Agent Lab / First week with your AI assistant
Practice catching the AI before it counts.
Try the kinds of AI tasks a new health plan employee might encounter. You’ll write a request, check a draft for mistakes and decide when a person needs to step in. The coach gives feedback after you’ve had a go, with suggestions for what to practice next.
The content: 12 design rules, a five-step path, 10 error types, 9 practice rounds with 30 planted errors and 30 correct lines, and a data trap and an agent exercise for each role.
Try it
Choose a role to watch a recorded practice session. Each step shows the learner’s answer and feedback, using saved coaching rather than live AI.
Choose a role and try the exercises yourself. Claude writes the coaching and hints; fixed answer keys score your work.
Under the hood
What the AI writes and the software checks
Written by Claude
- The coaching on your prompts
- The wording of each hint
- Your summary, the note for your manager and the practice plan
Checked by the software
- The answer keys never reach your browser before you submit
- Your score comes from the answer keys, not from Claude
- No hint before you try, and no hint names the line
- Feedback focuses on the work rather than praise of the learner
- The coach can offer hints but can’t write your prompt for you
These checks are tested automatically whenever the Lab is updated.
A demonstration with fictional people and cases, built from public research. Not any company’s training.