AI operations workshop

Scenario

Create the operating model for a private RAG service with an interactive objective, batch indexing, limited accelerator capacity, permission controls, and an approved maintenance window.

Tasks

1. Define successful and unacceptable outcomes.
2. Map request, dependency, policy, evaluation, change, incident, and evidence signals.
3. Create indicator definitions and service objectives with windows, exclusions, and owners.
4. Design privacy-preserving dashboards and alerts.
5. Run a tabletop exercise for capacity exhaustion followed by a grounding regression.
6. Define containment, safe degradation, rollback, recovery validation, and evidence.
7. Build a capacity envelope and unit-economics model with sensitivity ranges.
8. Specify release and reassessment triggers.

Deliverables

Submit the signal catalog, objective sheet, dashboards or wireframes, alert rules, incident timeline, recovery checklist, capacity model, cost model, and operational evidence report.

Passing work connects every metric to a decision and does not classify an unauthorized or unsupported response as success.