've stopped writing prompts from scratch and instead built something reusable? Walk me through it.Tool UsageEntry–leadership

Setup investment versus defaultsHow do you decide how much to invest in setting up an AI tool well versus just using it out of the box?Tool UsageMid–leadership

Your most critical AI toolWhat's the AI tool or feature you reach for most, and what would actually break in your week if it disappeared tomorrow?Tool UsageEntry–senior

AI in shared workflowsHow do you fold AI into a workflow that involves other people or systems without creating a mess for them?Tool UsageMid–leadership

Verifying large AI outputsWhen AI generates a large chunk of work in one shot — too much to read line by line — how do you decide what to actually verify?Error HandlingMid–leadership

Reviewing AI-generated workHow do you review work the AI produced that you didn't write yourself, piece by piece? What's your method?Error HandlingMid–leadership

Volume outrunning accountabilityAI lets you produce far more than before. How do you keep the volume from outrunning your ability to stand behind it?Error HandlingMid–leadership

Avoiding rubber-stamping polished AI outputHow do you avoid rubber-stamping AI output just because it looks polished and confident?Error HandlingEntry–senior

Catching subtle, plausible AI errorsWhat's your method for catching the subtle AI errors — the plausible, well-formatted ones — rather than the obvious howlers?Error HandlingMid–leadership

Handling unverifiable AI citationsHow do you handle citations or sources when an AI gives you facts you can't immediately verify?Error HandlingMid–leadership

Cost of AI fabricationHave you been burned by an AI fabricating something that sounded completely real? What did it cost, and what changed after?Error HandlingEntry–senior

Building repeatable AI output evaluationHave you ever built a repeatable way to test whether an AI's output is actually good — an eval, a checklist, a test set? Walk me through it.Error HandlingMid–leadership

Measuring AI output quality objectivelyHow do you measure whether an AI tool is genuinely producing good output beyond it just feeling right?Error HandlingMid–leadership

Buried error detectionWhen AI output is mostly right but has one quiet error buried in it, how do you make sure that error doesn't slip through?Error HandlingMid–staff+

Recognizing wrong AI answers intuitivelyWhat's your tell that an AI answer is probably wrong, even before you've checked it?Error HandlingEntry–senior

Tracking patterns in AI mistakesHow do you keep a record of AI mistakes you've caught so the same class of error doesn't keep getting through?Error HandlingMid–leadership

Distinguishing one-off from systemic errorWhen you catch an AI being wrong, how do you figure out whether it's a one-off or a pattern you need to design around?Error HandlingMid–leadership

Pre-customer quality gateHow would you set up a check so that AI-assisted work on your team gets caught before it reaches a customer, not after?Error HandlingSenior–leadership

Acting on suspected AI errorsWhat do you do when you suspect an AI output is wrong but you can't quite prove it yet?Error HandlingEntry–senior

Changed verification processHow has the way you double-check work changed now that more of your inputs come from AI rather than from you?Error HandlingMid–leadership

Disclosing AI involvement to customersHow do you talk to a customer or client about AI being part of the work or product they're paying for?Stakeholder CommunicationMid–leadership

Reassuring clients about AI-assisted workHow do you reassure someone worried that 'the AI did the work I'm paying a human for'?Stakeholder CommunicationMid–leadership

Managing executive AI expectationsHow do you set realistic expectations when an executive is over-excited about what AI can do for your team?Stakeholder CommunicationSenior–leadership

Pushing back on premature deploymentHow do you push back when leadership wants to deploy AI somewhere you think it's premature or risky?Stakeholder CommunicationSenior–leadership

Explaining AI failuresHow do you explain to a non-technical stakeholder why an AI tool got something wrong, without either excusing it or overcomplicating it?Stakeholder CommunicationMid–leadership

Proving AI effectiveness to skepticsHow would you prove to a skeptical leader that AI is actually making your team more effective — not just busier?Stakeholder CommunicationSenior–leadership

Declining unsuitable AI tasksWhen someone asks you to 'just have AI do it,' how do you respond if the task isn't a good fit?Stakeholder CommunicationMid–leadership

Communicating AI uncertaintyHow do you communicate the limits and uncertainty of an AI-assisted result to people who'll make decisions on it?Stakeholder CommunicationMid–leadership

Addressing job displacement fearsHow do you handle a stakeholder who's afraid AI is going to take their job, when it comes up in your work?Stakeholder CommunicationEntry–senior

Skeptical team adoptionHow do you communicate an AI policy or new AI workflow to a team that's split between enthusiasts and skeptics?Stakeholder CommunicationSenior–leadership

Distinguishing machine from human judgmentWhen you present AI-assisted analysis, how do you make clear which parts are machine-generated and which are your own judgment?Stakeholder CommunicationMid–leadership

Defending AI-assisted work against biasHow do you respond when a stakeholder dismisses solid AI-assisted work simply because AI was involved?Stakeholder CommunicationSenior–leadership

Realistic cost claimsHow do you keep stakeholders from assuming AI made something faster or cheaper than it actually did?Stakeholder CommunicationMid–leadership

Regulated AI disclosureIn a regulated or high-trust setting, how do you disclose AI involvement to the people affected by the output?Stakeholder CommunicationMid–leadership

Your value beyond AIAs AI gets better at the core of your job, where are you staking your value — what do you bring that the tool doesn't?UpskillingMid–leadership

Skills AI makes more valuableWhat part of your skill set do you think AI makes more valuable, not less?UpskillingEntry–senior

Getting ahead of AI automationHonestly, which parts of your job do you think AI could automate soon — and how are you getting ahead of that?UpskillingMid–leadership

Skills you deliberately keep sharpWhat's a skill you're deliberately NOT outsourcing to AI because you want to keep it sharp?UpskillingEntry–senior

Learning when tools keep changingHow has 'getting better at your job' changed for you in an era where the tools change faster than you can master them?UpskillingMid–leadership

Most useful AI lesson learnedWhat's the most useful thing you've learned about working with AI in the last few months, and how did you learn it?UpskillingEntry–leadership

Separating AI hype from substanceHow do you cut through the noise to tell which new AI capabilities are hype and which actually matter for your work?UpskillingMid–leadership

Building AI learning habitsWhat's a habit you've built to keep learning AI tools without it eating all your time?UpskillingMid–senior

Getting teams to adopt AIHow do you help your team level up on AI without turning it into yet another mandate that people resent?UpskillingSenior–leadership

Changing your mind about AIWhat's something you believed about AI a year ago that you've since changed your mind on?UpskillingEntry–senior

Learning a team's AI normsIf you joined a new team tomorrow, how would you get up to speed on the AI tools and norms they already use?UpskillingEntry–leadership

Spot-checking AI transaction categorizationYour close process uses an AI tool to auto-categorize thousands of transactions into GL accounts before you review the trial balance. Walk me through which categorizations you’d spot-check versus accept outright, and why.Tool UsageEntry–leadership

When you skip AI deliberatelyTell me about a task in your close or reporting cycle where you deliberately don’t use AI assistance at all, even though a tool exists for it. What makes that task different?Tool UsageMid–leadership

Catching AI lease classification errorsAn AI assistant drafts a journal entry to reclassify a lease payment and cites an ASC 842 treatment that doesn’t match your company’s actual lease terms. How would you catch that before it posts?JudgmentEntry–leadership

Verifying AI-drafted variance narrativesYou’re reviewing an AI-drafted variance analysis narrative for the board deck that attributes a margin swing to a plausible-sounding cause. Walk me through how you’d verify that explanation before it goes out under your name.JudgmentSenior–leadership

Explaining AI sampling to auditorsAn external auditor asks how your team flagged the sample of transactions under review. How do you explain that an AI anomaly-detection tool selected part of the sample, in a way that holds up to audit scrutiny?Stakeholder CommunicationEntry–leadership

Explaining AI in tax adviceA client asks whether the tax position you recommended was “just from ChatGPT.” Walk me through how you’d respond, and how you actually used AI in reaching that position.Stakeholder CommunicationMid–leadership

Responding to AI reconciliation errorsAn AI-assisted reconciliation tool mismapped a vendor account and it wasn’t caught until after the quarter closed and financials were distributed internally. Walk me through what you’d do next.Error HandlingEntry–leadership

Managing AI audit testing failuresYour team’s AI-assisted audit testing tool cleared a revenue recognition sample that turns out to include a bill-and-hold arrangement it wasn’t trained to flag, and this surfaces after sign-off. As the reviewing manager, walk me through your response.Error HandlingSenior–leadership

Learning new AI toolsNew AI-assisted research tools show up in your firm’s tech stack every few months. How do you decide which ones are worth learning, and how do you build that skill without slowing down billable work?UpskillingEntry–leadership

Training associates on AI draftingWalk me through how you’d train a first-year associate to use an AI drafting tool for audit workpapers without them losing the underlying judgment skills the job actually requires.UpskillingMid–leadership

Checkpoints for automated reconciliationYou set up an AI agent to pull data from the ERP, the bank feed, and the sub-ledger to complete a bank reconciliation end to end. What checkpoints do you build in before it can post the reconciling entries itself?Agentic WorkflowsEntry–leadership

Agentic escalation thresholdsYou’re rolling out an agentic tool that can draft and route journal entries for the whole close cycle, not just one account. Walk me through how you’d set the thresholds for what it can execute versus what it must escalate to a human.Agentic WorkflowsSenior–leadership

Unreleased statements in public AIBefore earnings are released, a teammate suggests dropping a draft of the unreleased financial statements into a public AI chatbot to get help tightening the language. Walk me through how you’d handle that request.Governance RiskEntry–leadership

Updating SOX controls for AIYour company’s SOX control matrix has to be updated now that an AI tool touches part of the close process. Walk me through what you’d document to show the control is still operating effectively.Governance RiskMid–leadership

Review process for AI-drafted footnotesYour team starts using AI to draft first-pass footnote disclosures for the financial statements. What review norm would you put in place so no AI-drafted language reaches the filing without a second set of eyes verifying it against the standard?Collaboration NormsEntry–leadership

Inconsistent AI answers across teamsTax and audit both use the same AI research assistant but sometimes get different answers to a similar accounting question. How would you set a team norm so those inconsistencies get caught before they reach a client or the financial statements?Collaboration NormsMid–leadership

Creative brief allocationWalk me through which parts of a creative brief you’d hand to a generative AI copy tool for first-draft headlines and taglines, and which parts of concepting you’d insist on doing yourself.Tool UsageEntry–leadership

Overriding AI media-buying recommendationsDescribe how you use an AI-assisted media-buying platform to set programmatic bid strategy, and where you still override its recommendations with your own read of the account.Tool UsageEntry–leadership

Verifying unconfirmed statistics from AIAn AI research assistant pulled a market-share statistic into a creative brief that you can’t find in any of your syndicated data sources. Walk me through what you do before that number goes into the deck.JudgmentEntry–leadership

Catching discriminatory AI audience segmentsYour AI targeting tool has generated an audience segment that performs well but skews heavily toward a protected demographic in a way that could read as discriminatory. How do you catch this and what do you do next?JudgmentMid–leadership

Disclosing AI-generated voice to clientsA client asks whether the actor’s voice in the radio spot is a real recording or an AI-generated clone. Walk me through how you’d answer, including what you’d have documented beforehand.Stakeholder CommunicationEntry–leadership

Explaining AI targeting to clientsA client pushes back after learning that an AI tool, not a human planner, set most of the audience targeting parameters on their campaign. How do you explain the role AI played and rebuild their confidence in the plan?Stakeholder CommunicationEntry–leadership

Brand safety violationAn AI-optimized programmatic buy placed your client’s ad next to content that violates their brand-safety guidelines, and it ran for two days before anyone caught it. Walk me through your response, from the client conversation to the process fix.Error HandlingMid–leadership

Handling unsubstantiated AI-generated claimsYou discover, after a campaign has launched, that an AI copywriting tool generated a claim that can’t be substantiated and may draw regulatory scrutiny. Walk me through how you handle it in the next 24 hours.Error HandlingEntry–leadership

Teaching AI toolsHow do you keep your own skills current on generative AI creative and media-buying tools as they change every few months, and how do you decide what’s worth teaching the rest of the team?UpskillingEntry–leadership

Training teams to prompt effectivelyWalk me through how you’d train a creative team to write prompts that consistently produce on-brand copy from a generative AI tool, rather than generic output that needs heavy rewriting every time.UpskillingEntry–leadership

A/B test automationDescribe how you’d set up an AI agent to run and report on a multistep A/B creative test — variant generation, budget allocation, and result summary — and where you’d insert a human checkpoint before it acts.Agentic WorkflowsEntry–leadership

Budget reallocation guardrailsAn AI agent has autonomy to pause underperforming ad variants and reallocate budget across channels in real time. Walk me through the guardrails you’d put around that agent before giving it live client budget.Agentic WorkflowsSenior–leadership

AI imagery copyright riskWalk me through how you assess the copyright and licensing risk of using AI-generated imagery in a client campaign before it ships, including what happens if the training data provenance is unclear.Governance RiskEntry–leadership

Evaluating AI lookalike audience complianceA client wants to feed their first-party customer data into an AI tool to generate more precise lookalike audiences. Walk me through how you’d evaluate whether that’s compliant before agreeing to it.Governance RiskEntry–leadership

Creative asset reviewWhat review standard would you put in place so that no AI-generated creative asset reaches a client without a human checking it for accuracy, brand fit, and rights clearance?Collaboration NormsEntry–leadership

Legal review alignmentHow would you get creative, media, and legal teams aligned on when an AI-assisted claim or visual needs to be flagged for legal review before a brief moves forward?Collaboration NormsMid–leadership

Verifying AI flight planning recommendationsYour flight-planning system uses AI to recommend optimal routes and fuel loads based on winds, weather, and traffic. Walk me through which parts of that recommendation you’d verify yourself before releasing the dispatch, and which part of the job you’d never hand to the tool.Tool UsageEntry–leadership

Predictive maintenance alertYour engine health monitoring system uses AI to flag a component trending toward failure days before any fault code would trip. Walk me through what you do with that alert versus what you’d only do once it’s confirmed through the certified inspection process.Tool UsageEntry–leadership

Disagreeing with AI turbulence forecastsAn AI turbulence-forecasting tool shows a smooth ride ahead, but the captain’s read of the cloud tops and pilot reports from earlier flights says otherwise. How do you think about a disagreement like that, and what does it tell you about the tool’s blind spots?JudgmentEntry–leadership

NOTAM summary misses runway closureAn AI tool summarizes the NOTAM package for a flight into a two-paragraph brief, and it’s usually accurate — but on this flight it dropped a runway closure buried in the raw data. Walk me through how you’d catch that before the crew briefs off the summary.JudgmentEntry–leadership

Regulator AI disclosureYou used an AI tool to help correlate flight data recorder trends across a fleet while building the causal-factors section of an internal safety investigation. How would you disclose that AI involvement to the regulator reviewing your report, and what would you make sure is independently verified first?Stakeholder CommunicationMid–leadership

Communicating AI crew reassignmentYour crew-scheduling system uses AI to propose reassignments during an irregular-operations day, and one proposal keeps a captain just inside legal rest minimums. How do you communicate that change to the crew member, and what do you disclose about how the system arrived at it?Stakeholder CommunicationEntry–leadership

Responding to AI weight-and-balance errorA load-planning tool’s AI-assisted weight-and-balance calculation turns out to be wrong after the aircraft has already pushed back. Walk me through what happens next, from the immediate response through how you’d prevent a repeat.Error HandlingEntry–leadership

AI draft contained wrong partYou used an AI tool to help draft the maintenance logbook entry closing out a repair, and a week later someone notices the entry references the wrong part number. Walk me through how you’d handle that discovery and what it changes about how you use the tool going forward.Error HandlingEntry–leadership

Keeping dispatchers sharp without AIYour operation is rolling out an AI flight-planning assistant that most dispatchers will lean on daily. How do you make sure the team stays sharp enough to plan a flight by hand the day the system is down?UpskillingEntry–leadership

Closing squawk without panel confirmationA junior mechanic wants to close out a squawk based on what the AI diagnostic tool recommends, without pulling the panel to confirm. How do you coach them on when the tool’s read is enough and when it isn’t?UpskillingEntry–leadership

Agentic passenger rebooking checkpointsYou’re introducing an agentic tool that can rebook and reroute passengers across a multi-hub disruption without a human touching each case. What decisions do you let it make on its own, and where do you insert a mandatory human checkpoint?Agentic WorkflowsSenior–leadership

Reviewing AI-drafted safety reportsAn agentic tool can now draft the initial hazard report from a raw safety submission, tagging severity and suggested corrective actions before a human reviews it. Walk me through what you check before that draft goes into the safety management system.Agentic WorkflowsEntry–leadership

Approving vendor access to flight dataA predictive-maintenance vendor wants direct API access to your fleet’s flight data recorder and FOQA streams to train their models. What do you need in place before you’d approve that, and what would make you say no?Governance RiskSenior–leadership

AI advisory in cockpit checklistLeadership wants an AI advisory layer added to the cockpit checklist workflow to speed up abnormal-procedure response. Where do you draw the line on what that tool is allowed to suggest versus what stays locked to the certified procedure, no exceptions?Governance RiskEntry–leadership

Dispatch release independent verificationWalk me through the norm you’d want on your dispatch desk for when a second dispatcher has to independently check an AI-assisted release before it goes out, versus when one person’s sign-off is enough.Collaboration NormsEntry–leadership

Review standard for AI-drafted justificationYour engineering group uses an AI tool to draft the technical justification for a continuing-airworthiness change, like an alternate means of compliance. What review standard do you set before that justification can move to the certifying engineer’s signature?Collaboration NormsMid–leadership

AI tool outside data-residency boundaryYour bank’s data-residency policy blocks account-level customer data from leaving the country, but the AI drafting tool your team wants to trial only runs on a vendor’s cloud outside that boundary. Walk me through how you’d get useful AI help onto this task without breaking the policy.Governance RiskEntry–leadership

AI anomaly detection model governanceYou’ve built an AI tool that helps loan officers flag anomalies in financial statement spreads before they go to underwriting. Walk me through how you’d get it through model risk management under SR 11-7 before it touches live deals.Governance RiskMid–leadership

