# AI agent scoping worksheet

One task, an expected outcome and explicit limits. Complete this worksheet before choosing a solution.

1. **Current situation**: who performs the task, with which tools and difficulties?
2. **Target task**: what starts it, what ends it and what outcome is observable?
3. **Examples**: collect ordinary, incomplete, conflicting and out-of-scope cases.
4. **Data**: useful sources, owners, access rights, updates and quality.
5. **Actions**: permitted reads, possible changes and required human approvals.
6. **Exceptions**: when should the system stop, ask a question or hand over?
7. **Evaluation**: success criteria, unacceptable errors and review process.
8. **Operation**: owner, monitoring, support and improvement.
9. **Initial scope**: included work, exclusions and assumptions to test.
10. **Next decision**: what should the pilot help you decide?

| Anonymised real request | Expected outcome | Required source | Approval | Success criterion |
| --- | --- | --- | --- | --- |
| | | | | |
| | | | | |
| | | | | |

Do not include secrets or unnecessary personal data in material sent to a provider. This worksheet helps scope a discussion; it does not authorise access to any tools.