AI can draft updates, summarize threads, sketch designs, and generate test ideas. It cannot own your ethics, your team relationships, or production risk. Leads who treat models as interns with no judgment ship plausible nonsense at scale.
Your job is to set norms: where AI is encouraged, where it is banned, how people review output, and how you talk about confidentiality.
Where AI helps leads this week
- Turn rough notes into a clear status update you still verify.
- Summarize long RFCs or incident timelines for stakeholders.
- Generate interview rubrics or scorecard drafts you edit.
- Brainstorm options and failure modes before a design review.
- Create first-pass docs, then make a human accountable for accuracy.
Where AI must not decide
- Hiring yes/no without human evidence and debrief.
- Performance ratings, PIP language, or exit decisions.
- Security-sensitive design without expert review.
- Customer or legal commitments.
- Anything that needs production authority without tests and owners.
If you would not sign your name under the output, do not paste it into a doc, PR, or message. AI drafts. You own.
Team norms that prevent mess
Write a one-page AI use policy for the team. Cover secret handling, citation of AI-assisted work when it matters, review expectations for code, and examples of good use. Update it when tools change.
- Never paste secrets, credentials, or private personal data into untrusted tools.
- Prefer enterprise or approved tools for company code.
- Require tests and human review for AI-generated code paths.
- Teach juniors to question fluent answers, not worship them.
AI is a power tool. Power tools without training still take fingers.