Loan summary misses subordination clauseAn AI tool summarized a loan agreement for you and the summary left out a subordination clause buried in an exhibit. What’s your process for catching gaps like this before the summary gets relied on downstream?JudgmentEntry–leadership

Tuning AI transaction monitoring alertsYour transaction monitoring system uses an AI model to score alerts, and it keeps flagging a legitimate import-export client’s wire pattern as suspicious. How do you decide whether to tune the model, document an exception, or escalate — and who signs off?JudgmentEntry–leadership

AI drafting in KYC filesWalk me through which parts of a KYC due diligence file you’d let an AI drafting tool touch, and which parts you’d insist stay entirely manual.Tool UsageEntry–leadership

AI copilot in regulatory reportingYou’re using an AI copilot to help draft commentary for a regulatory report like a Call Report schedule or a CCAR submission. Where does the copilot’s role end and your ownership of the numbers begin?Tool UsageMid–leadership

Disclosing AI in credit memoYou used AI to help draft sections of a credit memo. Does the credit committee need to know that, and how would you disclose it without sounding like you’re excusing the work?Stakeholder CommunicationEntry–leadership

Customer chatbot disclosure and boundariesYour bank is rolling out an AI chatbot to answer customer questions about account terms and fees. What do you need to get right about how it discloses that it’s AI, and what happens when a customer’s question edges into advice the bot shouldn’t give?Stakeholder CommunicationEntry–leadership

Missed suspicious activity patternA regulatory exam surfaces a suspicious activity pattern that your AI-assisted transaction monitoring system should have flagged months earlier but didn’t. Walk me through what you do in the first week, and what changes afterward.Error HandlingSenior–leadership

Remediating AI-generated loan document errorAn AI drafting tool generated a loan document with an incorrect interest rate index, and the loan was already booked before anyone caught it. Walk me through your remediation from the moment you discover it.Error HandlingEntry–leadership

Team resistance to AI spreadingHalf your underwriting team is resistant to using AI-assisted financial statement spreading, worried it’ll be blamed if the AI gets something wrong. How do you build their capability and trust at the same time?UpskillingMid–leadership

Adopting AI fraud detection safelyAI fraud detection capabilities are moving faster than regulatory guidance can keep up with. How do you personally stay current, and how do you decide what’s ready to bring into the team’s actual workflow?UpskillingEntry–leadership

Agentic AI credit memo oversightYou’re piloting an agentic AI tool that can pull a borrower’s financial documents, calculate ratios, and draft the first version of a credit memo end to end. Walk me through the oversight checkpoints you’d design before letting it run unsupervised on live deals.Agentic WorkflowsSenior–leadership

Agentic fraud alert closure boundariesThere’s a proposal to let an agentic AI system automatically close out low-risk fraud alerts without a human reviewing each one, to cut the backlog. Where would you draw the line on what it’s allowed to close on its own?Agentic WorkflowsEntry–leadership

Reviewing AI-assisted credit memosSet the norm for your team: when a credit memo has AI-assisted sections, what does the second reviewer need to check differently than on a fully manual memo?Collaboration NormsEntry–leadership

AI drafting adverse action noticesYour team wants to use AI to draft adverse action notices for declined loan applications faster. What review standard would you put in place before any of those go out the door?Collaboration NormsEntry–leadership

AI literature summarization in discoveryWalk me through how you use an AI literature-summarization tool during early target discovery, and where you still insist on reading the primary paper yourself before it enters your target rationale document.Tool UsageEntry–leadership

Selecting protein sequences for synthesisA generative protein-design model proposes ten candidate binder sequences ranked by predicted affinity. Walk me through your process for deciding which ones actually get synthesized and run through binding assays.Tool UsageMid–leadership

Low-confidence loop in structure predictionYou pull a structure prediction with high overall confidence but a low-confidence loop right at the binding interface you care about. How do you decide whether that region is trustworthy enough to guide mutagenesis design?JudgmentEntry–leadership

Overriding AI virtual screening rankingsYour virtual screening pipeline's AI scoring function ranks five hundred compounds, and you only have budget to synthesize and test twenty. Walk me through how you decide which of the model's rankings to trust and which to override with your own chemistry judgment.JudgmentSenior–leadership

AI assistance in regulatory documentsA section of your IND briefing document was drafted with help from an AI writing assistant, then heavily edited by you and reviewed by regulatory affairs. If a reviewer at the agency asked directly how the document was produced, what would you say and why does that matter?Stakeholder CommunicationEntry–leadership

Presenting AI-synthesized recommendationsYou used an AI tool to help synthesize conflicting efficacy signals across three preclinical studies into a single recommendation for a go/no-go committee. How do you present that recommendation so the committee understands what the AI contributed versus what your own scientific judgment added?Stakeholder CommunicationEntry–leadership

AI mislabeled sample IDsAn AI tool auto-summarized a week of raw instrument data into an ELN entry, and you later discover it mislabeled two sample IDs in a way that changed which lane was ‘treated’ versus ‘control.’ Walk me through what you do once you find this, and how far back you have to trace it.Error HandlingEntry–leadership

Recovering from missed AI-classified hitsA compound that an AI hit-classification model scored as inactive turns out, months later during a manual re-review, to have been a real hit that got shelved. Walk me through how you handle the fallout and what changes in your screening pipeline afterward.Error HandlingMid–leadership

Evaluating new protein design toolsNew AI-driven protein design tools seem to appear every few months, each claiming better binder success rates than the last. How do you personally decide which ones are worth learning versus which are hype, and how do you evaluate a new tool before it touches a real program?UpskillingEntry–leadership

Building skeptical team adoptionYou're asked to get a bench science team comfortable using AI-assisted sequence and structure analysis tools, and several senior scientists are skeptical the tools add value over their existing manual workflow. Walk me through how you'd build their trust and competence without mandating adoption.UpskillingSenior–leadership

Checkpoints for agentic literature workflowsYou set up an agentic AI workflow that searches the literature, extracts reported binding affinities across papers, and proposes a ranked shortlist of targets, running with minimal supervision overnight. What checkpoints do you build in so you trust the shortlist that's waiting for you in the morning?Agentic WorkflowsEntry–leadership

Autonomous AI protocol draftingAn agentic AI assistant can now draft an experimental protocol, flag the reagents you'll need, and queue instrument time, all without you touching each step. Where do you draw the line on what it's allowed to do autonomously versus what needs your sign-off first?Agentic WorkflowsEntry–leadership

Cloud AI access to proprietary SAR dataYour team wants to use a cloud-based AI tool to help analyze structure-activity relationship data for a compound series that's core to your IP position. Walk me through the questions you'd need answered before that data goes anywhere near the tool.Governance RiskEntry–leadership

AI in GLP study reportsYou want to introduce an AI drafting tool into the workflow for GLP study reports that eventually support a regulatory submission. What has to be true about that tool and its use before it's actually compliant, not just convenient?Governance RiskEntry–leadership

Sign-off requirements for AI protocolsYour lab has started letting an AI tool draft first-pass experimental protocols. What team norm did you put in place, or would you put in place, for who has to review and sign off before that protocol is actually run on the bench?Collaboration NormsEntry–leadership

Validation before AI go-no-go decisionsBefore an AI-assisted data analysis result is allowed to inform a go or no-go decision on a program, what does your team require happen first, and how did that bar get set?Collaboration NormsEntry–leadership

AI drafting MOC proceduresA process engineer on your team wants to use a generative AI tool to draft the operating-procedure changes after a management-of-change (MOC) review. Walk me through where you’d let the tool draft language and where you’d insist a human writes it from scratch.Tool UsageEntry–senior

Predictive maintenance recommendationsYour plant uses a predictive-maintenance model to flag reactors and rotating equipment at risk of failure. Walk me through how you’d decide which of its recommendations get auto-scheduled into the maintenance queue and which need an engineer’s sign-off first.Tool UsageMid–leadership

Overriding borderline AI assay resultsAn AI-assisted lab tool flags a batch’s assay result as within spec, but the raw chromatogram it summarized looks borderline to you. Walk me through what you do next.JudgmentEntry–mid

AI-recommended reactor temperatureA process-optimization model recommends raising a reactor’s operating temperature to increase yield, and the projected gain looks real. Walk me through how you’d decide whether to trust that recommendation before it touches the actual setpoint.JudgmentSenior–leadership

Disclosing AI use in certificates of analysisA customer’s quality team asks whether AI was involved in generating the certificate of analysis for a shipped lot. Walk me through how you’d answer, and what you’d want documented beforehand.Stakeholder CommunicationMid–senior

Operators adopting AI advisory systemsYou need to explain to plant operators why a new AI-based advisory system is now suggesting adjustments during their shift, when they’ve run the process manually for years. Walk me through that conversation.Stakeholder CommunicationEntry–mid

Correcting an AI-drafted batch record errorAn AI-drafted section of a batch record contained a unit-conversion error that made it past a first review before a second reviewer caught it. Walk me through the corrective action you’d put in place, beyond just fixing that one record.Error HandlingMid–senior

Near-miss from AI troubleshootingA plant experienced a near-miss after operators followed an AI-generated troubleshooting suggestion during a process upset. Walk me through how you’d run the investigation and what changes you’d consider for the tool’s role afterward.Error HandlingSenior–leadership

New hires skipping foundational skillsNew hires on your process engineering team have started leaning on AI assistants for basic mass-and-energy balance calculations before they’ve built the underlying skill themselves. Walk me through how you’d handle that.UpskillingEntry–mid

Building confidence in compliance reviewYou want your quality and EHS staff to get comfortable using AI tools for regulatory-document review, but several are wary given how much REACH and EPA compliance rides on getting the details exactly right. Walk me through how you’d build their confidence.UpskillingMid–leadership

Automating shift handover summaries safelyYou’re considering letting an AI agent pull data from the DCS historian, draft a shift handover summary, and flag anomalies for the incoming operator without a person assembling it first. Walk me through what oversight you’d build in before turning that on.Agentic WorkflowsMid–senior

AI agent adjusting process parametersWalk me through how you’d design guardrails for an AI agent that’s allowed to adjust minor process parameters — within a pre-approved range — on a running unit without waiting for operator confirmation on each change.Agentic WorkflowsSenior–leadership

Third-party AI and proprietary formulationsYour R&D team wants to use a third-party AI tool to help interpret proprietary formulation data ahead of a REACH submission. Walk me through the governance questions you’d need answered before approving that.Governance RiskMid–senior

Chatbot access during spill responsePlant operators want to use a general-purpose AI chatbot to quickly look up SDS guidance during an unexpected spill response. Walk me through whether you’d allow that, and what policy you’d put around it.Governance RiskEntry–mid

Preventing rubber-stamping of AI suggestionsSome engineers on your team review AI-suggested process changes carefully, while others rubber-stamp them because the output looks polished. Walk me through the team norm you’d put in place.Collaboration NormsMid–senior

Setting cross-functional AI sign-off normsWalk me through how you’d set cross-functional norms between process engineering, quality, and EHS for when an AI tool’s recommendation touches all three groups’ sign-off, so it doesn’t stall in disagreement about who owns the final call.Collaboration NormsSenior–leadership

AI-assisted civil design workflowWhich stages of preparing a site grading and drainage plan do you actually run through an AI-assisted CAD or civil design tool, and where do you switch back to manual review before it goes out for permit submittal?Tool UsageEntry–leadership

AI adoption in construction drawingsDescribe a specific task in producing construction drawings — rebar detailing, connection schedules, sheet notes — where you’ve adopted an AI drafting assistant, and one you’ve deliberately kept fully manual. What’s the difference between them?Tool UsageEntry–leadership

Verifying AI code citationsYou ask an AI assistant for the ASCE 7 wind exposure category rule and it gives a confident, specific citation. Walk me through exactly how you’d verify that before it goes into a load calculation package.JudgmentEntry–leadership

Geotechnical summary accuracyAn AI tool summarizes a geotechnical report and gives you the allowable bearing pressure and groundwater depth in one clean paragraph. What’s your process for catching the difference between what the report actually says and what a summary tool might have smoothed over?JudgmentEntry–leadership

Disclosing AI use to clientsA client asks whether the design alternatives you presented for their site layout were generated by AI. How do you answer, and how much do you disclose about the tool’s role versus your own engineering review?Stakeholder CommunicationEntry–leadership

Plan reviewer questions AI calculationsA plan reviewer at the building department asks pointed questions about whether AI tools were used to generate the structural calculations in your submittal package. As the engineer of record, how do you handle that conversation?Stakeholder CommunicationMid–leadership

Field-unbuildable AI-drafted detailA contractor calls mid-construction because a detail on the drawings — produced with AI drafting assistance — doesn’t match what’s buildable in the field. Walk me through what you do in the next 24 hours.Error HandlingEntry–leadership

Outdated code in stamped calculationsYou discover, after a set of structural calculations you stamped were submitted, that an AI-assisted load combination check used an outdated code edition. Walk me through how you handle disclosure, correction, and what changes in your process afterward.Error HandlingMid–leadership

Evaluating AI reliability for engineeringHow do you personally stay current on which AI tools are actually reliable for civil engineering tasks — versus ones that produce plausible-looking but wrong structural or geotechnical output — given how fast this space is changing?UpskillingEntry–leadership

Mentoring judgment over AI checkingYou’re mentoring a junior engineer who’s started leaning on an AI assistant for first-pass structural calculations. How do you make sure they’re building real engineering judgment and not just learning to check AI output?UpskillingEntry–leadership

Automating permit submittal package assemblyDescribe how you’d set up an AI agent to assemble a full permit submittal package — drawings, calculations, code compliance narrative — across multiple documents, and what checkpoints you’d build in before anything gets filed with the jurisdiction.Agentic WorkflowsEntry–leadership

Overriding automated site photo reviewYou’ve set up an AI agent to review daily site inspection photos and flag potential code or safety issues automatically. Walk me through a scenario where you’d want a human to override what the agent flagged, or catch what it missed entirely.Agentic WorkflowsEntry–leadership

AI policy for security-sensitive projectsYour firm works on projects with security-sensitive site data — critical infrastructure, government facilities. What policy would you put in place for which AI tools staff can use on those projects, and how would you enforce it?Governance RiskMid–leadership

Terms of service for AI-generated deliverablesA junior engineer used a generative AI tool to draft a structural framing scheme, and now there’s a question about whether that output can be used in a client deliverable given the tool’s terms of service. How do you think through that?Governance RiskEntry–leadership

Review standards for AI-assisted structural workAs a team lead, how would you set the standard for when AI-assisted structural or civil work requires a second engineer’s independent review versus your own sign-off alone?Collaboration NormsSenior–leadership

Peer engineer insufficient AI checkingA peer engineer keeps submitting AI-drafted calculations for your review with what feels like insufficient independent checking on their part. How do you raise that without it turning into a fight about whether AI tools are “allowed”?Collaboration NormsEntry–leadership

Trusting AI quantity takeoff softwareWalk me through how you use AI-powered quantity takeoff software on a new bid set, and where you stop trusting it and go back to the drawings yourself.Tool UsageEntry–leadership

AI schedule missing cure-time holdAn AI-generated CPM schedule sequences formwork strip immediately after pour with no cure-time hold, and it's already been shared with the crew. What's your process for catching this before it drives a Monday morning decision?JudgmentMid–leadership

Disclosing AI-drafted RFI responseYou used an AI assistant to draft a first response to an RFI. Do you tell the architect that, and how do you decide what to disclose?Stakeholder CommunicationEntry–leadership

Responding to AI errors discovered on siteAn AI-drafted RFI response got a rebar spacing detail wrong, and the error wasn't caught until the placing crew was already on site. Walk me through what you do in the next hour, and what changes afterward.Error HandlingEntry–leadership

AI clash detection field trustYour BIM coordination process now leans on AI-assisted clash detection. How do you help subcontractors and field staff who don't trust — or don't understand — what the tool is flagging?UpskillingEntry–leadership

Overseeing AI-generated punch list itemsYou set up an AI agent that reviews site photos from daily walkthroughs and auto-generates punch list items. What oversight do you build in before those items go on an official list a sub has to act on?Agentic WorkflowsEntry–leadership

Payroll data through general AIA sub asks you to run their certified payroll data through a general AI tool to help format it for submission. What's your answer, and why?Governance RiskEntry–leadership

Reviewing AI-drafted change order pricingYour team now uses AI to draft pricing on change order requests before they go to the owner. What review standard do you set so a bad AI number doesn't go out the door?Collaboration NormsEntry–leadership

Verifying AI percent-complete estimatesYou're using drone photogrammetry and an AI model to estimate percent-complete for pay application purposes. Where does that automated number need a manual site check before you sign off on it?Tool UsageEntry–leadership

Trusting unexpectedly low AI estimatesAn AI-assisted estimate comes back on an excavation package that looks significantly lower than your gut says it should be, but the geotechnical report doesn't flag anything unusual. How do you decide whether to trust the number before it goes into a bid?JudgmentSenior–leadership

AI bid scoring fairnessA sub finds out your bid leveling process now uses AI to compare and score subcontractor proposals, and asks whether that's fair to smaller bidders. How do you respond?Stakeholder CommunicationEntry–leadership

JHA missed overhead hazardAn AI-generated JHA for a trenching task didn't flag a nearby overhead power line as a hazard, and a crew came close to contact before someone caught it on site. What's your response, and what does it change about how your team uses AI for safety planning?Error HandlingSenior–leadership

Superintendent rejects AI schedulingA veteran superintendent tells you flatly that he doesn't trust AI scheduling tools and wants to keep building the schedule by hand. How do you handle that?UpskillingEntry–leadership

Auto-drafting RFI and submittal responsesYou're considering letting an AI agent monitor the RFI and submittal logs and auto-draft routine responses to keep them from going stale. As the person accountable for the project's compliance record, where do you draw the line on what it's allowed to do without a human in the loop first?Agentic WorkflowsSenior–leadership

Vetting an AI clash-detection toolA vendor offers an AI clash-detection tool that improves over time by training on the BIM models and as-built data you upload. What do you check before letting your team use it on a client's project?Governance RiskEntry–leadership

