Talk to procurement leaders instead of vendor marketing, and public sector AI adoption looks a lot more cautious — and a lot more reasonable — than the pitch decks suggest.
Every procurement technology conference this year has had some version of the same keynote: AI is transforming government buying, adoption is accelerating, the agencies that don’t move now will be left behind. Spend any real time with procurement leaders instead of vendor booths, and a different picture shows up.
Where Agencies Actually Are, Not Where the Pitch Decks Say
Public sector AI adoption right now is uneven and, for the most part, narrow by design. A handful of agencies are running AI on document review and bid comparison, where the tool’s job is legible and the output is easy to check. Very few are letting AI touch anything that looks like a decision — an award, an exception, a payment release. That’s not agencies falling behind. It’s agencies being appropriately skeptical of tools that can’t yet explain themselves well enough to survive an audit.
The Three Hesitations We Actually Hear
Procurement leaders aren’t resistant to AI in the abstract. The hesitation clusters around a few specific, reasonable concerns:
- “We can’t explain it later.” If a decision gets questioned in a protest or an audit two years from now, “the model recommended it” isn’t an answer anyone wants to give.
- “It wasn’t trained on how we actually buy.” Cooperative vehicles, statutory thresholds, and diversity requirements are specific to public procurement — and generic AI tools often weren’t built with any of that in mind.
- “Who’s accountable if it’s wrong?” A vendor’s model making a bad call doesn’t remove the agency’s accountability for that call. Most leaders aren’t willing to take on that risk for a workflow the vendor can’t fully explain.
None of these are technophobia. They’re the same due diligence a procurement leader would apply to any new system — applied, correctly, to AI.
What Responsible Adoption Actually Looks Like
The agencies making real progress aren’t the ones deploying AI everywhere at once. They’re starting with the highest-volume, lowest-ambiguity tasks — document extraction, bid comparison, anomaly flagging — where a person can verify the output quickly and the AI’s role stays confined to analysis. They’re asking vendors to show the audit trail before the first pilot, not after the first incident. And they’re treating human sign-off on every decision as a permanent design principle, not a training-wheels phase they’ll eventually remove.
The agencies getting AI adoption right aren’t the fastest movers. They’re the ones who decided in advance exactly where the AI’s job ends and a person’s begins — and haven’t moved that line since.
What the 2027 Budget Cycle Probably Looks Like
Expect more agencies to formalize AI governance policies before they expand AI usage, not after — procurement offices are increasingly writing the “what AI can and can’t decide” rule into policy before the RFP goes out, rather than discovering the gap during implementation. Expect vendor evaluations to ask harder, more specific questions about explainability, and expect AI tool budget requests to get more scrutiny than they did in 2026, not less, as more agencies compare notes on what actually worked.
Field Notes, Not Sales Pitches
This is where we’ll keep sharing what we’re actually seeing across public sector procurement — the adoption patterns, the hesitations, the things that work and the things that don’t. If that’s useful, subscribe for more of it.
Good breakdown of the shift from search to execution — that is where this actually starts changing day-to-day work.