A city AI CoE wants department product owners engaged with Foundry continuous monitoring. Which operational practice fits?

Holding a recurring continuous-monitoring review with department product owners is correct because continuous monitoring only improves civic GenAI when humans regularly interpret signals and own actions. A city AI center of excellence that wants department product owners engaged should institutionalize a cadence—weekly or biweekly—where Foundry continuous-monitoring dashboards, quality regressions, safety alerts, and cost trends are reviewed with the owning departments. That practice turns telemetry into governance and shared accountability across municipal programs.

Holding a recurring model-retraining review with department product owners focuses on retraining decisions rather than continuous monitoring; it misses the ongoing review of quality regressions, safety alerts, and cost trends that keeps deployed civic GenAI accountable. Holding a recurring dataset-labeling review with department product owners centers on labeling work rather than continuous monitoring; it does not give product owners a cadence to interpret live quality, safety, and cost signals and own resulting actions. Holding a recurring deployment-approval review with department product owners gates releases rather than sustaining oversight; it leaves already-deployed applications without regular review for regressions that emerge after approval.

Exam caveat: operationalize continuous monitoring with a recurring multi-owner review, not a model-retraining review, dataset-labeling review, or deployment-approval review.

Operational check: schedule a standing CoE–department monitoring review, share Foundry dashboards with read access for product owners, assign action items for any breached quality/safety/cost signals, track closure, and escalate persistent regressions to the municipal AI review board.