Approving AI-suggested means and methodsAn AI tool suggests an alternate means-and-methods approach for a shoring sequence that looks efficient. What's your team's standard for who has to sign off before that suggestion touches the site?Collaboration NormsEntry–leadership

Using AI for market-sizing modelsWalk me through how you’d use an AI tool to build a first-pass market-sizing model for a client’s growth-strategy engagement, and where you’d stop trusting its output and do the math by hand.Tool UsageEntry–leadership

AI-drafted steering committee deckYour team uses an AI tool to draft the first cut of a client steering-committee deck’s narrative and chart annotations — walk me through which parts of that deck you’d let the tool touch and which parts stay entirely yours.Tool UsageEntry–leadership

Verifying AI competitor benchmarking tablesYou asked an AI assistant to synthesize competitor benchmarking data pulled from several public filings and it produced a clean comparison table — walk me through how you’d verify each number before it goes into a client deliverable.JudgmentEntry–leadership

Catching misaligned AI issue treesAn AI tool generates a tidy issue tree for a client’s operating-model problem that looks structurally sound but doesn’t quite fit what you heard in the diligence calls — walk me through how you’d catch that and what you’d do next.JudgmentEntry–leadership

Disclosing AI in market sizingA client asks directly whether the market sizing in your deck was built with AI assistance — walk me through how you’d answer and what you’d already have in place to answer confidently.Stakeholder CommunicationEntry–leadership

Documenting AI use for counselYou’re the engagement lead and the client’s general counsel asks how AI tools were used anywhere in the engagement, from research to expert-interview notes — walk me through how you’d prepare that answer before the meeting, not during it.Stakeholder CommunicationSenior–leadership

AI proposal cited nonexistent statisticAn AI-drafted section of a proposal cited a market-growth statistic that turns out not to trace to any real source, and the proposal already went to the client — walk me through what you do in the next 24 hours.Error HandlingEntry–leadership

Correcting reversed AI interview summariesYour team’s recommendation to a client hinged partly on an AI-summarized quote from an expert interview, and post-delivery you discover the AI summary reversed the expert’s actual position — walk me through how you handle that with the client and the team.Error HandlingEntry–leadership

Keeping teams current on toolsHow do you keep your case teams current on which AI tools are actually useful for which parts of casework, given how fast both the tools and the firm’s approved list change?UpskillingEntry–leadership

Coaching overreliance on AI frameworksYou notice an associate on your case team defaulting to whatever framework an AI tool suggests instead of reasoning about the client’s specific situation — walk me through how you’d coach them.UpskillingEntry–leadership

Supervising autonomous competitive scansWalk me through how you’d set up and supervise an AI agent to run a multi-source competitive scan for a client engagement — what it’s allowed to do autonomously, and where you check in.Agentic WorkflowsEntry–leadership

AI drafting expert interview guidesYou give an AI agent the job of drafting expert-interview discussion guides and flagging follow-up questions across a dozen interviews in a diligence sprint — walk me through the guardrails you’d put around that.Agentic WorkflowsSenior–leadership

Confidential data into general chatbotA client’s data room includes financials marked highly confidential, and someone on your team wants to paste a summary table into a general AI chatbot to speed up analysis — walk me through what you’d do.Governance RiskEntry–leadership

Managing proprietary strategy overlap riskAn AI tool trained partly on your firm’s own frameworks generates client-ready content that looks close to another client’s proprietary strategy — walk me through how you think about that risk before it ships.Governance RiskSenior–leadership

Review standard for AI-output deliverablesWhat review standard would you set for any slide or number in a client deliverable that started as AI output, and how would you make sure the whole case team actually follows it?Collaboration NormsEntry–leadership

Firm-wide AI disclosure normsAs a director running multiple case teams, how do you set a firm-wide norm for when consultants must flag that a piece of analysis was AI-assisted, versus when it’s just background tooling?Collaboration NormsSenior–leadership

Routing fraud claims from chatbotsA customer's chat opens with a billing complaint that's actually a sensitive fraud claim buried in the second message. Walk me through how your AI chat routing decides that this ticket needs to leave the bot and land with a human agent.Tool UsageEntry–leadership

What to exclude from automationYour team uses an AI assistant to draft macro responses straight from the knowledge base. What do you deliberately keep out of that automation, and why?Tool UsageEntry–leadership

Catching outdated AI responsesAn AI-drafted response cites a 30-day return window, but your team just moved to 45 days for a product line and the knowledge base hasn't caught up. How would you catch that before it reaches a customer, and after it already has?JudgmentEntry–leadership

Preserving judgment beyond model scoresYour AI escalation model scores a complaint as low-risk, but the transcript shows a customer who's calm, precise, and clearly done — the classic profile of someone about to cancel rather than blow up. How do you make sure judgment like that doesn't get lost to the model's score?JudgmentEntry–leadership

Disclosing AI agents to customersWalk me through how and when you disclose to a customer that they're talking to an AI agent rather than a person — and where you draw the line if they ask directly.Stakeholder CommunicationEntry–leadership

Explaining AI recommendation overridesA customer disputes a billing charge, the AI assistant recommends a resolution, and you end up offering something different once you review the account. How do you explain that gap to the customer without undermining confidence in the tool?Stakeholder CommunicationEntry–leadership

Handling AI transcription errorsAn AI voice agent mis-transcribed a customer's request and processed a replacement for the wrong item. Walk me through how you'd handle it from the moment you spot it.Error HandlingEntry–leadership

Data breach in AI summarizationAn AI-drafted email response to a customer accidentally included another customer's order details — a data mix-up in the summarization tool. What's your first hour, and what changes after that?Error HandlingEntry–leadership

Retraining agents after AI deflectionAI deflection has cut your team's ticket volume by a third. Walk me through how you retrain frontline agents whose day-to-day used to be high-volume, low-complexity tickets, now that those are mostly gone.UpskillingSenior–leadership

Vetting AI call-scoring toolsA vendor pitches you an AI-based call-scoring tool to replace manual QA sampling. What do you check before rolling it out to the floor?UpskillingEntry–leadership

Guardrails for autonomous refund authorityYou're designing a fully agentic flow where the AI can autonomously issue refunds or replacements up to a set dollar threshold without a human in the loop. Walk me through the guardrails you'd put around that authority.Agentic WorkflowsSenior–leadership

Human checkpoints in cancellation flowsDescribe a multi-step AI agent handling a subscription cancellation end-to-end — retention offer, cancellation, confirmation. Where did you insert human checkpoints, and why there specifically?Agentic WorkflowsEntry–leadership

Governance for transcript-fed AI toolsSupport transcripts containing names, account numbers, and sometimes payment details now feed an AI summarization and QA tool. Walk me through the governance decisions you'd need signed off before that goes live.Governance RiskSenior–leadership

Verifying compliance in AI responsesA compliance reviewer asks whether your AI-generated responses to complaints include the disclosure language regulations require. How would you actually verify that, rather than assume the templates are right?Governance RiskEntry–leadership

When to edit AI draftsSet the norm for your support team: when can an agent send an AI-drafted response verbatim, and when does it require an edit or a second look? How do you get the team to actually follow it?Collaboration NormsSenior–leadership

Teaching AI knowledge-base judgmentYou're onboarding a new agent. How do you teach them when to trust an AI knowledge-base suggestion versus when to stop and ask a subject-matter expert?Collaboration NormsEntry–leadership

Generative AI in design conceptingWalk me through where generative AI earns a place in your concepting process — moodboards, rough layout exploration — and where you deliberately keep it out, like final brand-mark refinement or a client's hero visual.Tool UsageEntry–leadership

Vetting AI-generated product imageryA generative AI tool produced a batch of product lifestyle images for a client pitch. What do you check before those images go in front of the client?JudgmentEntry–leadership

Disclosing AI in client pitchesA client loves a concept from the pitch deck and later asks, point-blank, whether it was made with AI. How do you handle that conversation?Stakeholder CommunicationEntry–leadership

Catching AI errors before printAn AI-assisted layout went to print and the client calls to say a headline is subtly misspelled — the kind of error a generative retouch introduces and nobody caught in review. Walk me through what you do next.Error HandlingEntry–leadership

Staying current with AI toolsGenerative AI tools in this field change every few months — new models, new plugins, new client expectations. How do you keep your own skills current, and how would you bring a team along with you?UpskillingEntry–leadership

Building oversight into automated workflowsYou're asked to set up an agentic workflow that takes an approved hero design and automatically produces every resized, localized variant a campaign needs — social crops, print bleeds, translated copy. What oversight do you build in before you'd trust it unsupervised?Agentic WorkflowsMid–leadership

Protecting confidential data with AIA client's brief includes unreleased product photography and confidential positioning language. What's your policy on which AI tools that material is allowed to touch?Governance RiskEntry–leadership

Documenting AI-generated elementsHow does your team flag, inside a shared design file, which elements were AI-generated versus hand-built, so the next person working on it — or the client during handoff — isn't guessing?Collaboration NormsEntry–leadership

Standardizing AI tools across studiosAs a creative director, how do you decide which AI tools get standardized across the studio for brand identity work versus left as individual designers' personal experimentation?Tool UsageSenior–leadership

Reviewing AI-drafted brand guidelinesYou're reviewing an AI-drafted first pass of a brand guidelines document — logo usage, color codes, typography rules. What kinds of errors do you specifically look for before it's treated as the source of truth?JudgmentEntry–leadership

Disclosing AI to regulated clientsA client's industry has strict advertising or trademark review requirements. Before you use generative AI on any of their creative, what do you tell them, and what do you ask them?Stakeholder CommunicationEntry–leadership

Approved mockup has IP riskDuring final review, a designer notices an AI-generated background element in an approved mockup closely resembles another brand's recognizable pattern or mark. The client has already signed off. What do you do?Error HandlingEntry–leadership

Approving AI vendors for clientsYou're setting studio-wide policy on which generative AI vendors are approved for client work. Walk me through the criteria you'd weigh, given that briefs often include confidential product and positioning information.Governance RiskSenior–leadership

Vetting AI layout variationsYou set an AI agent to generate twenty layout variations of a landing page hero for A/B testing. Walk me through how you go from that raw output to what actually gets tested.Agentic WorkflowsEntry–leadership

Developing design craft beyond promptsAs a creative director, how do you make sure junior designers who lean heavily on generative AI for concepting are still building real design craft, not just prompt fluency?Collaboration NormsSenior–leadership

Reskilling displaced production designersA production designer on your team spent most of their time on manual asset resizing and format variations — work generative AI now handles well. How do you help them reskill instead of just becoming redundant?UpskillingEntry–leadership

AI drafting versus pedagogical thinkingWalk me through how you use an AI drafting tool to build a first-pass lesson plan or unit outline — and where you stop relying on it and do the pedagogical thinking yourself.Tool UsageEntry–leadership

Deciding what grading to automateWhich parts of grading and feedback do you let an AI tool touch — and which assessments do you keep entirely in your own hands?Tool UsageEntry–leadership

Validating AI-drafted IEP goalsAn AI tool drafts a goal for a student's IEP that sounds well-written but doesn't actually match the present-levels data in their file. How do you catch that before it goes to the team meeting?JudgmentEntry–leadership

Identifying generic report-card commentsYou review a batch of AI-drafted report-card comments and notice a few read as generic — they could describe almost any student. How do you decide which ones to trust and which to rewrite?JudgmentEntry–leadership

Disclosing AI in student documentsA student's family asks whether AI was involved in drafting their child's progress report or personalized learning plan. How do you answer, and what do you disclose?Stakeholder CommunicationEntry–leadership

Presenting AI-analyzed data to boardsYou're presenting attendance or achievement data to the school board and part of the analysis was generated by an AI tool. Walk me through how you frame that in the presentation.Stakeholder CommunicationSenior–leadership

Handling false AI detection flagsAn AI writing-detection tool flags a student's essay as likely AI-generated, but you suspect it's a false positive. Walk me through how you handle the situation from there.Error HandlingEntry–leadership

Catching AI placement errorsAn AI-assisted scheduling or placement tool puts a student in the wrong course track — say, missing a gifted or English-learner designation that should have routed them differently. How do you catch it and fix it?Error HandlingEntry–leadership

Professional development across skill levelsYour staff has wildly different comfort levels with AI tools — some teachers are building agentic workflows, others haven't touched it. How do you design professional development that meets both?UpskillingEntry–leadership

Keeping policy current without whiplashState guidance and your own district's academic-integrity policy on student AI use are both still shifting. How do you, as a curriculum leader, keep your own practice and your staff's current without whiplashing them every month?UpskillingSenior–leadership

Generating differentiated lessons with AIWalk me through using an AI agent to generate differentiated versions of a single lesson or assessment — say, three reading levels plus accommodated versions for IEP students — before it reaches your class.Agentic WorkflowsEntry–leadership

Human checkpoints in AI pipelinesYou're overseeing a multi-step AI pipeline that auto-grades formative quizzes, flags struggling students, and drafts outreach emails to parents. Where do you put human checkpoints, and why there?Agentic WorkflowsSenior–leadership

Vetting AI for student dataYou're vetting a third-party AI tool that would touch student rosters, grades, or IEP data. Walk me through what you check before it's allowed in the district, and what would disqualify it outright.Governance RiskSenior–leadership

Assignment-by-assignment AI integrity policyHow would you design an academic-integrity policy for student AI use that actually works assignment by assignment, rather than one blanket rule for the whole school?Governance RiskSenior–leadership

Defining acceptable AI collaborationYour grade-level team can't agree on how much of a lesson plan or set of feedback comments is acceptable to have AI draft before it stops being ‘your own work.’ How do you get the team to a shared standard?Collaboration NormsEntry–leadership

Trusting AI-recommended student groupingsYou want to use AI to analyze formative-assessment data across a grade level and recommend small groups for differentiated instruction. What norms do you set with your team about how much to trust the groupings it suggests?Collaboration NormsEntry–leadership

Drafting versus writing portfolio reviewsYou’re prepping a quarterly portfolio review for a client using an AI tool that pulls performance and allocation data from your CRM and custodian feeds. Walk me through which parts of that review you’d let the tool draft outright, and which parts — the market commentary, the recommendation — you’d insist on writing yourself.Tool UsageEntry–leadership

Validating AI retirement projectionsAn AI planning tool can run a Monte Carlo retirement projection in seconds. Walk me through how you decide when that output is ready to show a client versus when it needs more work from you first.Tool UsageEntry–leadership

Overriding AI annuity recommendationsYour firm’s AI tool flags a client as suitable for a higher-risk annuity product based on their stated risk tolerance. What would make you override that recommendation, and how do you decide?JudgmentEntry–leadership

Investigating unexpected VaR changesA risk model with AI-driven inputs shows your desk’s portfolio VaR dropping sharply overnight with no obvious market driver. Walk me through how you’d figure out whether that’s real or a model artifact.JudgmentEntry–leadership

Disclosing AI use to clientsA client asks directly whether the retirement plan you just presented was generated by AI. How do you answer, and what do you think firms owe clients in terms of disclosure here?Stakeholder CommunicationEntry–leadership

Documenting AI's role in recommendationsCompliance asks you to explain, in writing, how an AI tool contributed to a client recommendation that’s now under review. Walk me through what that explanation needs to include to hold up.Stakeholder CommunicationSenior–leadership

Responding to AI-drafted email errorsAn AI-drafted client email went out with a performance figure that turns out to be wrong — it used last quarter’s number instead of this quarter’s. Walk me through what you do in the first hour, and what changes after.Error HandlingEntry–leadership

Reviewing AML monitoring failuresYour AML transaction monitoring system, which uses AI to score alerts, cleared a pattern of transactions as low-risk that later turns out to have been structuring. Walk me through the review that follows and what you’d change.Error HandlingEntry–leadership

Training advisors on AI toolsYou’re rolling out an AI-assisted financial planning tool to a team of advisors with mixed comfort levels with the technology. Walk me through how you’d get the team proficient without creating inconsistent client experiences in the meantime.UpskillingMid–leadership

Implementing new SEC AI guidanceNew SEC guidance narrows what firms can say about AI-generated investment analysis in client-facing materials. Walk me through how you’d get your team current on the change and confident applying it.UpskillingEntry–leadership

Overseeing AI-assisted KYC onboardingYou’re considering letting an AI agent handle first-pass document review for new-account KYC onboarding — pulling ID data, flagging mismatches, drafting the file. Walk me through what oversight you’d put around it before it touches real accounts.Agentic WorkflowsEntry–leadership

Designing friction in rebalancing workflowsAn AI agent is proposed to automatically generate and queue rebalancing trades across a book of client accounts when drift exceeds a threshold, with an advisor sign-off step before execution. Walk me through where you’d want friction in that workflow and where you’d let it move fast.Agentic WorkflowsSenior–leadership

Fine-tuning AI with client dataYour firm wants to use client financial data — account balances, transaction history — to fine-tune or improve an internal AI planning tool. Walk me through what governance questions you’d raise before that happens.Governance RiskEntry–leadership

Model risk management standardsWalk me through how you’d apply model risk management standards to a third-party AI tool your firm is licensing for investment recommendations, versus a model your firm built in-house.Governance RiskMid–leadership

Standardizing trust in AI memosTwo advisors on your team have different personal standards for how much they trust AI-drafted investment committee memos — one rewrites almost everything, one barely edits. Walk me through how you’d get the team to a shared norm.Collaboration NormsEntry–leadership

Setting team AI normsSome advisors on your team use AI chat assistants to prep talking points before client meetings; others don’t trust the practice at all. Walk me through how you’d set a team norm that doesn’t just default to the most cautious person’s comfort level.Collaboration NormsEntry–leadership

Balancing AI forecasts with judgmentWalk me through how you use an AI demand-forecasting tool to set prep quantities and supplier orders for the week, and where you still rely on your own read of the calendar and the weather instead of the model.Tool UsageEntry–leadership

Overriding AI menu recommendationsWalk me through how you’d use an AI menu-engineering tool to suggest which dishes to cut, reprice, or promote based on margin and popularity data, and where you’d overrule it even though the numbers look clean.Tool UsageMid–leadership

