ai-fluency-diligencelisted

# Diligence — owning the result Diligence is taking responsibility for *what* you do with an agent and *how*. It has three parts, and the bar rises with the stakes and the audience. ## The three parts - **Creation diligence** — Is this the right tool for the task? Did you share anything confidential or sensitive that you shouldn't have? Did the collaboration build your understanding, or bypass it? (If you can't explain what the agent did and why, you haven't understood it.) - **Transparency diligence** — Who needs to know AI was involved? Is the disclosure specific enough to be meaningful (which tool, which tasks), and appropriate to the context? - **Deployment diligence** — Did you verify the factual claims where stakes are high? Are you prepared to stand behind this as if you wrote it yourself? For agentic output, did you review the *trajectory*, not just the final result? ## Pre-share checklist - [ ] Appropriate tool for the task; no unauthorized sensitive data shared. - [ ] Everyone who needs to know AI was involved is identified. - [ ] Disclosure is specific (not just "AI helped") and context-appropriate. - [ ] High-stakes factual claims verified independently. - [ ] You can personally vouch for every part of the output. - [ ] For agentic output: the trajectory was reviewed (see `ai-fluency-discernment`). ## Disclosure — match the context | Context | Disclosure level | |---|---| | Academic / research publication | Full: which tool, which tasks, what revie