Every request to GovAI is classified by department, kind of work and how complex it is.
Here’s what that shows about how work in local government is changing.
FPO image
41
teams across 9 departments
180+
local governments
23,000+
active users
2.8M+
real staff prompts
What we’ve learned
180+ local governments. 2.8M+ prompts. Six ways the work is changing.
What staff across the US and Canada ask GovAI for, and what it tells us about how work gets done in local government.
Every team starts with the same everyday asks.
Polishing their writing and drafting communications sit near the top of the list for nearly all 41 teams. Close behind: asking GovAI to walk them through how to do something.
Why it mattersThe first win is the same in every department, so start with everyone.
Every team has a signature ask.
Cybersecurity asks for security reviews about 20 times as often as the average team. Clerks draft council resolutions and bylaws 12 times as often. Legal drafts contracts 11 times as often. Procurement drafts RFP scopes 10 times as often.
Why it mattersAfter the basics, the gains come team by team. Pick one job per department, like the clerk’s resolutions or procurement’s RFPs, and build it into how the team works.
Half of all requests start from a document staff already have.
Staff paste in a draft, a policy or a resident’s email and ask GovAI to work on it. At the front counter, it’s close to nine in ten requests.
Why it mattersAI helps most with work already on someone’s desk. There’s no new process to set up first.
Sensitive information turns up in every department, not just HR.
Nearly half of all requests include sensitive information. Front counter requests include more than Finance’s. The most sensitive work sits in labor relations, EMS and social services.
Why it mattersSensitive information isn’t only in HR and Legal, so every staff member needs the same guardrails.
Some teams use AI where you’d least expect.
Procurement asks for translation about 8 times as often as the average team, and EMS more than 3 times as often. Real estate fixes Excel formulas. Emergency management makes disaster-scene images for training exercises.
Why it mattersYou can’t predict where AI will take off. Give everyone access, then build on what sticks.
Meet the unlikely twins: teams you’d never guess work alike.
Clerk and council works most like Licensing and bylaw: both draft bylaws and look up policy. Finance works like Internal audit. EMS, Public health and HR all build training material.
Why it mattersA prompt or template that works in one department can work in another, so share them across the org chart.
2.2×On weekends, staff are about twice as likely to ask GovAI for help writing a report. The long work waits for a quiet day. GovAI platform data, requests to August 2026
What the data says about how staff work.
When the work happens, who does it, and what they come back for. Use the arrows to move through them, and click a team to see its work.
Mornings go to the inbox. Afternoons go to reading.
42%of requests start between 7 and 11am, the busiest part of the day.
Before 7, it’s email drafts and tone rewrites. Mid-morning brings agendas and how-to questions. Around lunch, staff write public notices and solicitations. Late afternoon goes to summarizing reports and policies.
What it means
AI isn’t one job. It changes with the day, so the best first use case for a team is the work that fills its busiest hours.
Share of requests by local start time
12am6am12pm6pm11pm
Before 7 4%
Clearing the inboxEmail drafts · Tone rewrites · Proofreading
7–11am 42%
Getting the day set upAgendas · Database queries · Translation · How-to questions
11–2pm 29%
Writing for the public and for bidsPublic notices · Solicitations · Data breakdowns · Rollout plans
2–6pm 21%
Reading and summing upSummarizing reports · Summarizing policies · Pulling data into tables
Under each time: what’s more common then than across the week
After hours, staff bring the long documents.
2.5×as likely to ask for a contract summary outside office hours as during them.
About 9% of requests start before 7am, after 6pm or on weekends. Those lean toward slow, heavy work: contracts, reports, training material and presentations. On weekends, reports are about 2.2 times as likely. Quick emails and web copy drop off, and the requests carry more sensitive material.
What it means
The work that waits for a quiet hour is the long read and the first draft. Handing GovAI the first pass is how that work stops spilling into the evening.
Outside office hours, compared with during them
Summarize a contract2.5×
Write a report1.6×
Build training material1.4×
Make an image or graphic1.4×
Write speaking notes1.3×
Draft an email or letter0.85×
Write a formula or script0.77×
Write web or social copy0.69×
More likely after hoursLess likely
Fleet keeps the longest hours.
19%of Fleet requests start outside office hours, about twice the average team.
Shift teams lead: Fleet, EMS, water and utilities, and public works. Libraries are up there too.
What it means
Training, office hours and support built around a 9-to-5 day miss these teams. Plan the rollout around the hours shift teams work, or the people with the least desk time get the least help.
of requests start outside office hours
Managers bring AI in. Then it spreads to staff.
41%of a team’s first GovAI users are managers, directors or executives. Among everyone who joins after, it’s 27%.
Water and utilities shows it most: 61% of the first people to use GovAI are leaders, against 32% of those who follow. IT and HR start the same way. In Building and permits, frontline staff were there from the start.
What it means
Adoption mostly flows from the top down. Get the managers in each department using it on their own work first, and their teams tend to follow.
Share of each team’s GovAI users who are managers, directors or executives
Among the first usersAmong everyone who joined later
Repeat jobs become shared tools.
30%of Cybersecurity requests start from a shared template or an Agent the team set up.
The surprise is EMS, second only to Cybersecurity. Paramedics run 45% of their rewrites through an Agent their service set up, and about 24% of staff updates and schedule notices start from a template. Internal audit, IT and Accounting and payroll follow.
What it means
When a team asks for the same thing every week, a shared template or Agent means nobody starts from scratch, and every answer follows the same rules.
of requests start from a template or an Agent
Help one team, and the help ripples out.
Communications writes for Parks, Economic development and public safety. Legal reviews for the clerk’s office, bylaw and council. Procurement buys for Legal, Engineering and Finance. Much of what these teams ask GovAI for is another department’s work.
What it means
Start with the teams other teams wait on. Every hour they get back shows up in the departments they serve: the notice goes out sooner, the contract comes back sooner, the bid opens sooner.
TeamWorks on
The more local the rules, the more AI needs your documents.
1.6×the average team’s need for local know-how, in Legal.
Requests from these teams depend most on your own jurisdiction: your bylaws, your plans, your council’s decisions. A general answer won’t do.
What it means
For these teams, put your own bylaws, plans and policies in GovAI first, so answers come from your rules rather than someone else’s.
the average team’s need for local know-how
Economic development is the most polite.
1.7×as likely to say please to GovAI as expected.
Parks, administration and bylaw staff are polite too. IT, Data and GIS, and Cybersecurity say please about half as often as everyone else. The bigger pattern is length: the longer the request, the more likely it says please.
What it means
Mostly for fun. But staff who say please are writing to GovAI the way they’d brief a colleague, with context and detail, and that’s what gets a better answer. It’s a habit worth teaching.
Requests that say please, by length
Under 50 characters3.5%
400 characters or more24%
Say please more often
About half as often
1 / 8
Where the time goes
Some teams hand AI five-minute jobs. Others hand it half an hour.
Our task model estimates the minutes of by-hand work behind each request, based on how complex it is. Ranked by team, it shows where AI is already taking on the heaviest work. Pick a team to see what light and deep work look like there.