City leaders who assume AI policy requires a multi-year planning cycle often delay until the risk is no longer manageable.

From the wild west to council-ready in one quarter.

Oakridge has 160,000 residents and grows about 2% every year. Like most cities its size, it had no AI policy. What it had instead was a growing problem: employees across every department using AI tools on their own, drafting documents, answering resident questions, and summarizing records, with no training, no oversight, and no one accountable when something went wrong. The CAIO called it the wild west. Something had to change.

The problem wasn't that employees were using AI. It was that no one knew.

By the time Oakridge's CAIO sat down to assess the situation, AI tools had already spread through finance, communications, HR, and public works. Staff were summarizing resident feedback with ChatGPT. A department head had used an AI tool to draft a procurement memo. Someone in permitting was experimenting with an AI chatbot for public-facing questions, without telling anyone. None of it was documented. None of it had been reviewed. And none of it was going to stop on its own.

The risk wasn't just legal exposure, though that was real. It was that the city had no way to say yes or no to anything. No framework. No review path. No one with clear authority to decide what was acceptable and what wasn't. Every week the city waited, the problem compounded.

The CAIO made the call: AI policy had to become a priority, and it had to move faster than a traditional planning cycle. That's when the team found the repo.

A different kind of starting point

The team had looked at policy templates from other cities. They were either too simple to be useful or too specific to adapt. What the toolkit offered was different: a structured GitHub repo, built around five module families, with templates, prompts, and evaluation rubrics already in place. More importantly, it was built to be used with Claude Code. That meant the CAIO's team didn't have to figure out what to ask or where to start. They opened the repo, ran the first prompt, and the agent understood the structure immediately.

What the first quarter actually looked like

Week by week. Module by module. Tied to real files in the repo.

Oakridge moved quickly. Many cities will stretch this work across a longer cycle. The value here is the shape of the work and the order it moved in.

Week 1

The CAIO and two team members forked the repo and opened it in Claude Code.

They started with README.md and the docs folder. The agent read the structure and oriented them to the five module families: strategic leadership, governance and policy, workforce and learning, operations and service delivery, and community trust. In the first session, they ran the strategic leadership intake prompt and got back a scoped summary of Oakridge's current state: what was happening, where the gaps were, and which module to tackle first.

Weeks 2–3

A governance and policy draft was on paper. Legal had seen a first version.

Using the governance prompt pack, the team produced a first draft of Oakridge's core AI policy with risk classification tiers, review paths, prohibited uses, and documentation requirements. Claude Code pulled from the governance template and flagged every section that required local decisions, including legal obligations, union agreements, and department scope. The CAIO reviewed, made calls on the flagged items, and sent a first version to the city attorney by the end of week three.

Weeks 4–5

Workforce and operations drafts completed. The team had all five layers outlined.

With governance in review, the team moved to workforce and operations in parallel. The workforce prompt pack produced a staff enablement plan: what training was required, what managers needed to know, and what employees would be told. The operations template gave them a service delivery playbook for any AI pilot that wanted to launch. By the end of week five, all five module families had at least an outline. The shape of the full policy package was visible.

Weeks 6–7

Multi-stakeholder review. Legal, DEI, IT, HR, and communications all in.

The CAIO used the review sprint kit to route the draft to each stakeholder lane. Legal flagged two sections: public records obligations and vendor contract language. DEI raised a question about the benefits screening classification. IT requested clarification on the data handling requirements for external vendors. All of it was documented using the review sprint routing register. Revisions were made. A second draft went back out by the end of week seven.

Weeks 8–9

Community trust layer drafted. Public feedback plan built. Residents looped in.

The community trust module was the one the team had been nervous about. Using the community trust template and prompt pack, they produced a public engagement framework: how residents would be notified about AI use, what feedback channels would be open, and how the city would respond to public concerns. A 30-day public comment window was opened. This was the right bottleneck. Public trust takes time, and the team had built that time into the plan from the start.

Weeks 10–12

Council packet assembled. Leadership briefed. 90% draft declared ready.

The board review packet pulled together everything: the policy document, the risk tier framework, the workforce plan, the public comment summary, and the implementation timeline. The strategic leadership scorecard was used to confirm that all five layers met the release-gate standard. The CAIO presented to leadership at week ten. Council received the packet at week eleven. By week twelve, Oakridge had a 90% draft policy package, fully reviewed, publicly commented, and ready for formal adoption. The remaining 10% was the build and training that would follow adoption.

How the team used Claude Code through the process

It wasn't magic. It was a clear workflow. Here's what it actually looked like session to session.

Session 1: Orient and scope

The team opened the repo in Claude Code and asked the agent to read the structure and explain what to run first. The agent mapped the five module families, identified which prompt pack matched Oakridge's starting point, and produced a one-page scoping summary. The team reviewed it, made two corrections, and approved the sequence.

Each session: Draft and flag

Every working session followed the same pattern: run a prompt from the relevant module family, review the output, resolve the flagged local decisions, and move the draft forward. Claude Code never made policy decisions. It generated draft language, identified where local input was required, and organized the output so a human reviewer could act on it quickly.

Review sessions: Revise and route

After each stakeholder review, the team brought comments back into Claude Code and ran the revision prompts. The agent incorporated the changes, re-flagged anything that created new conflicts, and updated the review-readiness summary. The CAIO had a current status view at all times.

The draft was 90%. That's the point.

Oakridge left week twelve with a policy package ready for formal adoption, not a finished implementation. The remaining work is the build: standing up the training program, implementing the review routing system, monitoring the first pilots, and hardening the community trust layer as real feedback comes in. That work takes months. But it starts from a council-approved foundation instead of from nothing.

The thing that surprised the team most wasn't the speed. It was the quality. Using a repo built from best-practice city models, reviewed by Claude Code prompt by prompt, and run through a real multi-stakeholder review process, the output was something leadership could defend. That's what the toolkit is for.

Explore each module in depth

The Oakridge story covers all five module families at a sprint pace. These pages go deeper on the parts your city is focused on.

Start From Zero

If Oakridge's week-one orientation is where your city needs to begin, this page walks through the full starting sequence — what to open first and how to run the first session.

Start from zero →

Program Includes

The six-part program structure Oakridge built — governance, workforce, operations, and community trust — is described in full here, with links to each module.

See the full program →

Workforce Training

The staff enablement plan and training path Oakridge completed in weeks four and five is documented in the workforce guidance pages, including manager guidance and training paths.

Open workforce guidance →

Rules and Review

The risk classification tiers and reviewer lanes Oakridge's legal team worked through in weeks two and three are covered here — approved uses, prohibited uses, and escalation paths.

Open rules and review →

Your city can run this same path.

Start with the repo

Fork it, open Claude Code, and follow the five-module sequence. The first session gives you a scoped plan. The first two weeks give you a draft.

Open the repo

Want to run this with expert support?

If your city wants a guided first session, someone who has run this path before and can help your team move faster and avoid the avoidable mistakes, that's the guided option.

Get guided help