Intake
Give teams a clear path for new tool ideas, vendor proposals, expanded uses, and service changes. If intake is informal, the city will lose track of what started where and why.
Open intake and working filesCities that adopt AI policy without building the operational layer — intake, review, escalation, documentation — see drift and unauthorized use within months of rollout.
Operations and service delivery are where city AI policy either holds or drifts.
This page is about day-to-day AI operations after launch: intake, human review, escalation, vendor standards, documentation, and monitoring. If the city cannot run those things clearly, the policy does not really exist in practice yet.
What operational alignment usually requires
Keep the path simple enough to use and clear enough to audit. This is where the city turns policy into repeatable operations instead of hoping teams will interpret the rules the same way on their own.
Human review and approval
Put human review in front of important operational uses so AI work does not move straight from idea to implementation without oversight, especially when residents, benefits, records, or frontline service are involved.
See the review pathEscalation
When a team hits a use case outside the approved path, the city needs a real escalation route instead of informal exceptions or hallway approvals.
Open rules and reviewDocumentation
The city should be able to show what was approved, under what conditions, and what guardrails were attached. That record is what makes later review and accountability possible.
See documentation materialsMonitoring after launch
Monitoring is how the city catches drift, unexpected expansion of use, vendor behavior changes, and service delivery problems before they become normalized.
See the public-facing side of operationsUse the repo when your city is ready to build the operational layer.
This page should explain the shape of the work. The repo should help your city produce the intake forms, escalation rules, documentation patterns, and monitoring materials that make operations hold up.
- Use the working files when you need operational templates instead of more explanation.
- Use the prompts to turn policy into daily practice.
- Use the review path before operational changes go live.
Before the city says operations are covered, make sure these are true
These checks help test whether the operational layer is actually ready to use.
- Intake gives teams a clear path for new AI ideas, vendor proposals, or expanded use.
- Human review happens before important work moves without oversight.
- Escalation gives teams a path when they hit a use case outside the approved rules.
- Documentation lets the city show what was approved and under what conditions.
- Monitoring catches drift or unexpected expansion of use after launch.
What to track once operations are running
The right indicators depend on your city's context and what best practice looks like at that time. Specific targets will shift as tools, risks, and guidance evolve. Start with questions the city should be able to answer rather than a fixed number to hit.
- Can the city show which AI uses are active and documented?
- Can the city show which uses have gone through the review path?
- Do staff know how to escalate an uncertain use, and does that path get used?
- Has the policy been reviewed since launch, and is a next review date named?
- Are there open complaints, incidents, or public questions the city still needs to address?
What counts as a satisfactory answer to these questions will shift as AI technology and city practice evolve. Build the tracking structure now so the city has a baseline to improve from.
If the city wants policy to hold in service delivery, it needs the operational layer too.
The repo should help your city build the operational materials that make AI use reviewable and consistent after launch. If your city still needs help standing up that process after using the repo, get help at the end.