My 2023 AI rule is why my best dev went quiet
Your AI policy is a measurement, not a rulebook. The wider the gap between what it says and how your best people actually work, the more of them go silent. Mine hit 3 years wide before I noticed.
I wrote our AI guidance in 2023. It assumed AI drafts and a human ships. The tools became agents that run the task themselves, and the rule never moved. My strongest person stopped documenting her workflow because the stale rule made honesty look like going off-script.
The fix wasn't a tighter rule. It was asking her to show me how she actually works, with a promise it would make things safer.
What's the most out-of-date assumption still baked into your team's AI rules?
I like the distinction that policy should measure reality instead of trying to force it. Curious, what changed in your guidance once you started working with AI agents?
The stalest assumption in most AI policies is the 'human approves everything' gate. It assumes the AI suggests and the person decides, but the real productivity gain comes when you trust the agent enough to execute without a sign-off on every output. The smoothest agent deployments we see have policies that are guardrails not playbooks. They say 'dont go past these boundaries' instead of 'do it exactly this way.' It lets people adapt as the tools evolve faster than the rules can.