Agent Lab / AI Use Coach
AI helps. People stay accountable.
Bring a list of tasks and the coach will suggest where AI could help. It looks at the information each task uses and what could happen if the result is wrong. You’ll see a Green, Yellow or Red rating, with the rule behind it and any review needed.
The guide: 30 public rules, 40 team tasks with recommended answers, 22 trick cases and 21 points still to confirm with compliance.
Try it
- Green Use it and check the result.
- Yellow Use an approved tool, limit the data and name a reviewer.
- Red A person decides this.
Choose a recorded team example to see the coach’s recommendations and a manager’s review. No live AI needed.
Tasks AI can help with and tasks a person must handle
How the tasks compare
Under the hood
What the AI writes and the software checks
Written by Claude
- The two answers for each task: what goes in, and what happens if it’s wrong
- A suggested rating, with an explanation and links to the relevant rules
- How AI could help, and what to check
- Any part of a Red task that AI can help with separately
Checked by the software
- The tier can’t be lower than the guide’s tier table allows
- Unclear data counts as the stricter case
- A Red task never gets a suggestion for how AI could help
- Anything that decides something about a person is Red
- Cited rule numbers must exist in the guide
- Checks for personal data run before anything is sent
- Every Yellow task needs a named reviewer role
These checks are tested automatically whenever the Lab is updated.
A demonstration built from public rules and frameworks. Not legal advice or any company’s policy.