Verifying AI shelf-life estimatesAn AI tool estimates a four-day refrigerated shelf life for a new sous-vide chicken dish based on similar recipes in its training data. Walk me through how you’d verify that number before it goes into your prep and labeling standards.JudgmentEntry–leadership

Catching AI allergen-tagging errorsAn AI allergen-tagging tool marks a new pasta dish as gluten-free, but one of the sauce components was reformulated last month and now contains a wheat-based thickener. How do you catch that before the menu prints?JudgmentEntry–leadership

AI disclosure for food allergiesA regular guest with a shellfish allergy asks whether the seasonal special’s ingredient list was checked by a person or generated by AI. How do you answer, and what does that tell you about how your kitchen should be using these tools?Stakeholder CommunicationEntry–leadership

Surge pricing during emergenciesYour AI-driven dynamic pricing tool is recommending surge pricing on your best-selling items during a documented local emergency. Walk me through how you’d handle that recommendation and communicate the decision to ownership and the public.Stakeholder CommunicationMid–leadership

Stock-out during peak serviceThe AI forecasting tool under-ordered chicken ahead of a holiday weekend and you’re going to run out mid-Saturday dinner service. Walk me through what you do in the next two hours, and what you change afterward.Error HandlingEntry–leadership

Cooler failure despite AI monitoringYour AI-assisted HACCP monitoring system shows a walk-in cooler holding at safe temperature all night, but a manual spot-check the next morning finds it warm and product that should be discarded. Walk me through your response.Error HandlingEntry–leadership

Adoption resistance in the kitchenYou’re rolling out a new AI-powered inventory and ordering app to a kitchen team where a few cooks are uneasy about it. How do you get everyone using it well within the first month?UpskillingEntry–leadership

Maintaining fundamentals with AI toolsAs kitchen or F&B operations leadership, how do you keep your team’s AI fluency current without letting it erode the fundamentals — knife skills, palate, food safety instincts — that AI tools can’t replace?UpskillingSenior–leadership

Oversight for autonomous orderingYou’ve set up an agentic ordering system that monitors inventory levels and automatically places restock orders with approved suppliers when thresholds hit. Walk me through what oversight you build in so it doesn’t order badly.Agentic WorkflowsEntry–leadership

Reviewing AI-drafted seasonal menusWalk me through how you’d oversee a multi-step AI agent that drafts a new seasonal menu — generating recipe concepts, estimating food cost, and checking supplier availability — before any of it reaches your culinary team for testing.Agentic WorkflowsSenior–leadership

Proprietary recipes in public AIA sous chef wants to paste your restaurant group’s proprietary recipe formulations into a public AI tool to get help scaling a recipe from 10 to 200 portions. What’s your policy, and how do you make it easy to follow?Governance RiskEntry–leadership

Allergen changes from substitutionsAn AI ingredient-substitution tool suggests swapping a cream-based thickener for a starch-based one in a recipe to cut cost, but you haven’t verified whether the substitute changes the dish’s allergen profile. Walk me through the governance process before that substitution is approved.Governance RiskEntry–leadership

Approving AI recipe modificationsYour kitchen has started using an AI tool to suggest recipe modifications for cost or seasonality. What team norm do you set for who reviews and approves an AI-suggested change before it’s tested in the kitchen?Collaboration NormsEntry–leadership

Kitchen and front-of-house handoffFront-of-house wants to use AI-generated menu descriptions and allergen call-outs pulled straight from a recipe database, but the kitchen hasn’t signed off on the wording. How do you set up the handoff between kitchen and front-of-house so this doesn’t create a mismatch guests notice?Collaboration NormsEntry–leadership

AI dialogue versus hand-written textWalk me through where you’d use an AI assistant to draft NPC dialogue or quest text, and where you’d insist on writing it by hand instead.Tool UsageEntry–leadership

Procedural generation in level designDescribe how you’d bring a generative or procedural AI tool into level design — say, for populating a large open world with encounters or terrain variation — while making sure the result still plays the way you intend.Tool UsageEntry–leadership

Catching AI concept art errorsAn AI concept-art tool hands you a set of character designs that look great at a glance but have subtle anatomy or perspective errors an artist wouldn’t make. Walk me through how you catch that before it reaches the art director.JudgmentEntry–leadership

AI playtesting versus real playersYour AI-driven playtesting bots report a level as well balanced, but real player telemetry after launch shows a very different picture. Walk me through how you’d catch that gap earlier, and what you’d trust the bots for going forward.JudgmentMid–leadership

Disclosing AI-generated voice workYour team used a generative AI voice model to produce some NPC barks and filler dialogue. Walk me through how you’d disclose that to players and the community, and what you’d do differently for a game with a name-brand voice cast.Stakeholder CommunicationEntry–leadership

Clearing generative art with legalYou want to bring a generative AI art tool into the concept pipeline for an upcoming title. Walk me through how you’d raise that with legal and the publisher before it’s greenlit, and what would make you hold off.Stakeholder CommunicationMid–leadership

Sequence-breaking from AI dungeonsA procedural or AI-assisted content generator you used for a dungeon or mission layout shipped with a sequence-breaking exploit players are already using to skip content. Walk me through the incident from discovery to fix.Error HandlingEntry–leadership

AI economy tuning gone wrongLive-ops used an AI model’s recommendation to retune drop rates for a seasonal event, and it visibly wrecked the in-game economy within a day. Walk me through how you’d respond, and what changes in that process going forward.Error HandlingEntry–leadership

Artist anxiety about AI toolsSome of your concept artists are anxious that AI-assisted texturing and ideation tools threaten their role. Walk me through how you’d help the team build real skill with these tools without it feeling like deskilling.UpskillingEntry–leadership

Adopting design AI toolsAs a lead, how do you decide which generative AI tools are worth adopting across the studio’s art and design pipelines this year versus which ones you’d pass on, given how fast the tooling landscape moves?UpskillingMid–leadership

Autonomous testing and bug filingYou want to set up an AI agent that autonomously runs regression and exploit testing across nightly builds and files bug reports. Walk me through what you’d let it do on its own versus what always needs a human in the loop.Agentic WorkflowsEntry–leadership

Checkpoints in asset production pipelineDescribe a multi-step agentic pipeline you’d build for asset production — say, concept generation through to a game-ready model — and where you’d put checkpoints for a human to approve before the agent continues.Agentic WorkflowsEntry–leadership

Voice actor consent and IPA publisher wants to use a text-to-speech model trained on a voice actor’s past recordings to generate new NPC lines without bringing the actor back into the booth. Walk me through the consent and IP issues you’d flag before that ships.Governance RiskEntry–leadership

Handling assets from unlicensed trainingLegal flags that the generative AI concept-art tool your studio adopted may have been trained on copyrighted artwork without clear licensing. Walk me through how you’d handle the assets already in production that used it.Governance RiskMid–leadership

Flagging AI involvement in workYour studio doesn’t yet have a norm for how artists and designers should flag AI-assisted work during review. Walk me through the norm you’d propose for tagging AI involvement in a submitted asset or design doc.Collaboration NormsEntry–leadership

Review standards for AI-generated codeA gameplay programmer wants AI-generated code to go through the same review bar as any other pull request, but a teammate argues it needs extra scrutiny because it could hide subtle exploits or performance issues. Walk me through where you land and why.Collaboration NormsEntry–leadership

Chatbot scope for constituent requestsYour agency deploys an AI chatbot to triage 311 constituent requests before they reach a caseworker. Walk me through which categories of requests you’d let the bot fully resolve versus route straight to a human, and why.Tool UsageEntry–leadership

Verifying AI-summarized public commentsYou asked an AI assistant to summarize public comments on a proposed rule for a briefing memo, and it reports that ‘the vast majority of commenters opposed’ the rule. How do you verify that characterization before it goes to the decision-maker?JudgmentEntry–leadership

Disclosing AI in policy documentsA journalist files a public records request asking whether AI was used to draft a policy document you worked on. Walk me through how you’d respond, and what you’d have needed to have documented in advance.Stakeholder CommunicationEntry–leadership

Screening tool wrongly flagged applicantAn AI screening tool used to flag likely-ineligible benefits applications wrongly flagged a valid applicant, delaying their payment by three weeks. Walk me through what you do in the first 48 hours, and what changes afterward.Error HandlingEntry–leadership

Training caseworkers on AI draftingYour department wants to roll out an AI drafting tool to caseworkers who’ve never used one. What does a training plan look like, and how do you know it worked?UpskillingEntry–leadership

Oversight for AI procurement agentsYou’re piloting an AI agent that drafts sections of a request for proposals and checks vendor submissions against compliance requirements automatically. What oversight do you build in before letting it touch a live procurement?Agentic WorkflowsSenior–leadership

Case details in chatbotA caseworker wants to paste details from a constituent’s benefits case file into a general-purpose AI chatbot to help draft a response letter. What do you tell them, and what’s the actual policy behind it?Governance RiskEntry–leadership

Consistent AI drafting normsYour team is drafting language for a new regulation, and some staff use AI tools to speed up first drafts while others don’t. How do you set a team norm so the final rule reads consistently and holds up to legal review?Collaboration NormsEntry–leadership

Machine translation for benefits noticesYour office serves a constituency where a third of residents aren’t primarily English speakers, and you’re considering AI translation for benefits notices. Where would you trust machine translation outright, and where would you insist on human review?Tool UsageEntry–leadership

Bias in AI inspection prioritizationAn AI tool your agency uses to prioritize which neighborhoods get inspection resources keeps ranking historically under-resourced areas lower, because it’s trained on past inspection data where those areas were visited less. How do you catch and correct for that?JudgmentEntry–leadership

Explaining AI in benefits cutsYou’re testifying at a public hearing where a resident asks pointedly whether ‘a computer’ decided their benefits were being cut. How do you answer honestly without either overstating or understating the AI’s role?Stakeholder CommunicationSenior–leadership

Vendor protest over AI-generated RFPA vendor protests a contract award, arguing that AI-generated language in the RFP was ambiguous and disadvantaged their bid. Walk me through how you’d investigate that claim.Error HandlingEntry–leadership

AI literacy curriculum for analystsYou’re asked to build the AI literacy curriculum for policy analysts across the agency, most of whom have never used these tools. What do you prioritize in the first module, and what’s explicitly out of scope?UpskillingSenior–leadership

AI agent for FOIA redactionsYou’re evaluating an AI agent that searches agency records for a FOIA request and proposes redactions for exempt material. What would make you comfortable expanding its role, and what would always stay a human call?Agentic WorkflowsEntry–leadership

Vendor contract for constituent dataYour agency is evaluating a vendor’s AI-powered case-management add-on that would process constituent data. What do you need to see in the vendor contract before you’d sign off, from a governance standpoint?Governance RiskSenior–leadership

Shared tool across sensitivity levelsTwo program offices in your agency want to share one AI drafting tool for correspondence, but they handle different types of cases with different sensitivity levels. How do you set norms so one office’s shortcuts don’t create risk for the other?Collaboration NormsEntry–leadership

Ambient scribe during patient visitWalk me through how you use an ambient AI scribe during a patient visit — what it captures automatically, what you double-check in the note before signing it, and any part of the encounter you’d never let it summarize unsupervised.Tool UsageEntry–leadership

Catching scribe errors before signatureAn ambient scribe’s note says a patient “denies chest pain,” but you recall the patient actually mentioned mild chest tightness during the visit. Walk me through how you’d catch an error like that before it becomes part of the permanent record, and what you do once you find one.JudgmentEntry–leadership

Overriding clinical decision support alertsA clinical decision support alert flags a drug interaction as high risk, but in your clinical judgment it’s manageable with monitoring. Walk me through how you decide whether to override the AI’s flag, and what you document to justify that decision.JudgmentMid–leadership

Trusting clinical decision alertsAcross the clinical decision support alerts built into your EHR — drug interactions, sepsis scores, deterioration alerts — which do you act on by default, and which do you deliberately still verify yourself before trusting? Walk me through where you draw that line.Tool UsageEntry–leadership

Disclosing AI listening to patientsHow do you tell a patient, in plain language, that an AI tool is listening to or summarizing their visit, and what do you do if they decline?Stakeholder CommunicationEntry–leadership

Framing AI in tumor boardYou’re presenting an AI-assisted differential diagnosis to a multidisciplinary tumor board. Walk me through how you frame the AI’s contribution versus the clinical team’s own reasoning so nobody in the room mistakes a suggestion for a diagnosis.Stakeholder CommunicationSenior–leadership

Discovering an AI-suggested billing errorAn AI coding assistant suggested a billing code that upcoded a visit, and it slipped through into a submitted claim. Walk me through what you’d do once you discovered it, from the correction to who you tell.Error HandlingEntry–leadership

Outdated guideline in AI appealAn AI-drafted prior-authorization appeal cited a clinical guideline that turned out to be outdated, and the appeal was denied because of it. How do you handle the immediate fix, and what changes so it doesn’t recur?Error HandlingEntry–leadership

Tracking approved AI documentation toolsHow do you stay current on which AI documentation or coding tools are actually approved for use, versus ones being piloted informally by staff?UpskillingEntry–leadership

Rolling out ambient scribingYou’re rolling out a new ambient scribing tool to a nursing unit that’s wary of AI after a bad experience with a previous system. Walk me through your training and change-management plan.UpskillingSenior–leadership

Delegating care coordination to AIDescribe a care-coordination task — referrals, prior-authorization requests, discharge follow-up calls — you’d delegate to an AI agent, versus one you’d keep fully manual, and what oversight you’d put on the agent’s actions.Agentic WorkflowsEntry–leadership

Agent misses worsening symptomsAn AI agent drafts replies to patient portal messages for clinician review before sending. Walk me through what happens when a message describes worsening symptoms the agent isn’t equipped to triage.Agentic WorkflowsEntry–leadership

De-identification before model fine-tuningYour health system wants to use de-identified patient encounter data to fine-tune a documentation model. Walk me through how you’d evaluate whether the de-identification is actually sufficient before that data leaves controlled systems.Governance RiskEntry–leadership

Using vendor tool without BAAA vendor pitches an AI note-summarization tool that isn’t yet covered by a signed business associate agreement. What do you do if a clinician on your team wants to start using it today?Governance RiskEntry–leadership

Review standard for AI notesWhat review standard does your team apply before an AI-assisted clinical note or coded claim is considered ready to sign, and who’s accountable if something in it turns out to be wrong?Collaboration NormsEntry–leadership

Setting AI-use policy across departmentsAs the person setting AI-use policy across a multidisciplinary care team, how do you decide where AI assistance should be standard practice department-wide versus where each clinician needs individual discretion?Collaboration NormsSenior–leadership

Auto-drafting personalized guest messagesYour property's AI concierge tool can auto-draft personalized welcome messages and local recommendations pulled from a guest's stay history. Walk me through which of those messages you'd let go out untouched and which ones you'd want a person to read first.Tool UsageEntry–leadership

AI rate undercuts negotiated contractYour AI-driven revenue management system recommends dropping tonight's rate to fill remaining rooms, but that rate would undercut a negotiated corporate or group contract still on the books. Walk me through how you'd catch that before the new rate goes live across your channels.JudgmentMid–leadership

Handing off from booking assistantA guest has been chatting with your property's AI booking assistant for ten minutes about a complicated multi-room, multi-date request before you step in. Walk me through how and when you'd let them know a person is now handling it.Stakeholder CommunicationEntry–leadership

AI response went live wrongAn AI-drafted response to a one-star review went live on a public review site and apologized for the wrong issue entirely, making it look like nobody actually read the guest's complaint. Walk me through what you'd do in the next hour, and what you'd change about the review-response workflow.Error HandlingEntry–leadership

Guest preference tool feels impersonalYou're rolling out an AI tool that surfaces guest preferences and stay history to front-desk and concierge staff in real time. Walk me through how you'd train the team so they use it to sound more personal, not less.UpskillingEntry–leadership

AI-automated rebooking for missed check-insYou're considering letting an AI agent handle end-to-end rebooking when a guest's flight delay means they'll miss their original check-in window — finding a new room, adjusting the folio, and notifying housekeeping automatically. Walk me through where you'd still want a person in that loop.Agentic WorkflowsEntry–leadership

Guest data through AI personalizationGuest passport numbers, payment details, and stated preferences all flow through your property's AI-personalization and chatbot stack. Walk me through the guardrails you'd want in place before rolling that out across a multi-property portfolio.Governance RiskSenior–leadership

Department heads trusting AI forecastsHousekeeping staffing and F&B prep quantities are now driven by an AI occupancy and demand forecast instead of managers' own read of the calendar. Walk me through how you'd get skeptical department heads to actually trust and use those numbers.Collaboration NormsMid–leadership

Algorithm conflicts with loyalty promiseYour AI dynamic-pricing tool wants to push a rate that's technically compliant with your rate-parity agreements but would price out the loyalty members you've promised preferential rates. Walk me through when you'd override the algorithm's recommendation.Tool UsageMid–leadership

AI misses safety issueYour AI sentiment-analysis tool scores a new review as low-priority and routine, but reading it yourself, the guest is describing a safety issue — a broken lock, a smoke detector that didn't work. Walk me through what you do.JudgmentEntry–leadership

Personalization crosses into creepyYour AI personalization engine can generate a highly specific upsell offer for a guest — say, one referencing their anniversary date pulled from a past reservation note. Walk me through where that crosses from thoughtful into unsettling, and how you'd explain that line to a revenue team that just wants the conversion lift.Stakeholder CommunicationEntry–leadership

Overbooking from sync errorA sync error between your AI booking assistant and the property management system oversold a room category, and a guest is now standing at the front desk with a confirmation for a room that doesn't exist tonight. Walk me through handling the guest in the moment and fixing the underlying gap.Error HandlingEntry–leadership

Team ignoring AI forecastYour housekeeping and F&B leads are quietly ignoring the AI demand forecast and scheduling off their own instincts instead, because a bad forecast burned them once. Walk me through how you'd rebuild that trust.UpskillingEntry–leadership

Translation loses emotional toneAn international guest is using your property's AI translation-and-chat tool to file a complaint about a billing issue, and something in the translated message suggests real frustration that may not be coming through in the auto-translated tone. Walk me through how you'd handle the handoff.Agentic WorkflowsEntry–leadership

