When not to use agents
Your org's culture has swung to 'AI-first': there's subtle pressure to use an agent for every task, and engineers feel they're underperforming if they hand-write code. As a staff engineer, articulate the categories of work where reaching for an AI agent is actively the wrong call on your team — not because it can't produce output, but because the output is a liability. Be specific about the failure modes, and address the cultural pressure, not just the technical line.
classify_delegation(has_undocumented_rules: list[bool], is_security_boundary: list[bool], is_novel_design: list[bool], reviewer_can_evaluate: list[bool]) → list[str][[false],[false],[false],[true]]out["agent_ok"][[false],[true],[false],[true]]out["hand_write"][[true],[false],[false],[true]]out["agent_with_spec"]Treat the AI’s output as a draft to verify, not an answer to trust. Name the specific flaw and the input that triggers it, say how you’d catch it (tests, edge cases, reading critically), and how you’d re-prompt or decompose to get it right.
Vibe & agentic: describe the solution in plain language (or narrate it) and the coach grades your approach.