Van Buren County, Michigan (pop. 75,000)
Discussions about AI in government often drift into the abstract, like a strategic plan that mentions “leveraging synergies” without ever explaining who’s doing the leveraging, or what exactly is being synergized. Phrases like “AI is transforming the public sector” and “Copilot is everywhere” float around conference rooms and PowerPoint decks with all the specificity of a weather forecast for “cloudy with occasional disruption.”
But what does a typical day of AI use inside county government actually look like?
In Van Buren County, Michigan, a community of about 75,000 people and a government that prefers practical results to theoretical revolutions, a recent analysis of firewall and web traffic data offers a refreshingly grounded answer.
The Numbers Behind the Buzz
On an average workday, the county network logs about 2,500 AI-related access events, touching roughly 80 active employees. That may not sound like much to a Silicon Valley product manager, but in the world of rural local government, where IT budgets are modest and attention spans are frequently occupied by zoning maps and pothole complaints, it’s a meaningful signal.
The usage pattern follows a familiar shape: a core of power users doing the bulk of the work, surrounded by a periphery of casual or curious dabblers. In other words, your classic technology adoption curve, just with more public records and fewer venture capitalists.
ChatGPT Reigns Supreme
Of all the AI tools being used, ChatGPT absolutely dominates, accounting for over 70% of all usage. And this isn’t just passive traffic or idle curiosity. Employees are regularly engaging with various ChatGPT endpoints (chatgpt.com, ws.chatgpt.com, realtime.chatgpt.com) throughout the workday, suggesting real task-oriented interaction: drafting emails, summarizing documents, brainstorming responses to angry constituents who write in ALL CAPS.
Other tools, Google Gemini, Claude, Scribe, and a few smaller platforms, round out the AI ecosystem. But the overwhelming majority of activity points in one direction, which has practical implications: if you’re looking to focus training, policy, or security reviews, start where the usage actually is, not where the marketing videos are slickest.
The Copilot Mirage
Microsoft Copilot also made a frequent appearance in the raw data. Or so it seemed. Once analysts peeled back the layers, they discovered that nearly all Copilot entries were just digital background noise, triggered automatically when someone opened Word or Excel, regardless of whether they interacted with AI at all.
The result? Actual human-initiated Copilot use was virtually nonexistent.
This is a useful reminder that not all AI traffic is meaningful engagement. Sometimes, it’s just your software muttering to itself in the background while you’re trying to find last year’s budget spreadsheet.
Why This Matters
So why bother measuring all this? Because clarity matters. Especially in government, where every new initiative is expected to be simultaneously innovative, cost-neutral, risk-free, and universally beloved.
- Strategic clarity: You can’t make smart decisions about training, policy, or budget without knowing what’s actually happening.
- Governance focus: If 70% of real use is ChatGPT, maybe we don’t need an AI oversight framework that treats every tool as equally probable and perilous.
- Scalable success: Your power users are the beachhead for wider adoption. Support them first. Let them test the boundaries so others don’t have to.
The Bottom Line
A typical day of AI use in Van Buren County government is practical, focused, and refreshingly free of hype. Real engagement is happening, mostly with ChatGPT, mostly from a concentrated group of users, and mostly for work that needs doing.
AI isn’t revolutionizing government overnight. It’s helping a small group of staff do their jobs a little faster, a little better, and with slightly less keyboard-induced sighing. And frankly, that’s a very good place to start.
