Know what you know. Mark what you don't.
You don't know if what AI told you is true, so you either trust it because it sounds confident, or distrust everything. Neither is useful.
Ask: 'Do I know enough to evaluate this?'
You have expertise — verify the specific claims.
Treat the output as a starting hypothesis, not an answer.
When passing output downstream, flag what's unverified.
Understand the domain before shipping unfamiliar work.
You're honest about what you know. You don't pretend AI gave you expertise you don't have. You manage uncertainty instead of being paralyzed by it.
Drafting a policy brief on tax law changes. You understand the operational impact — that's your domain. But legislative details? Use AI to help synthesize, but mark uncertainty: 'We recommend legal review before relying on this for operational decisions.' That honesty protects your credibility.