Facial recognition check-in rolloutCorporate wants to roll out facial-recognition check-in kiosks powered by AI to speed up arrivals. Walk me through the consent, accessibility, and data questions you'd want answered before supporting that rollout.Governance RiskMid–leadership

Reviewing AI-drafted guest communicationsWalk me through your team's review standard for AI-drafted guest communications — welcome messages, concierge recommendations, review responses — before anything reaches a VIP or top-tier loyalty guest.Collaboration NormsEntry–leadership

AI touchpoints in recruiting pipelineWalk me through which parts of your requisition-to-offer pipeline you let an AI copilot touch — job req drafting, resume screening, scheduling — and which step you've deliberately kept AI out of entirely.Tool UsageEntry–leadership

Resume screener misses title mismatchAn AI resume-screening tool ranks a candidate low because their title history doesn't match the keywords in the req, but you can see they've done the actual job under a different title. How do you catch cases like this before they're auto-rejected?JudgmentEntry–leadership

Candidate asks about AI screeningA candidate in a jurisdiction covered by NYC Local Law 144 asks what role AI played in screening their application. What do you tell them, and what do you make sure you can actually back up?Stakeholder CommunicationEntry–leadership

AI summary drops prior complaintAn AI tool used to summarize employee relations case notes drops a key detail — a prior complaint against the same manager — and the case gets closed without it. How do you handle it once you discover the miss?Error HandlingEntry–leadership

Keeping up with AI regulationsAI recruiting tools and the regulations around them — bias audits, disclosure laws — are both changing fast. How do you keep your team current on both without turning it into a compliance seminar nobody wants to sit through?UpskillingEntry–leadership

Guardrails for autonomous candidate sourcingYou're piloting an AI agent that can source passive candidates, draft outreach, and schedule first-round screens without a recruiter touching each step. What guardrails do you put around it before letting it run?Agentic WorkflowsEntry–leadership

Preparing for bias audit requirementYou're the person accountable for your organization's compliance with an AI bias audit requirement like NYC Local Law 144. Walk me through what you actually need in place before an AI-driven screening tool can go live.Governance RiskSenior–leadership

Legal blocks recruiting tool rolloutRecruiting wants to roll out an AI tool for interview note summarization; legal wants a six-month review process first. How do you get the team to a working agreement instead of a standoff?Collaboration NormsEntry–leadership

Validating AI compensation benchmarkingYou're using an AI tool to benchmark a role's compensation range against market data. What do you check before you trust the number it gives you enough to put it in front of a hiring manager?Tool UsageEntry–leadership

Survey summary misses team complaintsAn AI tool summarizes your quarterly engagement survey and reports morale as stable, but you're hearing more informal complaints from a specific team than the summary suggests. How do you reconcile that?JudgmentEntry–leadership

Hiring manager skips reviewA hiring manager wants to skip your team's review and go straight off an AI candidate-ranking tool's top pick to save time. How do you handle that conversation?Stakeholder CommunicationEntry–leadership

Responding to AI chatbot misinformationNew hires start reporting that the AI onboarding chatbot gave them wrong information about benefits enrollment deadlines. What's your response, both to the affected employees and to the tool itself?Error HandlingEntry–leadership

Sharing anonymized data with vendorYour team wants to feed anonymized resume and performance data into a third-party AI vendor to build a better candidate-matching model. What do you need resolved before that data leaves your systems?Governance RiskMid–leadership

AI scheduling agent hits limitYou've set up an AI agent to handle interview scheduling across candidates, panelists, and multiple time zones, including rebooking when someone cancels. Walk me through what happens when it hits a case it can't resolve on its own.Agentic WorkflowsEntry–leadership

Job description review thresholdSet the norm for your recruiting team on when an AI-drafted job description needs a second human read before it posts, versus when it can go live as-is. How do you decide where that line sits?Collaboration NormsMid–leadership

Validating AI bias reduction claimsYou're evaluating a new AI hiring tool a vendor claims reduces bias in candidate ranking. What evidence do you actually require before you believe that claim, versus what the sales deck shows you?JudgmentSenior–leadership

Underwriting model uses stale photosAn AI-assisted underwriting model recommends declining a homeowners application, but the property inspection photos it scored look outdated and don’t match the address on file. Walk me through what you’d do before that decision goes out.JudgmentEntry–leadership

Explaining underwriting score denialA policy applicant is denied coverage in part because of a machine-learning underwriting score, and state law requires you to give a specific reason in the adverse action notice. How do you make sure the reason you send is accurate and defensible?Governance RiskEntry–leadership

AI intake handoff to adjusterWalk me through how you use an AI assistant during first notice of loss intake, and where you stop relying on it and hand the claim to a human adjuster.Tool UsageEntry–leadership

Explaining AI involvement to policyholdersA policyholder asks whether an AI system was involved in denying part of their claim. How do you explain that to them in a way that’s honest and holds up if the state insurance department asks the same question?Stakeholder CommunicationEntry–leadership

Correcting AI fraud detection delaysAn AI fraud-detection model flags a legitimate claim as suspicious, and the claim sits unpaid past the state-mandated prompt-payment deadline while your team investigates. Walk me through how you handle it once you realize the flag was wrong.Error HandlingEntry–leadership

Oversight for autonomous claims paymentYou’re rolling out an agentic tool that can auto-adjudicate and pay small property claims end to end below a dollar threshold. What oversight do you build in before letting it touch money without a human in the loop?Agentic WorkflowsMid–leadership

Teaching actuaries machine-learning modelYour actuarial team is being asked to incorporate a new machine-learning pricing model alongside the traditional GLM rating plan. How do you help less AI-comfortable actuaries get up to speed without losing rigor?UpskillingEntry–leadership

AI coverage recommendation reviewYour underwriting team starts using an AI assistant to draft coverage recommendations for complex commercial risks. What review norm do you put in place before those recommendations go to the underwriter of record?Collaboration NormsEntry–leadership

Pricing model diverges from lossAn AI-driven pricing model’s indicated rate for a segment diverges sharply from what your actual loss experience shows over the last two quarters. Walk me through how you’d investigate before trusting the model’s number in a rate filing.JudgmentMid–leadership

Proxy variables in fairness auditsA fairness audit shows your AI underwriting model is using a variable that correlates strongly with a protected class, even though the variable itself seems neutral, like credit-based insurance score or zip code. How do you decide whether it stays in the model?Governance RiskSenior–leadership

AI policy summarization workflowWalk me through how you use AI to summarize a lengthy commercial policy with multiple endorsements before a coverage determination, and what you still read line by line yourself.Tool UsageEntry–leadership

Stale data in fraud scoringYour special investigations unit relied on an AI fraud score to refer a claim for investigation, and it turns out the score was based on a data feed that hadn’t updated in months. What do you do once you find that out?Error HandlingEntry–leadership

Agent resistance to triage toolsYou’re introducing an AI claims-triage tool to a network of independent agents and producers who are worried it will second-guess their judgment. How do you communicate the rollout so it lands as support rather than a threat?Stakeholder CommunicationMid–leadership

Autonomous claims workflow boundariesWalk me through a multi-step claims workflow you’d let an AI agent run on its own, from document collection to a payment recommendation, and the exact point where a human has to sign off.Agentic WorkflowsEntry–leadership

Model validation documentation trailYour compliance team asks for documentation showing how an AI underwriting model was validated before it went into production, in case a state Department of Insurance asks for it during a market conduct exam. What does that documentation trail look like?Governance RiskEntry–leadership

Cross-functional testing standardsActuarial, underwriting, and compliance all have a stake in a new AI rating model, and they don’t agree on how much testing is enough before it ships. How do you get them to a shared standard?Collaboration NormsEntry–leadership

Comparable-companies analysis splitWalk me through which parts of building a comparable-companies analysis you'd hand to an AI tool versus keep manual, and why the line falls where it does.Tool UsageEntry–leadership

AI drafting qualitative CIM sectionsYour team is deciding whether to let an AI assistant draft the qualitative sections of a CIM — business overview, management bios, market positioning. How do you decide what's safe to delegate?Tool UsageEntry–leadership

Covenant extraction accuracy checkAn AI tool summarizing a target's credit agreement lists a leverage covenant at 4.5x when the actual document says 4.0x. Walk me through how this gets caught before it shapes your LBO model.JudgmentEntry–leadership

Buried DCF assumptions in reviewA junior analyst submits an AI-assisted DCF where the terminal growth rate assumption is buried three tabs deep and doesn't match what's stated in the pitch book narrative. What's your review process, and what would make you send it back?JudgmentEntry–leadership

Disclosing whether comps were AI-generatedA client asks whether the comp set in the valuation section of your presentation was selected by an analyst or generated by an AI tool. How do you answer, and how does that shape what you disclose going forward?Stakeholder CommunicationEntry–leadership

AI sensitivity analysis disclosureYou're presenting a fairness opinion to a board, and part of the valuation cross-check was run through an AI-assisted sensitivity analysis. How much of that process do you surface to the board, and how much stays in methodology?Stakeholder CommunicationSenior–leadership

Outdated data in pitch materialsAn AI-generated first draft of a pitch book quotes a market share figure that turns out to be from an outdated industry report. It's already gone out to the client in advance materials. Walk me through what you do next.Error HandlingEntry–leadership

Double-counted EBITDA add-backsYour team discovers, two days before signing, that an AI-assisted adjustment to EBITDA add-backs in the deal model was double-counting a one-time expense reversal — and it's already flowed into the valuation the client has anchored on. How do you manage the correction with the deal team and the client?Error HandlingMid–leadership

Maintaining modeling fundamentalsHow do you keep your modeling skills sharp — building an LBO or DCF from scratch — when AI tools can generate a working model shell in minutes?UpskillingEntry–leadership

Pre-AI training for new analystsYou're asked to set the AI-tool training plan for a new analyst class before they touch live deal models. What do you make sure they can do without AI assistance before you let them use it?UpskillingEntry–leadership

Data room document flagging supervisionWalk me through how you'd supervise an AI agent tasked with scanning a virtual data room and flagging documents relevant to a specific diligence question, from setup through final sign-off.Agentic WorkflowsEntry–leadership

Autonomous precedent comp populationYou're considering letting an AI agent autonomously pull precedent-transaction comps from a subscription database and populate the first draft of a transaction summary, running end to end with no human touching individual entries. What oversight would you build in before you'd sign off on that workflow?Agentic WorkflowsMid–leadership

Unapproved AI for deal correspondenceA generative AI tool your firm hasn't formally approved would save real time drafting deal correspondence, but running it means pasting in details from an active mandate. Walk me through your decision.Governance RiskEntry–leadership

AI tools across information barriersYour desk is on one side of an information barrier during a live deal, and someone proposes using a firm-wide AI assistant that indexes shared drives to speed up research. What questions do you need answered before that tool touches anything connected to your deal?Governance RiskEntry–leadership

Flagging AI-assisted versus manual sourcingWhat's your team's norm for flagging in a model or memo which numbers came from an AI-assisted pull versus a manually sourced figure, and why does that distinction matter when someone else picks up the file?Collaboration NormsEntry–leadership

Reviewing AI-drafted models and memosYou're setting review standards for a deal team where analysts increasingly use AI to build first drafts of models and memos. What changes about how you review their work compared to when everything was built by hand?Collaboration NormsMid–leadership

AI equity screening workflowWalk me through how you’d use an AI tool to build a first-pass equity screen across a broad universe, and where in that process you’d switch back to doing the work yourself.Tool UsageEntry–leadership

Earnings call preparation with AIWhen you’re prepping for an earnings call, what do you let an AI summarizer do with the transcript and prepared remarks, and what do you always listen to or reread yourself?Tool UsageEntry–leadership

Verifying AI debt covenant summariesAn AI drafting tool summarizes a company’s latest 10-K and states its debt covenants are unchanged from last year. How would you verify that before it shapes your credit view?JudgmentEntry–leadership

Reviewing AI-assisted valuation modelsYou’re reviewing a junior analyst’s AI-assisted valuation model and the DCF output looks reasonable, but something about the growth assumptions feels off. As the senior reviewer, how do you decide whether to dig in or let it pass?JudgmentMid–leadership

AI disclosure to limited partnersA limited partner asks whether AI tools were involved in generating the return attribution in their quarterly letter. How do you answer that, and how does it change what you’d disclose going forward?Stakeholder CommunicationEntry–leadership

Explaining AI-validated alternative signalsHow would you explain to a skeptical investment committee that the alternative-data signal you’re proposing to add to the model was validated using AI-assisted backtesting, without losing their confidence in the result?Stakeholder CommunicationEntry–leadership

Correcting AI-generated client commentaryAn AI-generated portfolio commentary draft misattributed a sector’s outperformance to the wrong driver, and it went out to a subset of clients before anyone caught it. What do you do in the following 24 hours?Error HandlingEntry–leadership

Responding to risk model bugsYou discover that a risk model an AI agent has been auto-updating with new factor exposures introduced a subtle bug three weeks ago, and every VaR estimate since then has been understated. Walk me through what you do.Error HandlingMid–leadership

Maintaining fundamental research skillsHow are you keeping your fundamental research skills sharp now that AI tools can produce a first-draft investment thesis in minutes — what are you doing to make sure you can still build one from scratch?UpskillingEntry–leadership

Retraining analysts on manual researchAs a team lead, how are you retraining research associates who came up relying heavily on AI-assisted screening, so they can still build conviction the old-fashioned way when the model’s wrong or the data’s thin?UpskillingSenior–leadership

Autonomous portfolio rebalancing guardrailsYou’re considering letting an AI agent autonomously rebalance a portfolio sleeve within pre-set bands between review cycles. What guardrails would you need in place before you’d sign off on that?Agentic WorkflowsEntry–leadership

Delegating multi-step research to AIDescribe a multi-step research task — say, building a comparable-company set and flagging outliers — that you’d delegate to an AI agent end to end, and where you’d insert a human checkpoint.Agentic WorkflowsEntry–leadership

MNPI and expert-network AI restrictionsYour firm handles material non-public information from expert-network calls. What restrictions would you put on feeding AI tools with notes or summaries from those calls, and why?Governance RiskEntry–leadership

Validating AI-assisted trading signalsAs the person accountable for model risk, how would you structure a validation and sign-off process for a new AI-assisted signal before it’s allowed to influence live portfolio decisions?Governance RiskSenior–leadership

Disclosing AI authorship in memosWhat norms has your research team set for disclosing, in an investment memo itself, which sections were AI-drafted versus analyst-written?Collaboration NormsEntry–leadership

Credit research AI assistance thresholdHow does your team decide what counts as an acceptable level of AI assistance on a credit research note before it needs a second analyst’s review, versus routine work that ships with a single sign-off?Collaboration NormsEntry–leadership

AI ticket triage handoff pointWalk me through how you’d use an AI assistant to triage an incoming ticket queue, and where you’d stop trusting its severity call and reroute to a human dispatcher.Tool UsageEntry–leadership

Verifying AI-summarized runbooksYou’re building the first draft of a runbook from a batch of closed tickets using an AI summarizer. What do you personally verify before it goes into the knowledge base other technicians will follow during an outage?Tool UsageEntry–leadership

Plausible but wrong AI suggestionsAn AI assistant suggests a fix for a recurring server issue, and it’s plausible, well-formatted, and wrong about the root cause. What tips you off before you apply it to a client’s environment?JudgmentEntry–leadership

Trusting instinct on database migrationsYou’re leading a migration and the AI-assisted script for moving a client’s database looks clean, but something about the schema mapping bothers you. Walk me through how far you’d push that instinct before greenlighting the run against production.JudgmentMid–leadership

Disclosing AI use to clientsA client asks whether AI was involved in producing part of their deliverable — say a piece of a runbook or an audit report you generated with AI assistance. How do you answer, and what do you disclose without being asked?Stakeholder CommunicationEntry–leadership

Drafting client AI usage policiesA long-standing managed-services client, under a strict SLA, wants a written policy on how and when your team uses AI tools against their environment. How do you approach drafting that policy, and where do you push back on requests that would slow your team down too much?Stakeholder CommunicationSenior–leadership

Handling AI-caused production incidentsAn AI-suggested configuration change goes out during a routine ticket and it breaks something else in the client’s environment a few hours later. Walk me through what you do in the first thirty minutes, and how you write it up afterward.Error HandlingEntry–leadership

Patch rollout bug client notificationDuring a scheduled patch rollout across a managed-services fleet, an AI copilot’s suggested script introduces a bug that only shows up on a subset of client machines after the fact. How do you handle notifying affected clients, and in what order?Error HandlingEntry–leadership

Staying current on AI capabilitiesHow do you personally stay current on what your AI copilots can and can’t reliably do as the tools change month to month, and how do you tell the difference between a real capability improvement and marketing?UpskillingEntry–leadership

Skill atrophy in AI-assisted supportYou’re leading a support team where junior technicians increasingly lean on AI for diagnosis and are starting to lose the muscle for manual root-cause troubleshooting. How do you redesign training so AI adoption doesn’t hollow out the team’s underlying skill?UpskillingSenior–leadership

Unattended AI diagnostics with oversightYou give an AI agent permission to run a multi-step diagnostic and remediation sequence against a client’s server unattended. What oversight do you build in before you let it run, and what stops it partway through?Agentic WorkflowsEntry–leadership

Agentic AI rollout governanceYou’re rolling out agentic AI tooling that can propose and stage changes across your entire managed-services client fleet. What governance do you put in place before any client’s environment is in scope, and how do you stage the rollout?Agentic WorkflowsSenior–leadership

Pasting logs into public AIA technician wants to paste a chunk of a client’s error logs into a general-purpose AI chat tool to get help debugging faster. What’s your answer, and what would you have them do instead?Governance RiskEntry–leadership

AI access across competing clientsYour firm supports several clients who are direct competitors. How does that shape which AI tools and shared knowledge bases you’re willing to let touch each client’s account, and where do you draw the line on cross-client learning?Governance RiskEntry–leadership

