AI·Personal working standard · used on Use Summary Table & ARS Automation
AI Pre-Flight Checklist
A portable checklist for shipping agentic AI — distilled from building and deploying agents in regulated life sciences.
| Before | Now | |
|---|---|---|
| Governance | Reinvented on each agent | One checklist, about 30 items |
| Applied to | — | Use summary and ARS |
| What it forces | A slide | A kill switch, an owner, an audit trail |
Problem
When you ship an agent, the same questions keep coming back: can you stop it, who reviews the output, what did it cost, who owns it, and did you check for prompt injection? Without a shared checklist, every project reinvented those answers — or skipped them.
Value
A reusable Pre-Flight board I can run before a pilot or a broader rollout. It turns governance into concrete work (a kill switch in the repo, an audit trail, a named owner) instead of a slide. I have already applied it to the use-summary extractor and ARS Automation.
How I built it
I did not create an official corporate standard from scratch. I built a generic, portable version from shipping agentic tools in regulated workflows — turning Agentic AI / AI Governance practice and GenAI learning into a personal checklist with placeholders for whichever organisation you are in (risk gate, identity, catalog, help channel).
Deployment
Interactive checklist on this site, password-protected. Also available as a private Claude Code skill (/ai-preflight) against real repos. A practitioner synthesis, not an official policy document. Reference apps: Use Summary Table and ARS Automation.
Stack
- AI governance practice
- HITL · kill switch · audit
- Portable across employers
- Agent Card / sign-off
- Claude Code skill