Getting started 3 min read
Why would a local government use AI?
Every department has a job of its own that AI can take the first pass at. Here’s what local government staff actually ask for.
Part of the path I don’t know where to start with AI
Because the work keeps growing and the team doesn’t. Every department has reading and writing it can’t keep up with: reports, notices, contracts, replies to residents. AI takes the first pass at that work, so staff spend their time on the parts that need judgment: the decision, the phone call, the council question.
Why local government feels this more than most
A company with a backlog can hire, drop a product, or turn a customer away. Local government mostly can’t do any of those. The permit still has to be reviewed. The records request still has a legal deadline. The council packet goes out on the same night whether the clerk’s office is fully staffed or not.
The teams doing this work are often small, and one person may cover three jobs. When that person is out, the work waits. That’s the gap AI fills first: the reading and writing that sits between a request coming in and a person making a decision about it.
Every team starts with the same four asks
Across 41 teams using GovAI, from finance to fleet, the same four requests sit in the top eight for 39 to 41 of them: rewriting for tone, restructuring a document, proofreading, and drafting emails.
It means every department has a use for AI on day one, without a project plan. It also means those four asks are where most people stop, and most of the value sits elsewhere.
Then each team finds a job that’s its own
Compared with the average team, some departments ask for one kind of help far more often:
| Team | The job they ask for | How often, compared with the average team |
|---|---|---|
| Cybersecurity | Security reviews | About 20 times |
| Clerks | Drafting resolutions and bylaws | About 12 times |
| Legal | Drafting contracts | About 11 times |
| Procurement | Drafting solicitations | About 10 times |
| Procurement | Translation | About 8 times |
These are the jobs with a known format, a deadline, and someone who has to sign off. That’s what makes them good work for AI: there’s a clear shape for the first draft, and a person already checks every one.
Source: GovAI requests from US and Canadian local governments, all time to August 2026, including free-tier use. Shares describe what staff using GovAI ask for, not local government as a whole.
The work doesn’t stop at five
About 9% of requests in GovAI start before 7am, after 6pm, or on a weekend. In fleet, it’s about 19%. Requests arrive in every hour of the day.
That’s a picture of small teams covering work that doesn’t fit inside office hours: the mechanic writing up a repair note at the end of a shift, the supervisor drafting tomorrow’s schedule at home. When the help is there at 9pm, the work doesn’t wait for morning.
Local time is estimated from each organization’s location.
A few real kinds of requests
- “Rewrite this bylaw amendment in plain language, with a short ‘What’s changing’ section for residents.”
- “From these 75 resident emails, group the main themes and suggest answers for our project FAQ.”
- “Turn this council report into a one-page notice for the website.”
- “From these procurement specs, draft a scope of work and a list of evaluation criteria.”
Why some AI efforts return nothing
MIT’s Project NANDA studied enterprise AI pilots in its State of AI in Business 2025 report and found that 95% returned nothing measurable. A pilot with no named job has nothing to measure. “Try AI and see” rarely survives the next budget conversation.
The local governments that get further start with the jobs above: work they can name, that someone owns, with a result they can point to. At a typical GovAI customer, nearly 20 departments are using it. It spreads as each department finds a job of its own.
Try this with your team
At your next team meeting, ask each person one question: which document do you write every week, from the same template, that someone else has to check?
Write the answers down. A council report, a permit summary, a reply to the same three resident questions, a monthly variance note. Those are your team’s first real candidates for AI, and most teams find more than they expect. The next step is picking one: How do we choose the first job for AI?
What stays with your staff
Approvals, records, and the final word. AI gives you a strong first draft. Your team decides what goes out, and your team is still the one residents call.
Before anyone starts, it helps to agree on one thing: What should staff never put into a public AI tool?