If unsure, ask. Never guess. — AI Agent Pre-Execution Checklist

A new AI agent pre-execution checklist proposes that humans declare intent and boundaries in natural language, which AI converts into executable JSON to operate autonomously but stop when uncertain, enabling safer physical AI deployment and clearer regulatory and legal responsibility.

Unknown = Stop When outside the boundary or uncertain, always stop and ask. Human declares in natural language → AI converts it into executable JSON → AI executes autonomously within that structure → AI stops at Unknown Humans define the two things AI should not invent: intent and boundary AI converts those declarations into structured execution conditions AI operates freely within those conditions When uncertain, it stops — it does not guess AI is free within declared intent and boundary Physical AI deployment becomes easier to scale Regulatory review becomes simpler Legal responsibility becomes easier to separate Non-developers can design AI behavior through natural language This is not “let’s control AI better.” This is“let’s build a structure where AI can operate freely — under clearly separated responsibility and executable conditions.” AI can run freely because there is a brake. You need brakes to drive on a highway. The title “If unsure, ask. Never guess.” looks like a constraint on AI. But it is actually the prerequisite for AI to act freely and responsibly in the real world.