Approving AI-drafted runbooksYour team has started using AI to draft first-pass runbooks and KB articles. What review standard do you set before one of those gets marked approved for other technicians to follow?Collaboration NormsEntry–leadership

Code review for AI-drafted scriptsHow does your team’s code review process change, if at all, when the script being reviewed was drafted by an AI assistant and is headed for a client’s production environment?Collaboration NormsEntry–leadership

AI contract redlining and negotiationWalk me through how you use an AI drafting or redlining tool on a commercial contract, and where in the negotiation you deliberately go back to doing the comparison by hand.Tool UsageEntry–leadership

AI research assistance for motionsWhen you're building the research for a motion, which parts do you hand to an AI research assistant and which parts do you still run through case-law databases and read yourself?Tool UsageEntry–leadership

Verifying AI-cited case lawAn AI drafting tool hands you a brief section that cites a case supporting your argument perfectly. Walk me through how you verify that case actually exists and says what the citation claims before it goes anywhere near a filing.JudgmentEntry–leadership

Trusting AI due diligence summariesYou're reviewing an AI-generated summary of a due diligence data room, and it flags no material issues. How do you decide how much of that clean bill of health to trust before you sign off?JudgmentMid–leadership

Disclosing AI use to clientsA client asks whether AI touched the memo you just billed them for. Walk me through how you answer that, and what you'd want documented beforehand to make the answer easy.Stakeholder CommunicationEntry–leadership

Certifying AI use in filingsA judge's standing order requires you to certify whether generative AI was used in preparing a filing. Walk me through how you'd decide what counts as 'used' and how you'd document it so the certification is accurate every time.Stakeholder CommunicationEntry–leadership

Responding to fabricated citationsA fabricated citation from an AI drafting tool made it into a filed brief before anyone caught it. Walk me through what you'd do in the following twenty-four hours, from the court to the client to your own team.Error HandlingSenior–leadership

AI document review privilege breachAn AI-assisted document review tool clears a batch as non-privileged, and one of those documents turns out to be genuine attorney work product that's now in opposing counsel's hands. What's your process from that moment forward?Error HandlingEntry–leadership

Staying current on AI reliabilityHow do you keep your own judgment about when an AI research or drafting tool is reliable up to date, given that the tools change versions every few months?UpskillingEntry–leadership

Developing judgment with AI toolsYou're supervising a first-year associate who leans on an AI tool to draft their first motions. Walk me through how you make sure they're still building real legal-writing and analytical judgment underneath the tool.UpskillingSenior–leadership

AI-assisted document triageWalk me through how you'd set up an AI agent to triage a document production for responsiveness and privilege across multiple custodians, and where you'd insert a required human checkpoint before anything moves downstream.Agentic WorkflowsEntry–leadership

QC for AI contract abstractionAn AI agent is abstracting change-of-control and assignment clauses out of hundreds of contracts for a due diligence summary. How do you decide your sampling and QC process so the deal team can actually rely on the output?Agentic WorkflowsMid–leadership

Protecting privileged data in AIWalk me through how you decide whether a document containing privileged client information is safe to put into an AI tool at all, and what you check about that tool before you trust it with anything client-related.Governance RiskEntry–leadership

Applying policy to edge casesYour firm or legal department has a written policy on which AI tools are approved for which kinds of work. Walk me through what you do when a task in front of you doesn't cleanly fit any category in that policy.Governance RiskEntry–leadership

Reviewing AI-drafted contract sectionsWhat review norm does your team use to make sure an AI-drafted section of a contract or brief gets the same scrutiny as one written from scratch, and how do you keep people from skipping it under deadline pressure?Collaboration NormsEntry–leadership

Enforcing client AI restrictionsA client's outside-counsel guidelines restrict which AI tools you can use on their matters. Walk me through how that gets communicated to and enforced across the whole team working the file, including contract staff and vendors.Collaboration NormsEntry–leadership

Trusting computer vision defect flagsYour line just deployed a computer-vision system that flags surface defects on stamped parts in real time. Walk me through how you decide which of its flags to trust outright and which still need a human inspector’s eyes before a part gets scrapped or shipped.Tool UsageEntry–leadership

Automating production schedulingYou’re evaluating an AI-driven scheduling tool that promises to re-optimize the production plan against machine capacity, changeovers, and due dates faster than your planners can. What would make you trust it enough to let it publish a schedule directly to the shop floor versus just recommending one?Tool UsageMid–leadership

Resolving AI and technician disagreementA predictive-maintenance model flags a critical stamping press as low risk for failure, but a technician notices a vibration signature that worries them anyway. How do you decide whose read wins, and what do you do next?JudgmentEntry–leadership

Verifying AI root-cause diagnosisAn AI-assisted SPC dashboard flags a process shift on a machining line and auto-suggests the tool wear is the root cause. Walk me through how you’d verify that diagnosis before you pull the tool and stop the line.JudgmentEntry–leadership

Disclosing AI use in PPAPA customer auditing your PPAP submission asks whether any of the process capability data or root-cause analysis was AI-generated. How do you answer, and how do you document AI’s role in the package so it holds up under a supplier audit?Stakeholder CommunicationEntry–leadership

Presenting AI-drafted 8D reportsYou used an AI tool to draft the containment and root-cause sections of an 8D report after a customer complaint. Walk me through how you’d present that report to the customer’s quality team so the AI’s role doesn’t undercut their confidence in it.Stakeholder CommunicationEntry–leadership

AI scheduling skipped inspectionAn AI scheduling tool sequenced a changeover in a way that skipped a required first-article inspection step, and a bad lot reached the warehouse before anyone caught it. Walk me through your response from the moment you find out.Error HandlingEntry–leadership

Forecast model underestimated demandThe AI demand-forecasting model your planning team relies on badly underestimated a customer’s ramp, and now the MRP run is showing shortages on long-lead components. What do you do in the next 48 hours, and what changes after?Error HandlingMid–leadership

Adopting AI work instructionsYou’re rolling out an AI-assisted work-instruction system that walks operators through changeovers step by step. How do you get a veteran operator who trusts their own muscle memory over a screen to actually adopt it?UpskillingEntry–leadership

Auditing AI quality verdictsAs AI tools get folded into your quality systems, walk me through how you’d build your quality engineers’ ability to audit and challenge a model’s output, not just accept its verdict.UpskillingEntry–leadership

Checkpoints for autonomous purchasingYou’re considering letting an AI agent monitor inventory levels, generate purchase orders, and send them to approved suppliers without a human touching each one. Walk me through where you’d put a human checkpoint back into that loop.Agentic WorkflowsEntry–leadership

Autonomous parameter adjustment authorityAn AI agent on your line can auto-adjust process parameters, like torque or temperature setpoints, in real time when it detects drift, without waiting for an engineer’s sign-off. Walk me through how you’d decide which parameters, if any, it’s allowed to touch on its own.Agentic WorkflowsSenior–leadership

Sharing sensor data with vendorsA predictive-maintenance vendor wants your machine sensor data and historical failure logs to train their model, and some of that data reflects proprietary process parameters. Walk me through how you’d evaluate that request before agreeing.Governance RiskEntry–leadership

Auditing AI-generated quality recordsYou work in a regulated environment where quality records need to be traceable and audit-ready. Walk me through the guardrails you’d put around using an AI tool to help write nonconformance reports and CAPAs.Governance RiskEntry–leadership

Using AI risk scores in material reviewQuality, production, and engineering all sit on the material review board that decides whether a suspect lot gets scrapped, reworked, or shipped with a deviation. What norms would you set for how AI-generated risk scores factor into that decision?Collaboration NormsMid–leadership

Reviewing AI-assisted root-cause analysisWalk me through the review standard you’d set for engineers who use AI to help generate fishbone diagrams or 5-Whys during root-cause investigations, so the analysis stays rigorous rather than just plausible-sounding.Collaboration NormsEntry–leadership

AI-first versus human-first contentWalk me through how you decide whether a piece of content — a landing-page headline, a paid social ad variant, a subject line — gets first-drafted by a generative AI tool versus written by a human on your team from the start.Tool UsageEntry–leadership

Manual control in bid optimizationYour team runs paid search and paid social through an AI-driven bid-optimization platform that reallocates budget across campaigns in real time. As the person accountable for the media budget, where do you keep manual control instead of letting the algorithm run unchecked?Tool UsageMid–leadership

Fact-checking AI-cited statisticsYou ask a generative AI tool to draft a client thought-leadership blog post, and it cites an industry statistic that looks impressive but you can’t find its source anywhere. What do you do before that post goes anywhere near publish?JudgmentEntry–leadership

Validating attribution model claimsA multi-touch attribution model shows paid social driving a spike in conversions last quarter, and the recommendation is to shift budget there — but you suspect the model is over-crediting the channel. How do you validate that before signing off on the reallocation?JudgmentEntry–leadership

Disclosing AI involvement to clientsA client asks directly whether the campaign concepts you’re presenting were written by a person or generated by AI. How do you answer, and how does that shape what you show them?Stakeholder CommunicationEntry–leadership

Disclosing synthetic testimonialsYour team is using an AI voice-clone or synthetic-avatar tool to produce testimonial-style content for a client’s paid campaign. Walk me through the disclosure obligations you’d make sure are handled before that runs, and who signs off.Stakeholder CommunicationMid–leadership

Personalized email merge field brokeAn AI-personalized email — dynamically inserting customer names, past purchases, and product recommendations — goes out to your full lifecycle segment with a broken merge field, so subscribers see literal placeholder text or a nonsensical recommendation. Walk me through your next hour.Error HandlingEntry–leadership

Campaign under-delivered to demographicThree weeks into a flight, you discover an AI-optimized programmatic campaign has been systematically under-delivering impressions to a demographic your client specifically wanted reached, because the algorithm optimized purely for click-through rate and that group converts differently. How do you handle this with the client?Error HandlingEntry–leadership

Reluctant writers avoiding AI toolsA few writers on your content team are quietly avoiding the AI drafting tools the team has adopted, worried it threatens their role or produces work they don’t trust. How do you bring them along?UpskillingEntry–leadership

Keeping AI skills currentGenerative AI content tools and their capabilities are changing every few months. As a marketing leader, how do you keep your team’s skills and your creative standards current without chasing every new tool that launches?UpskillingMid–leadership

Checkpoints for autonomous publishingYou’re evaluating an AI agent that can research trending topics, draft social posts, and publish them automatically across your brand’s channels with minimal human review. What checkpoints do you build in before you’d let it run unattended?Agentic WorkflowsEntry–leadership

Guardrails for agentic bid managementYou’re designing a multi-step agentic workflow that adjusts SEM bids, pauses underperforming ad sets, and reallocates budget across a portfolio of client accounts with only periodic human check-ins. As the owner of this system, what guardrails do you put on the agent’s authority, and what triggers an immediate human escalation?Agentic WorkflowsMid–leadership

Uploading customer data to AIA team member wants to upload a spreadsheet of customer purchase history and email addresses into a public generative AI tool to help build lookalike audience segments faster. Walk me through how you’d handle that request.Governance RiskEntry–leadership

AI copy mirrors competitor phrasingYou discover that a generative AI copywriting tool your team uses produced ad copy that closely mirrors a specific competitor’s tagline and brand phrasing — close enough that it could read as IP infringement or, at minimum, brand confusion. What do you do?Governance RiskEntry–leadership

Quality standards for AI-drafted contentYour content team ships far more volume now that generative AI drafts a first pass of blog posts, social copy, and email content. What review standard do you put in place so speed doesn’t quietly become a substitute for quality?Collaboration NormsEntry–leadership

Cross-team review for personalized campaignsCreative, legal, and the data science team that builds your audience-targeting models rarely talk day-to-day, but AI-driven personalization now touches all three — the copy, the compliance, and the targeting logic. What norms do you put in place so a personalized campaign doesn’t slip through with one of those three never having looked at it?Collaboration NormsEntry–leadership

AI drafting design history fileWalk me through how you'd use an AI drafting tool to help assemble a design history file summary for a submission, and which sections you'd never let it touch without a line-by-line trace back to source records.Tool UsageEntry–leadership

AI literature search in evaluationDescribe how you'd use an AI literature-search tool to support a clinical evaluation report, and where you'd still insist on pulling the primary source yourself.Tool UsageEntry–leadership

Catching gaps in AI-drafted FMEAsAn AI tool helped generate a first draft of your design FMEA and it's missing a failure mode you know has come up in complaints for a similar device. How do you catch gaps like that before the risk file goes to review?JudgmentEntry–leadership

Verifying predicate device comparisonsYou're reviewing an AI-drafted predicate device comparison table for a 510(k) and one of the technological characteristics is subtly mischaracterized in a way that makes your device look more similar to the predicate than it is. How do you catch this before it reaches regulatory affairs leadership?JudgmentMid–leadership

Auditor asks about AI useA notified body auditor asks whether AI tools were used anywhere in producing the design control documentation they're reviewing. How do you answer, and what do you have ready to back it up?Stakeholder CommunicationEntry–leadership

Addressing AI-generated training materialsA physician on your clinical advisory board notices that draft training materials for a new device read like they were AI-generated and asks about it directly. How do you handle that conversation?Stakeholder CommunicationEntry–leadership

AI complaint triage missed reportabilityAn AI-assisted complaint triage tool miscoded a customer complaint as non-reportable, and it should have been evaluated as a possible medical device report. What's your process once you discover the error?Error HandlingEntry–leadership

Responding to AI-missed test case failuresAn AI-generated software test suite for an infusion pump's dosing logic missed an edge case that later showed up as a field issue. Walk me through how you'd lead the response, from containment through the process fix.Error HandlingSenior–leadership

Validating AI-assisted test generationHow would you help a verification and validation team get comfortable using AI-assisted test case generation without it becoming something an ISO 13485 auditor flags as an uncontrolled tool?UpskillingEntry–leadership

Keeping procedures current with guidanceRegulatory guidance on AI use in submissions and quality systems keeps shifting. How do you keep your team's practices current without constantly rewriting your validated procedures?UpskillingMid–leadership

AI assembles design history fileDescribe how you'd set up an AI agent to assemble supporting documentation for a design history file, and where you'd put a mandatory human checkpoint before it moves forward.Agentic WorkflowsEntry–leadership

Overseeing AI device classification researchYou've asked a multi-step AI agent to research and draft a rationale for why a new device should be classified Class II rather than Class III. What oversight do you put around that before it becomes the basis for a submission strategy?Agentic WorkflowsSenior–leadership

Fine-tuning AI with complaint narrativesYour team wants to use complaint and adverse event narratives to fine-tune an AI tool that helps draft CAPA investigations. What governance questions do you need answered before that data goes anywhere near a model?Governance RiskEntry–leadership

Pasting study data into toolA third-party AI writing tool would speed up drafting sections of a PMA submission, but doing so means pasting unpublished preclinical study data into it. How do you evaluate whether that's acceptable?Governance RiskMid–leadership

Review standards for AI-drafted ECOsWhat review standard would you set for engineering change orders that were drafted with AI assistance, so the team trusts them the same as any other ECO?Collaboration NormsEntry–leadership

AI test protocol acceptance normsHow would you set cross-functional norms between quality, regulatory, and software engineering for when AI-generated test protocols are acceptable to use as-is versus requiring independent authorship?Collaboration NormsSenior–leadership

AI drafting donor appeal lettersWalk me through how you use AI tools to draft a year-end appeal letter to a mid-level donor, and where you draw the line on what AI shouldn’t touch before it goes out under your name.Tool UsageEntry–leadership

AI involvement in grant writingWalk me through which parts of the grant-writing process you’ll let an AI tool touch — research, first draft, budget narrative — and which parts you insist stay entirely in your own hands.Tool UsageEntry–leadership

Fact-checking AI grant proposalsYou used an AI assistant to draft a foundation grant proposal, and it states you served ‘500 families last year’ — but your database shows 340. Walk me through how you catch this before the proposal goes out.JudgmentEntry–leadership

Trusting AI fundraising projectionsYou’re looking at an AI-generated dashboard projecting your year-end fundraising total from current pledges and historical giving patterns. What makes you trust or distrust that number before you share it with your board chair?JudgmentEntry–leadership

Disclosing AI-drafted board summariesYou used AI to draft the executive summary of program outcomes in this quarter’s board deck. How do you disclose that to the board, and what do you personally verify before presenting it as the staff’s assessment?Stakeholder CommunicationMid–leadership

Chatbot disclosure on donation pageYour organization is piloting an AI chatbot on the donation page to answer donor questions. Walk me through how you decide what it discloses about being AI, and what it should never say on the organization’s behalf.Stakeholder CommunicationEntry–leadership

Correcting AI-assisted impact story errorsYour annual report includes an AI-assisted ‘impact story’ that paraphrased a beneficiary quote in a way that overstates what actually happened, and a board member catches it after publication. Walk me through how you handle the correction with donors and the board.Error HandlingMid–leadership

AI expenditure report misclassified spendingA program officer emails to flag that the AI-assisted expenditure report you submitted misclassified a restricted grant’s spending. Walk me through what you do in the next forty-eight hours.Error HandlingEntry–leadership

Introducing AI to wary teamYour program team has never used AI tools, and a funder is starting to ask how you’re using AI to extend your reach. Walk me through how you’d introduce these tools to a team that’s wary of them without losing their trust.UpskillingEntry–leadership

AI volunteer onboarding training moduleWalk me through how you’d use AI to build a volunteer onboarding training module, and how you make sure volunteers in the field aren’t relying on something the model got wrong.UpskillingEntry–leadership

AI shift-matching agent with oversightWalk me through how you’d set up an AI agent to match volunteers to open shifts and send reminder texts, and what oversight you’d keep in place before you let it run unsupervised.Agentic WorkflowsEntry–leadership

AI drafts prospect lettersYou’re piloting an AI agent that researches foundation prospects and drafts letters of inquiry for your review queue. Walk me through how you design the human checkpoints so nothing goes out that misrepresents your programs.Agentic WorkflowsMid–leadership

Organization-wide AI use policyAs executive director, the board asks you to approve an organization-wide AI use policy covering donor data, beneficiary data, and public communications. Walk me through what it needs to cover and where you draw the hard lines.Governance RiskSenior–leadership

