A city CAIO is building the AI roadmap. When should product owners, model stewards, and ethics reviewers be defined?
Select an answer to reveal the explanation.
Short Explanation
Naming who owns the product, who stewards the model, and who does ethics review isn’t a post-launch scavenger hunt. Bake those seats into the roadmap before anything goes live.
Full Explanation
AI operating models in city government are part of strategy, not an afterthought bolted on after production. Defining product owners, model stewards, and ethics reviewers during roadmap planning—before go-live—establishes who owns outcomes, who monitors model behavior and change requests, and who escalates fairness, privacy, and public-trust concerns while the initiative can still be scoped, staffed, and funded. Embedding these roles in the CAIO’s roadmap also clarifies RACI across departments, vendors, and shared services so accountability is contractual and budgetary, not informal hallway consensus.
Defining roles only after the first production outage reveals ownership gaps fails because incidents are the most expensive way to discover missing accountability; outages expose gaps after residents and operators already absorbed harm. Leaving roles informal forever in municipal IT fails by concept: civic AI affects rights, budgets, and equity obligations that informal “we’ll figure it out” ownership cannot satisfy under audit or public records scrutiny. Defining roles only for vendor staff and never for city employees fails because the city retains residual risk and policy authority; vendor-only ownership leaves the municipality without named internal decision rights when contracts end, disputes arise, or policy exceptions are needed.
Exam caveat: naming roles early does not require fully staffing every seat on day one of strategy drafting, but the roadmap must identify the seats, decision rights, and interim coverage before production use. Operational check: before any municipal AI go-live, confirm named product owner, model steward, and ethics reviewer (or equivalent) appear in the roadmap/RACI, contact paths are current, and escalation for model drift or ethics concerns is documented and tested.