Most government leaders picture “buying AI” as a big, deliberate decision — a project, a committee, a line item. In reality, it almost never arrives that way. It shows up quietly, folded into tools you already use. The scheduling software adds a recommendation engine. The case-management system starts “prioritizing” files. The hiring platform begins ranking applicants. Nobody signed off on “adopting AI.” It simply came bundled in.

That matters because governing something starts with knowing you have it. Right now, in most agencies, there is no single person who can say where AI touches a decision about the public — and that gap is exactly what an auditor, a reporter, or a resident will find first.

The reassuring news is that you don’t need a technology overhaul to fix this. You already own the strongest control there is: the contract. Every purchase your office makes comes with the power to ask questions, set expectations, and require proof before you sign. Governing AI is mostly a matter of turning that power toward a few new questions — what the tool decides, what it learned from, how you can check its work, and who is responsible when it’s wrong.

The agencies getting this right aren’t the ones buying the least AI, and they’re certainly not the ones who understand the math behind it. They’re the ones who decided, early, that any tool making decisions about people has to be able to explain itself — and who wrote that expectation into the paperwork before the tool ever went live.

You can start this week without a task force. Make a simple list of the systems your office relies on, and ask each vendor one plain question: “Where, if anywhere, does your product use AI to make or shape a decision?” The answers will tell you where to look first. Everything else — the standards, the clauses, the reviews — builds out from that list.

Schedule a 45-minute consultation →  (or ask about the AI Procurement Readiness Audit)