Pasting client notes into chatbotsA case worker wants to paste anonymized client intake notes into a general AI chatbot to help draft a service plan for a survivor of domestic violence. Walk me through how you’d evaluate whether that’s safe, and what you’d put in place if it isn’t.Governance RiskMid–leadership

Donor emails need reviewYour small development team is split on whether AI-drafted donor emails need a second read before they go out. Walk me through how you’d set the team norm, and why.Collaboration NormsEntry–leadership

Client story numbers divergeProgram staff and development staff are drafting different AI-assisted versions of the same client success story for different audiences, and the numbers don’t quite match. Walk me through how you’d set a cross-functional norm so that doesn’t happen again.Collaboration NormsEntry–leadership

AI assistance in deviation investigationsWalk me through which parts of a deviation investigation you’d let an AI assistant draft — say, summarizing batch history or proposing root-cause categories — and which parts you’d never hand off, like the final root cause conclusion or CAPA justification.Tool UsageEntry–leadership

Vision AI flags cosmetic defectsA vision-AI system flags cosmetic defects during in-process inspection of filled vials. Walk me through how you’d use that tool’s output and where a human quality decision still has to sit downstream.Tool UsageEntry–senior

AI cited wrong ICH guidelineYou asked an AI assistant to summarize the stability data section for an ANDA submission and it cited an ICH guideline that doesn’t actually apply to the dosage form. How would you catch that before it goes to regulatory affairs for filing?JudgmentMid–leadership

Verifying AI root-cause suggestionsAn AI-assisted investigation tool suggests a probable root cause for an out-of-specification lab result that sounds plausible but doesn’t match the equipment log. Walk me through how you’d verify or reject that suggestion before it goes into the OOS report.JudgmentEntry–mid

FDA asks about AI useAn FDA investigator asks during an inspection whether AI tools were used anywhere in preparing the batch record review or the CAPA narrative in front of them. How do you answer, and what do you need to have ready to back it up?Stakeholder CommunicationSenior–leadership

AI drafted executive summary flaggingYou used an AI tool to draft the executive summary of an annual product review before your quality reviewer signs off on it. How do you flag that to the reviewer so their sign-off reflects an informed decision, not a rubber stamp?Stakeholder CommunicationEntry–senior

CAPA effectiveness check failedAn internal audit finds that a CAPA effectiveness check written with AI assistance recommended a monitoring period too short to catch a recurring deviation. Walk me through how you’d handle the finding and what changes to your process afterward.Error HandlingMid–leadership

AI mis-transcribed batch record weightDuring digitization, an AI tool mis-transcribes a weight value from a paper batch record into the electronic system, and it isn’t caught until a later reconciliation. Walk me through what you’d do once you find it and how it changes your reconciliation approach.Error HandlingEntry–mid

Staying current on AI capabilitiesHow do you stay current on what AI-assisted tools can and can’t be trusted for in a GMP environment, given how fast the tools change and how slowly validated systems change?UpskillingEntry–mid

Training investigators on AI toolsYou’re rolling out an AI-assisted root-cause tool to a team of deviation investigators. Walk me through how you’d train them so the tool sharpens their investigations instead of letting them accept the first plausible answer it produces.UpskillingMid–leadership

Agentic deviation routing and classificationYou’re considering an agentic tool that can read incoming deviation reports, classify severity, and route them to the right investigator automatically. Walk me through what oversight you’d build in before letting it run unsupervised.Agentic WorkflowsEntry–senior

Agentic manufacturing line pausesAn agentic monitoring system watches real-time process data from a continuous manufacturing line and can pause the line on its own if it detects a trend excursion. Walk me through how you’d decide what it’s allowed to act on versus what it just has to flag.Agentic WorkflowsEntry–senior

Confidential data in public chatbotsA formulation scientist wants to paste process parameters and impurity profiles into a public AI chatbot to help troubleshoot a scale-up issue. Walk me through how you’d handle that request.Governance RiskEntry–mid

AI audit trail review validationYour site wants to use an AI tool to help review audit trails in the electronic batch record system for anomalies. What has to be true about that tool’s own validation and audit trail before it’s allowed near a GMP record?Governance RiskEntry–leadership

Reviewing AI-assisted CAPA narrativesWhat team norm would you set for reviewing any deviation or CAPA narrative that had AI assistance in its drafting, before it goes to the quality unit for approval?Collaboration NormsEntry–mid

Cross-functional AI governance normsQuality, manufacturing, and IT each have a different comfort level with adopting AI tools on the shop floor. Walk me through how you’d get them to a shared set of norms for what’s allowed near GMP systems.Collaboration NormsEntry–senior

AI media lists and pitchesWalk me through how you use AI tools when building a media list and pitch angles for a launch, and where you still rely on your own read of a reporter’s beat rather than what the tool surfaces.Tool UsageEntry–leadership

Automated sentiment monitoring accuracyYour media monitoring platform tags sentiment automatically across thousands of mentions during a launch. Walk me through how you decide when the automated read is good enough to report to the client and when you pull a human read instead.Tool UsageMid–leadership

Catching fabricated CEO quotesAn AI drafting tool generates a press release quote attributed to your CEO that sounds plausible but the executive never actually said those words. Walk me through how you’d catch that and what you’d change before it ships.JudgmentEntry–leadership

Verifying AI competitor summariesYou ask an AI tool to summarize a week of competitor coverage and recommend a share-of-voice narrative for your client’s board deck. What do you personally verify before that summary goes in front of the board, and why?JudgmentSenior–leadership

Disclosing AI assistance to clientsA client asks whether the crisis holding statement you sent them was drafted with AI assistance. Walk me through how you answer that, and how it changes if the matter is headed toward litigation.Stakeholder CommunicationEntry–leadership

Executive briefing on AI draftsYou’re prepping an executive for an all-hands where several talking points were AI-drafted from prior interview transcripts. How do you brief the executive so nothing sounds unfamiliar when they’re asked a follow-up on stage?Stakeholder CommunicationEntry–leadership

Responding to wrong embargo datesAn AI-assisted pitch tool auto-populated the wrong embargo date on a batch of releases to your media list, and two outlets have already published early. Walk me through what you do in the next hour.Error HandlingEntry–leadership

Retro on missed sentiment escalationYour AI sentiment-monitoring system failed to flag an escalating complaint thread until it was already trending, and the client found out from a reporter before you called them. Walk me through the retro you’d run and what changes as a result.Error HandlingSenior–leadership

Staying current on AI detectionHow do you keep your own skills current on how AI-generated or AI-assisted language gets detected and perceived by journalists and audiences, and how has that changed the way you draft?UpskillingEntry–leadership

Preserving voice with AI draftingWalk me through how you’d bring a comms team up to speed on using AI drafting tools without letting institutional voice and craft erode across the account roster.UpskillingEntry–leadership

Social media agent reply guardrailsYou set up an AI agent to monitor live social mentions during a product launch and draft first-response replies. Walk me through the guardrails between what the agent drafts and what actually gets posted under the brand’s name.Agentic WorkflowsEntry–leadership

Checkpoints for end-to-end AI workflowsYou’re considering letting an AI agent handle research, first-draft messaging, and target-list building end to end for a major announcement, with you reviewing only the final package. Where do you insert checkpoints, and what’s the one step you’d never let the agent run unsupervised?Agentic WorkflowsSenior–leadership

Unannounced merger leaked to chatbotA junior teammate pastes details of an unannounced merger into a public AI chatbot to help draft the announcement. What’s the exposure, and how do you handle it once you find out?Governance RiskEntry–leadership

AI and fair disclosure rulesYou’re using AI tools to help draft materials for a publicly traded client ahead of an earnings call. What governance do you put in place so nothing drafted with AI risks a fair-disclosure violation before the numbers are public?Governance RiskEntry–leadership

Press release review chainWalk me through your team’s review chain for a press release that started as an AI-generated first draft, from that draft to the version that actually goes out to media.Collaboration NormsEntry–leadership

Crisis team disagrees on AITwo senior members of your crisis team disagree about whether AI tools should touch any drafting on a currently-active reputational crisis account. How do you resolve that and set the norm going forward?Collaboration NormsMid–leadership

Using AI for listing descriptionsWalk me through how you use AI when drafting a listing description — from pulling property details to publishing copy — and where in that sequence you deliberately keep a human check.Tool UsageEntry–leadership

AI-powered valuation and broker opinionsDescribe how you’d use an AI-powered comps or automated valuation tool to build a broker’s opinion of value, and which parts of that valuation you still do by hand.Tool UsageEntry–leadership

Stale comps in market analysisAn AI-generated comparative market analysis adjusts a subject property upward because it used a comp that sold six months ago in a market that’s since cooled. How do you catch that before you present a listing price to the seller?JudgmentEntry–leadership

Lease abstract misses co-tenancyYou’re reviewing an AI-generated lease abstract for a commercial acquisition and it summarizes the renewal terms but leaves out a co-tenancy clause buried in an exhibit. What’s your process for catching gaps like that before the abstract goes into the deal model?JudgmentEntry–leadership

Disclosing AI valuation to sellersA seller asks whether the price recommendation you gave them came from you or from an AI valuation tool. How do you answer that, and how does it change what you say next?Stakeholder CommunicationEntry–leadership

AI-assisted property disclosuresWalk me through how you’d use AI to help draft a property disclosure statement, and what you tell the buyer about which parts were AI-assisted versus verified firsthand.Stakeholder CommunicationEntry–leadership

Responding to fair housing violationsAn AI-drafted listing description you published turns out to include phrasing a fair housing attorney flags as steering language. Walk me through what you do in the hours and days after that’s identified.Error HandlingEntry–leadership

AI pro forma underestimated expensesYou discover after closing that an AI-assisted pro forma underestimated operating expenses on an acquisition because it pulled a comparable’s expense ratio instead of the subject property’s actual trailing costs. What do you do, and what changes in your underwriting process going forward?Error HandlingEntry–leadership

Training agents on AI toolsAs a team lead, how are you upskilling agents on your team to use AI valuation and comp tools well, and how do you tell whether it’s working?UpskillingMid–leadership

Tenant screening rules keep changingAI-powered tenant screening tools and the fair housing rules around them keep changing. How do you stay current, and how do you make sure the rest of your property management team does too?UpskillingEntry–leadership

Agentic comp and valuation workflowDescribe how you’d design an agentic workflow that runs comp searches and drafts a first-pass broker’s opinion of value autonomously across a portfolio, including where you’d insert mandatory human checkpoints.Agentic WorkflowsSenior–leadership

Guardrails for portfolio lease abstractionYou’re rolling out an AI agent to abstract leases across an entire acquired portfolio instead of one file at a time. What guardrails do you put around it, and how do you decide when it’s earned less oversight?Agentic WorkflowsEntry–leadership

Approving AI listing description toolYour brokerage is evaluating an AI copywriting tool for listing descriptions at scale. As the person responsible for compliance, what do you need to see before you approve it for the whole team?Governance RiskSenior–leadership

Screening tool disparate impactHow do you evaluate whether an AI tenant screening tool is producing a disparate impact on protected classes, and what do you do if you find one?Governance RiskEntry–leadership

Pro forma review before committeeWhat team norm do you set for reviewing AI-assisted pro formas before they go in front of an investment committee, and how do you enforce it without slowing every deal down?Collaboration NormsMid–leadership

Publishing AI-drafted property contentWhat’s your team’s norm for when an agent can publish AI-drafted property content directly versus when a licensed team member has to personally verify and take ownership of it first?Collaboration NormsEntry–leadership

Literature review with AI assistantYou're scoping a literature review with an AI research assistant that can search and summarize abstracts across academic databases. Walk me through which parts of the review you hand to it and which parts you insist on doing yourself.Tool UsageEntry–leadership

Survey analysis speed versus depthA client wants topline results within 48 hours of survey close, and your AI tool can auto-generate a full analysis write-up from the raw dataset. Where do you use it, and where do you slow down and do the analysis by hand?Tool UsageEntry–leadership

Citation doesn't match any paperAn AI drafting tool hands you a literature review paragraph with a citation that reads perfectly but doesn't match any paper you can find. Walk me through how you catch this before it goes into a report.JudgmentEntry–leadership

Causal language in correlational outputAn AI summary of your regression output describes a predictor as 'driving' the outcome, when your model only supports a correlational claim. How do you catch language like that before it reaches a client deck?JudgmentEntry–leadership

Disclosing AI use to clientsA client asks whether any part of their custom report — the executive summary, the findings, the recommendations — was drafted by AI. Walk me through how you answer, and what you'd have done differently if you expected the question.Stakeholder CommunicationEntry–leadership

Framing AI contribution in meta-analysisYou're presenting synthesized findings from an AI-assisted meta-analysis to a funder who wants to know how much of the conclusion is machine-generated versus human-verified. How do you frame that answer?Stakeholder CommunicationSenior–leadership

Data cleaning script dropped responsesTwo weeks after a report ships, you discover the AI-assisted data-cleaning script silently dropped a batch of valid survey responses, changing a topline number. Walk me through what you do next.Error HandlingEntry–leadership

Miscoding in AI qualitative analysisAn AI tool you used to code open-ended qualitative responses grouped several distinct themes under one label, and that miscoding shaped a published finding. How do you handle it once you catch it?Error HandlingEntry–leadership

Evaluating new AI literature toolsNew AI literature-synthesis tools keep shipping features that promise to replace parts of a systematic review. How do you decide which ones are worth learning versus which ones you skip?UpskillingEntry–leadership

Training juniors on AI limitationsYou're bringing junior researchers onto a team that now uses AI heavily for early-stage drafting and coding. Walk me through how you'd train them to know when the AI's output can't be trusted yet.UpskillingSenior–leadership

Checkpoints for multi-step AI agentYou give an AI agent a multi-step task — search the literature, extract study characteristics into a table, and flag studies that meet inclusion criteria — with minimal supervision. Where do you put checkpoints before you'd trust the output?Agentic WorkflowsEntry–leadership

Survey pipeline with sign-offsWalk me through how you'd set up an agentic pipeline that takes a survey dataset all the way from raw responses to a drafted insights report, and where a human has to sign off before it moves to the next step.Agentic WorkflowsSenior–leadership

Identifiable data in AI analysisA survey dataset contains identifiable respondent information, and you want to use an AI tool to help spot patterns in the open-ended comments. Walk me through how you'd decide whether that's allowed and how you'd set it up.Governance RiskEntry–leadership

Proprietary data and general-purpose AIA client's proprietary dataset is licensed for use only by your firm, and someone on your team wants to run it through a general-purpose AI tool to speed up analysis. How do you handle that request?Governance RiskEntry–leadership

AI paraphrasing and inter-rater reliabilityYour team is split on how much AI-assisted paraphrasing is acceptable when coding qualitative transcripts for a study with inter-rater reliability requirements. How do you resolve it?Collaboration NormsEntry–leadership

AI policy for deliverablesYou're asked to set the team's standard for where AI is allowed to touch a research deliverable — literature review, analysis, write-up — before it ships to a client. Walk me through what that policy looks like.Collaboration NormsSenior–leadership

Opening order without demand forecastYou're setting the initial buy quantity for a new-to-world product with no sales history for the demand-forecasting model to learn from. Walk me through how you'd set that opening order without leaning on the algorithm.Tool UsageEntry–leadership

Catching blind markdown recommendationsYour markdown-optimization tool recommends a deep discount on a product line the week before a competitor's planned major sale event, which it has no visibility into. How do you catch that before the markdown goes live?JudgmentEntry–leadership

AI-generated planogram rollout communicationA planogram you're rolling out to stores was generated by an AI space-planning tool with minimal manual adjustment. How do you communicate that to store managers who are being asked to reset their floor around it?Stakeholder CommunicationEntry–leadership

Incorrect AI-assisted price shipped liveAn AI-assisted pricing update pushed an incorrect price to your e-commerce site and a handful of orders went through before anyone caught it. Walk me through what you do in the first hour, and what changes after.Error HandlingEntry–leadership

Closing the AI skill gapYou're leading a merchandising and planning team where half the group is fluent with the new AI forecasting and assortment tools and half is still working entirely off spreadsheets and gut instinct. How do you close that gap?UpskillingMid–leadership

Autonomous replenishment guardrailsYou're evaluating an agentic replenishment system that can autonomously place reorders for fast-moving SKUs without a human confirming each one. What guardrails do you require before letting it write purchase orders on its own?Agentic WorkflowsMid–leadership

Vetting loyalty dataYour personalization engine uses loyalty program purchase history to tailor product recommendations and marketing. What do you check before feeding that data into a new AI-powered recommendation model?Governance RiskEntry–leadership

Reviewing AI-generated contentYour team now uses AI to draft product descriptions and tune search-relevance rules at scale. What's the review standard before either of those ships to the live site?Collaboration NormsEntry–leadership

Scoping visual search appropriatelyCustomers can now use a visual search tool to find products from an uploaded photo. Where do you rely on that tool, and where do you deliberately keep it out of the loop?Tool UsageEntry–leadership

Handling AI theft-detection false positivesYour loss-prevention system uses computer vision to flag potential theft on the sales floor, and it's just flagged an employee for a pattern the model reads as suspicious. How do you handle that before anyone acts on it?JudgmentEntry–leadership

Disclosing AI-driven orders to vendorsYou're heading into a vendor negotiation where the allocation and order quantities you're proposing were set largely by an AI planning tool. How much of that do you disclose to the vendor, and why?Stakeholder CommunicationSenior–leadership

AI product misinformationA customer tells you the AI shopping assistant on your site told them a product was machine-washable when it isn't, and they've already ruined it following that advice. Walk me through how you respond, and what you fix upstream.Error HandlingEntry–leadership

Expanding agentic pricing automationYou're deciding whether to expand agentic pricing and markdown automation from a pilot category to the entire chain. What has to be true before you approve that, and what stays under human control regardless?Agentic WorkflowsSenior–leadership

Vetting competitor data scraping toolsA team member wants to use an AI tool to scrape and summarize competitor pricing and assortment data at scale to inform your buying decisions. What do you need to check before that becomes a standard practice?Governance RiskMid–leadership

