
A government AI strategist asked a simple question on LinkedIn last week. Had anyone moved from OpenAI's business plan to Enterprise? The license minimum had dropped, he noted, but Enterprise meant switching from per-seat pricing to consumption. And consumption, in his words, is one of the hardest things to calculate.
He's right. In fact, it's a complete mismatch for government budgeting.
Consumption pricing works fine for a startup with a credit card and a flexible burn rate. It does not work for a county that sets its budget a year in advance, in public, with a vote. When you can't estimate a cost, you can't budget it. When you can't budget it, you can't responsibly buy it.
Here's how most governments fund technology. Departments submit requests months ahead. Finance builds a budget. Council or the board adopts it in a public meeting. That number becomes a line item, and the line item is the ceiling.
There is no "we'll true it up next quarter." Going over means a budget amendment, a supplemental appropriation, or pulling money from somewhere else. Every one of those is a public conversation, often on an agenda, sometimes in the local paper.
Consumption pricing asks government to do something it is built not to do: commit to an unknown number. A vendor quote that says "estimated $X, actual cost may vary based on usage" is a quote a finance director can't sign with a straight face. Procurement wants a not-to-exceed amount. Auditors want to see that what was approved is what was spent. Elected officials want to know the cost before they vote, not after.
Government demand doesn't run on a smooth curve. It spikes, and it spikes for reasons nobody controls.
Under consumption pricing, every one of those moments costs more. The bill goes up exactly when the community is under the most stress and staff are stretched thinnest. A bad storm becomes a weather event and a budget event.
That's backwards. Surge moments are when AI earns its keep. They're when a front desk team of four can't answer 400 calls, and residents need accurate answers anyway. Pricing that penalizes those moments penalizes the reason you bought the tool.
The unpredictable bill is only the first problem. Consumption pricing creates quieter ones too.
Teams start rationing. When every query has a price tag, someone eventually asks staff to use the tool less. Departments get told to hold back. Residents get pushed back to the phone queue. The technology you bought to expand service ends up capped to protect the budget.
Success gets punished. Launch well, and residents use it more. Use it more, and costs climb. Now the department that did everything right is the one explaining an overage at the next budget hearing.
Finance inherits a forecasting job. Someone has to track usage monthly, model the rest of the year, and flag risk before the line item runs dry. That's staff time most finance teams don't have.
Procurement gets harder. Many solicitations and contracts expect a fixed price or a firm not-to-exceed. An open-ended usage model can slow approvals, trigger extra legal review, or stall a purchase entirely.
Government doesn't need cheaper meters. It needs no meter at all.
Pricing built for the public sector should do three things:
That's the model Polimorphic is built on. Agencies know what they'll pay before they sign. Then they're free to do the thing they bought AI for: answer every resident, every time, especially when it matters most. Our customers see 50–90% fewer calls reaching staff. That relief shouldn't come with a surprise bill attached.
If the answers start with "it depends," you're looking at a meter.
Your residents don't call less because the budget is tight. Your AI shouldn't cost more because they called.
Book a demo and see what predictable pricing looks like for your agency.





