Follow-up drafts from call notes
You're an account executive juggling forty open conversations. The prospects who said circle back next quarter slip through because your follow-ups are rushed and sound like everyone else's. Design an agent that reads your call notes and CRM history and drafts a follow-up worth sending. Tell me what data it touches, what stays in your hands before anything goes out, and how you'd prove the drafts are good before trusting them. Close with your one-minute pitch: problem, what you built, payoff.
score_followup_specificity(draft: str, prospect_terms: list[str], generic_phrases: list[str]) → int["Great talking about the SOC 2 audit deadline and the Okta migration. I will send the security packet and we can revisit in October, right after your board review.",["SOC 2 audit","Okta migration","board review"],["just checking in","touching base"]]out30["Just checking in to see if now is a better time. Touching base as promised. Let me know your thoughts.",["SOC 2 audit","Okta migration"],["just checking in","touching base","let me know"]]out0["You mentioned the Okta migration lands in Q3 and that renewals are the pressure point. Just checking in on whether a short call in October works.",["Okta migration","renewal","board review"],["just checking in"]]out15Turn the fuzzy goal into a bounded agent task. Say what the agent reads, what it may do on its own versus draft for a human, and what it must never touch. Then earn the trust: how you’d verify it before rollout, and the pitch (the problem, what you built, why it matters).
Agent build: design the agent in plain language (or narrate it). The coach grades the decomposition, the guardrails, the verification plan, and the pitch.