Resolving forecast versus gut disagreementsIn the weekly buy meeting, a planner's gut read on a category disagrees with the AI forecast, and the room has to decide which number goes into the order. What's the norm you set for resolving that?Collaboration NormsEntry–leadership

Overriding AI planogramsThe AI assortment and space-allocation tool recommends a standard planogram for a new small-format store, but you know the tool was trained almost entirely on full-size store data. Do you override it, and how do you decide?JudgmentMid–leadership

When to use pair-programming assistanceWalk me through how you decide when to let an AI pair-programming tool write a function outright versus when you write it yourself line by line — say, generating a REST endpoint handler versus implementing token validation in your auth middleware.Tool UsageEntry–leadership

Standardizing AI coding toolsYou lead a small platform team. How do you decide which AI coding tool to standardize on for autocomplete versus for larger agentic tasks like multi-file refactors, and how do you revisit that choice as the tooling landscape shifts every few months?Tool UsageMid–leadership

Reviewing AI-generated caching logicYou're reviewing a pull request where the author used an AI assistant to write a caching layer. The code looks clean and the tests pass, but something about the invalidation logic feels off. What do you actually check before approving it?JudgmentEntry–senior

Unfamiliar package suggested by copilotAn AI coding assistant suggests importing a package to solve a problem, and the package name looks plausible but you've never heard of it. Walk me through what you do before it ends up in a dependency file.JudgmentEntry–leadership

Enforcing AI-generation disclosure in PRsYour team requires PR authors to note when a change was significantly AI-generated. Why does that norm matter to you, and how do you handle it when a teammate's PR clearly used heavy AI assistance but doesn't disclose it?Stakeholder CommunicationEntry–mid

Communicating AI-generated code incidentsA production incident traces back to code your team shipped that was largely AI-generated and passed review. How do you communicate that root cause to leadership and the affected customers without it reading as either an excuse or an overcorrection to ban the tools?Stakeholder CommunicationSenior–leadership

AI diagnosis during production incidentYou're on call and a service is throwing errors in production. You ask an AI copilot to help diagnose it and it confidently proposes a root cause and a fix. What's your process before you act on that, given the pressure to resolve the incident fast?Error HandlingEntry–mid

AI code accepted blindlyIn a postmortem, you find the root cause was that an engineer accepted an AI-suggested code change without fully understanding it, and it shipped a race condition. How do you write that finding up, and what changes to process rather than to the individual?Error HandlingEntry–senior

Maintaining debugging skillsHow do you make sure you're still building real debugging and system-design skill when an AI assistant can generate most of your day-to-day code, and how do you know if you're starting to lose that muscle?UpskillingEntry–mid

Coaching AI-dependent juniorsYou're mentoring a junior engineer who leans heavily on an AI assistant and can ship working code quickly, but struggles when you ask them to explain why it works. How do you coach that without slowing them down or making them defensive about using the tools?UpskillingMid–leadership

Scoping agentic code migration tasksYou're asked to use an agentic coding tool to migrate a legacy module to a new framework across dozens of files. Walk me through how you scope that task for the agent and what you check before merging its output.Agentic WorkflowsEntry–mid

Guardrails for autonomous code remediationYour org wants to let an autonomous coding agent auto-remediate certain failing CI checks and, eventually, some on-call alerts without a human in the loop each time. What guardrails do you insist on before agreeing to that, and where do you draw the line on full autonomy?Agentic WorkflowsSenior–leadership

Detecting AI-generated licensed codeAn AI assistant generates a function that looks suspiciously close to a snippet you recognize from an open-source project with a restrictive license. What do you do before that code goes into your company's proprietary codebase?Governance RiskEntry–mid

Securing code in AI toolsHow do you set and enforce a policy on what code and data engineers can paste into external AI coding tools, given that your codebase includes proprietary algorithms and some customer data flows through it?Governance RiskMid–leadership

Review standards for AI-generated codeYour team is debating whether AI-generated code should require more reviewer sign-off than hand-written code, or whether that's an unnecessary double standard. Where do you land, and how would you actually implement it?Collaboration NormsEntry–mid

Scaling AI adoption unevenlyYou're setting AI tool usage standards across a fast-growing engineering org where teams have wildly different comfort levels — some already delegate heavily to agents, others barely use autocomplete. How do you build a policy that raises the floor without capping the teams already ahead?Collaboration NormsSenior–leadership

Trusting AI for analytics queriesYou're pulling a growth metric for a quarterly review and an AI assistant can draft the SQL against the events warehouse in seconds. Walk me through which parts of that query you'd let it write outright, and where you'd insist on writing the logic yourself.Tool UsageEntry–leadership

Trust boundaries for copilot suggestionsYour team ships to a service handling millions of requests a day, and everyone's using an AI coding copilot. Walk me through where you'd trust its suggestions without a second look, and where you'd slow down regardless of how confident it sounds.Tool UsageEntry–leadership

Validating AI-generated experiment claimsAn AI-generated experiment summary tells you your new ranking model 'caused' a lift in session length, based on a single A/B test. Walk me through how you'd check that claim before it goes into a launch decision.JudgmentEntry–leadership

Checking AI analysis for biasAn AI-assisted review of your content moderation classifier's false-positive rate looks clean in aggregate, but you suspect it's uneven across language communities. Walk me through how you'd dig in before trusting the aggregate number.JudgmentSenior–leadership

Explaining unexplainable personalizationA user asks your support team why they stopped seeing a particular type of content in their feed, and the honest answer involves a personalization model none of you can fully explain. Walk me through how you'd respond to them.Stakeholder CommunicationEntry–leadership

Tracking AI training data usageLegal asks whether user messages have ever been used to train or fine-tune one of your product's AI features, ahead of a regulatory inquiry. Walk me through how you'd track down a defensible answer.Stakeholder CommunicationMid–leadership

Responding to gamed experiment rolloutAn AI agent auto-rolled out a personalization experiment to 100% of traffic overnight after a green early read, and it turns out the metric was gamed by a bot cluster. Walk me through what you'd do in the first hour after catching it.Error HandlingEntry–leadership

Handling mass moderation false positivesYour AI moderation model mass-removed a batch of posts that turn out to be a false-positive spike from a model update, and creators are publicly noticing. Walk me through how you'd handle the next 24 hours.Error HandlingEntry–leadership

Aligning on AI analyticsHalf your growth analysts have started using AI to draft experiment hypotheses and interpret results, and half are skeptical it just produces plausible-sounding noise. Walk me through how you'd get the team to a shared, disciplined way of using it.UpskillingEntry–leadership

Supervising AI triageYou're introducing an AI-assisted triage tool to a trust and safety operations team that's used to manual review queues. Walk me through how you'd train them to supervise it well instead of just rubber-stamping its output.UpskillingEntry–leadership

Guardrails for autonomous anomaly pausingYou're piloting an AI agent that can detect anomalies in your event pipeline and automatically pause the affected experiment. Walk me through the guardrails you'd want in place before you let it act without a human approving each pause.Agentic WorkflowsEntry–leadership

Human checkpoints in data deletionYou're considering letting an AI agent handle end-to-end data deletion requests, from verifying the user's identity to confirming deletion across every downstream system. Walk me through where you'd still require a human checkpoint.Agentic WorkflowsSenior–leadership

Pasting user data into AIAn engineer wants to paste a sample of real user session data into an AI assistant to debug a personalization bug faster. Walk me through how you'd decide whether that's acceptable and what you'd have them do instead if not.Governance RiskEntry–leadership

Vetting third-party AI model rightsYou're about to ship a third-party AI model as part of your ranking pipeline, and legal wants to know what happens if it was trained on data your company doesn't have rights to. Walk me through how you'd build confidence before launch.Governance RiskMid–leadership

Reviewing AI pull requestsYour engineering org is debating whether AI-generated pull requests need the same review bar as human-written ones. Walk me through the norm you'd argue for and why.Collaboration NormsEntry–leadership

Evaluating AI-suggested experimentsProduct and growth teams are using AI to generate a stream of new experiment ideas, and some of the suggestions nudge toward engagement tactics that feel manipulative. Walk me through the norm you'd set for evaluating AI-suggested experiments before they ship.Collaboration NormsEntry–leadership

Trusting AI routing and dispatchWalk me through how you decide when to trust an AI-generated route and dispatch plan as-is versus when you'd override it during a live shift.Tool UsageEntry–leadership

Automating demand planning forecastsWhich parts of your demand-planning process do you let an AI forecasting model own outright, and which do you deliberately keep manual?Tool UsageMid–leadership

Validating AI shipment ETAsA shipment-tracking AI tells you a container will clear customs on time, but you know the terminal is congested from a recent storm. Walk me through how you'd validate the AI's ETA before it goes out to the customer.JudgmentEntry–leadership

Validating an AI cost-cutting recommendationAn AI network-design tool recommends consolidating two distribution centers to cut cost. What real-world constraints would you check before taking that recommendation to your VP?JudgmentSenior–leadership

Explaining AI-scored carrier bid rankingsYou've started using an AI tool to score and rank carrier bids. How would you explain that scoring to a carrier who's upset about losing a lane they've run for years?Stakeholder CommunicationEntry–leadership

Auditing AI classificationA customs auditor asks how much of your HTS classification work is AI-assisted versus human-reviewed. Walk me through how you'd answer that question.Stakeholder CommunicationEntry–leadership

AI dispatch temperature failureAn AI dispatch system auto-assigned a temperature-sensitive shipment to a truck without refrigeration, and it wasn't caught until the load was already out. Walk me through what you'd do in the first hour.Error HandlingEntry–leadership

Diagnosing systematic AI under-orderingYour AI-driven inventory replenishment system has been systematically under-ordering a SKU for three weeks, causing repeated stockouts. How would you diagnose and fix this?Error HandlingEntry–leadership

Building dispatcher trustHow would you get a team of experienced dispatchers, who've built their routing instincts over years, to actually trust and use a new AI routing recommendation engine?UpskillingEntry–leadership

Evaluating supply chain vendorsHow do you evaluate whether a new AI-powered supply chain planning vendor is worth adopting versus hype, and how do you build that evaluation muscle in your team?UpskillingMid–leadership

Guardrails for autonomous shipment rebookingYou're rolling out an agentic tool that can automatically rebook a carrier and adjust the delivery promise when a shipment exception is detected. What guardrails would you put around what it's allowed to do without a human in the loop?Agentic WorkflowsEntry–leadership

Overseeing AI customs documentation filingWalk me through how you'd design oversight for an AI agent that pulls together and files customs entry documentation across multiple shipments without a human touching each one.Agentic WorkflowsMid–leadership

Protecting customer dataYour fleet telematics and shipment data include customer-identifying information. What would stop you from feeding that data into a general-purpose AI tool to get a quick answer?Governance RiskEntry–leadership

Detecting carrier model biasAn AI carrier-scoring model has started consistently down-ranking carriers based in a particular region. How would you investigate whether that's a legitimate performance signal or a bias problem?Governance RiskEntry–leadership

Reviewing AI load plansWhat review standard would you set for a load plan that was generated by AI before it goes out to drivers?Collaboration NormsEntry–leadership

Aligning misaligned AI planning toolsYour demand planning team and your warehouse ops team are using AI tools that don't talk to each other, and their numbers have started to diverge. How would you get the teams aligned?Collaboration NormsEntry–leadership

Overriding day-ahead load forecastsYour day-ahead load forecasting model runs on an AI/ML pipeline that ingests historical consumption, weather, and calendar data. Walk me through a scenario where you’d override or set aside the model’s forecast entirely rather than trust it.Tool UsageEntry–leadership

Discounting renewable energy forecastsYou use an AI model to forecast next-day solar and wind output for unit commitment planning. Describe the specific conditions under which you’d discount that forecast and how much lead time you’d build in as a result.Tool UsageEntry–leadership

Validating transmission warningsAn AI-based contingency analysis tool recommends rerouting load away from a transmission line it’s flagged as at risk of overload. Walk me through what you check before you act on that recommendation.JudgmentSenior–leadership

Failed predictive maintenanceA predictive-maintenance model rated a distribution transformer as low risk two months before it failed and caused a multi-hour outage. Walk me through the post-incident review you’d lead.Error HandlingEntry–leadership

Skeptical operators adopting AIYour control room is rolling out an AI-based anomaly detection tool on top of SCADA telemetry. How would you get a team of veteran operators, some skeptical of it, actually proficient with it?UpskillingEntry–leadership

Overseeing autonomous demand response agentsYou’re designing a demand response program where an AI agent monitors forecasted peak load and can autonomously send curtailment signals to enrolled commercial customers. What oversight would you build into that agent before letting it act without a human in the loop?Agentic WorkflowsMid–leadership

AI copilot on OT networkA vendor is pitching an AI copilot that would sit on top of your SCADA/OT environment to help operators query historical telemetry faster. What would you need to see before approving it anywhere near that network?Governance RiskEntry–leadership

Reviewing AI restoration estimatesYour team started using an AI tool to generate estimated restoration times during outages, and those estimates go out to customers and dispatch. What review norm would you put in place before an estimate is published?Collaboration NormsEntry–leadership

AI sequence conflicts with fieldDuring a major storm, an AI model generates a prioritized outage restoration sequence based on customer count and criticality. A field supervisor tells you the sequence doesn’t match what crews are actually seeing on the ground. How do you handle that conflict in the moment?JudgmentSenior–leadership

Correcting AI-drafted NERC CIP documentationYou used an AI tool to help draft evidence documentation for a NERC CIP compliance audit, and after submission you discover it mischaracterized a control that wasn’t actually in place. Walk me through what you do next.Error HandlingEntry–leadership

Resistant engineers adopting AIYour distribution engineering team has relied on manual line-loss analysis for years, and you want to move them to an AI-assisted workflow. Two senior engineers are openly resistant. How do you bring them along?UpskillingEntry–leadership

AI agent work order authorityAn AI agent reviews drone and LiDAR imagery of transmission corridors and automatically opens work orders for vegetation clearance or hardware defects it detects. What boundaries would you put around its authority to open those work orders on its own?Agentic WorkflowsEntry–leadership

Data pooling with external vendorsA load forecasting vendor’s AI model is trained partly on aggregated data from other utility customers, and it’s being pitched as more accurate because of it. What would you want resolved before your operational data becomes part of that pool?Governance RiskEntry–leadership

Sign-off on dispatch recommendationsGrid operators and the data science team both touch AI-driven dispatch recommendations before they reach the control room. Where would you draw the line on who owns final sign-off, and why?Collaboration NormsMid–leadership

Disclosing AI in regulatory filingsYou used an AI tool to draft the reliability-metrics narrative in a filing to the state public utility commission. Would you disclose that AI was involved in drafting it, and how would you decide?Stakeholder CommunicationEntry–leadership

Human review of curtailment signalsA customer on your demand response program asks whether a human reviewed the AI-generated curtailment signal that shut down part of their operation last week. How do you answer, and what would make that answer credible?Stakeholder CommunicationEntry–leadership

AI in deal sourcing funnelWalk me through which parts of your deal-sourcing funnel you let an AI tool touch — flagging companies, drafting outreach — and which parts you keep fully manual, and why.Tool UsageEntry–leadership

Manual vs. AI-populated model cellsWhen an AI copilot helps you build out a financial model — an LBO build or a SaaS metrics model — which cells or assumptions do you always enter by hand, and which do you let it populate?Tool UsageEntry–leadership

Verifying AI contract summariesAn AI tool summarized fifty pages of a target's customer contracts and flagged no unusual termination clauses. Walk me through how you'd verify that before it goes into the diligence memo.JudgmentEntry–leadership

Reconciling AI market sizing estimatesYou ask an AI research tool to size a target's addressable market and it returns a number well above what your own bottoms-up sense suggests. Walk me through how you'd reconcile that before it lands in the investment memo.JudgmentSenior–leadership

Disclosing AI-drafted memo sectionsHow do you disclose to your investment committee that part of a memo — say, the competitive landscape section — was AI-drafted, and does that change how they should read it?Stakeholder CommunicationEntry–leadership

AI in valuation methodology disclosureAn LP asks whether AI tools touched the valuation methodology behind a portfolio company's reported mark. Walk me through how you'd answer that honestly without undermining their confidence in the number.Stakeholder CommunicationEntry–leadership

Incorrect AI-generated comp setYou discover after an investment committee meeting that an AI-assisted comp set in the memo included a company that isn't actually a comparable business. Walk me through what you do next.Error HandlingEntry–leadership

AI recommendation hurt a customerA portfolio company's leadership team acted on AI-generated pricing recommendations you'd shared as a discussion starting point, and it hurt a customer relationship. Walk me through how you handle that.Error HandlingSenior–leadership

Evaluating new AI diligence toolsHow do you evaluate whether a new AI-powered diligence or sourcing tool is actually worth adopting into your workflow versus being noise?UpskillingEntry–leadership

Teaching AI tool disciplineYou're onboarding a new analyst who's used to letting an AI tool draft first-pass company summaries. What do you teach them about using it well from day one?UpskillingEntry–mid

Supervising AI filing research agentsWalk me through how you'd set up and supervise an AI agent tasked with pulling public filings, news, and litigation records on a target across multiple sources — where do you check its work?Agentic WorkflowsEntry–leadership

AI monitoring portfolio company signalsYou've set an AI agent to continuously monitor your portfolio companies' public signals — hiring pages, review sites, news — and alert you to red flags. Walk me through what happens when it flags something, and what happens when it misses something.Agentic WorkflowsSenior–leadership

Confidential data in AI toolsA target company shares a data room full of confidential financials for diligence. What's your rule for what does and doesn't go into an AI tool, and why?Governance RiskEntry–leadership

Deal signal leakage riskWalk me through how you think about the risk that an AI tool trained or fine-tuned on your firm's historical deal data could leak signal about a live, undisclosed deal.Governance RiskMid–leadership

Flagging AI assistance in memosWhat's your team's norm for flagging which parts of an investment memo were AI-assisted when a partner or IC member reviews it?Collaboration NormsEntry–leadership

Addressing suspected AI-generated summariesA junior analyst hands you a diligence summary that reads well but you suspect was mostly AI-generated with light editing. Walk me through the conversation you have with them.Collaboration NormsEntry–leadership