Local AI Agent Readiness Planner

Prepare one bounded Ollama agent workflow before expanding access.

1. Define the job

The agent should help us:

The finished result should be:

2. Choose the route

☐ Local model

☐ Cloud model

☐ Hybrid with web search

Reason:

3. Choose the model

Model:

☐ Tool support verified

☐ Fits hardware

☐ Context is sufficient

☐ License reviewed

4. Working directory

Approved directory:

☐ Test copy

☐ No unnecessary secrets

☐ Backup or version control

☐ Easy to restore

5. Connected services

Web search, APIs, cloud models, files, email, calendar, or other tools:

Permission and Approval Gate

6. Permission level

☐ Read

☐ Draft

☐ Execute in sandbox

☐ Publish with approval

7. Credentials

☐ Read-only where possible

☐ Separate test account

☐ Scoped permissions

☐ Expiration / rotation

8. Evidence required

☐ Files changed

☐ Commands run

☐ Tests completed

☐ Sources used

☐ Remaining uncertainty

9. Recovery

☐ Backup

☐ Rollback command

☐ Audit log

☐ Kill switch

Named recovery owner:

10. Pilot result

Time saved:

Errors and review burden:

Quality and business result:

Decision:

☐ Stop

☐ Improve and retest

☐ Approve limited rollout