The Funding Model Was Built for a Different World
Traditional state IT financial operations were built around predictability. Annual appropriations are fixed at the start of the fiscal year. Operating appropriations lapse at year-end, meaning unrecovered costs from prior periods cannot easily be billed retroactively. IT Internal Service Funds are structured to recover costs through interagency billing, but those mechanisms were designed for stable, foreseeable charges.
Cloud consumption does not behave that way. Usage fluctuates hourly based on citizen demand and events entirely outside IT’s control. Artificial intelligence intensifies this mismatch further. AI token costs do not follow predictable patterns. A single AI-assisted citizen service can consume dramatically different compute depending on query complexity, model version, and concurrent user volume. An autonomous agent processing a backlog of administrative requests may run for hours, generating costs that no line-item appropriation anticipated. There is no fixed unit cost to budget against. There is only consumption, measured after the fact.
The 2026 NASCIO State CIO Top 10 Priorities directly reflect this pressure. Artificial intelligence ranks first. Budget and cost control returned to the top three for the first time in years. The NASCIO-PTI State of the States Tech Forecast found that only 41 percent of states have updated procurement contract terms for AI, meaning financial accountability structures are absent in the majority of states, even as deployment accelerates. The technology investment is moving forward. The financial infrastructure underneath it largely is not.
The Difference Between Spend and Allocation
This distinction matters more than most billing conversations acknowledge: the cloud bill is not the problem. Allocation is.
Most state governments operate enterprise cloud agreements with hyperscalers that include committed spending thresholds or volume discounts. The vendor is paid regardless of fluctuations in internal consumption. What those agreements do not solve is the internal question of which agency drove which portion of that consumption.
When a state’s tax filing portal approaches its annual deadline, citizen traffic surges predictably in the final days. The cloud vendor bills for that consumption, and central IT pays. But the financial question that matters is whether the Department of Revenue, the Labor agency processing wage filings, and the Treasury office managing payment reconciliation each received accurate chargebacks for the consumption their systems generated, within their fiscal period, documented at a level of detail that would survive an interagency dispute. A single filing deadline can touch multiple agencies simultaneously. In most states operating manual reconciliation processes, the answer to at least one of those questions is no. That is where the IT fund absorbs costs it should have recovered.
The enterprise commitment makes accurate allocation more urgent, not less. When the spending is happening regardless, the only financial lever central IT has left is attribution.
Why Manual Reconciliation Breaks
State IT billing operations that rely on manual reconciliation face three compounding challenges that did not exist in the fixed-cost world the systems were designed for.
- Volume is only the first problem. A state operating 60 agencies across multiple cloud and AI platforms generates hundreds of billing line items per month under normal conditions. When a consumption spike occurs, driven by a policy announcement, a service launch, or a federal enrollment deadline, that volume can multiply several times within a single billing period. Human reconciliation capacity does not scale with consumption volume.
- Complexity is the harder problem. Each cloud or AI usage record requires a judgment call: which agency, which cost center, which interagency agreement. When consumption spikes across multiple agencies simultaneously, every record appears to be a potential anomaly. Error rates climb even when experience and effort remain constant.
- Timing makes errors permanent. Operating appropriations lapse at the fiscal year-end. A billing error not caught within the same fiscal period often cannot be recovered retroactively. What began as a reconciliation error becomes an unrecoverable absorbed cost, carried into the next year as an unexplained variance.
The Problem Does Not Appear on Any Dashboard
IT fund variances absorbed due to billing errors do not appear as line items in any budget document. They surface as unexplained variances at fund close. The Annual Comprehensive Financial Report captures the IT Internal Service Fund at the aggregate level. It does not decompose cloud versus non-cloud spend, nor does it surface misallocation between agencies.
The NASCIO-PTI Forecast identified “funding does not match expectations” as a top enduring challenge in state IT leadership. More than two dozen state CIO transitions have occurred since 2024 alone, and new CIOs consistently discover that cost allocation for cloud and AI services is, in their words, a black box. That pattern of discovery, not catastrophic failure but invisible inadequacy, is the shape revenue leakage takes in state government IT.
Because no single document or dashboard captures the gap between what was consumed and what was correctly recovered, there are no published benchmarks for state IT billing leakage rates. That absence is not evidence that the problem is small. It is evidence that the problem has not yet been measured.
The Question Worth Asking Before the Next Audit
The diagnostic question is straightforward: if your state experienced its largest cloud and AI usage spike of the year this month, could you produce an agency-level attribution report showing which agency consumed what, reconciled against their current appropriation, before your next budget meeting?
If the answer is no, or if it would take six weeks of manual reconciliation to produce, that is the gap this series is examining.
The next piece looks at the specific scenario where that gap is most exposed: a citizen-facing digital service hit by an uncontrolled demand spike. The cloud vendor gets paid. The question is what happens to the internal allocation, and who absorbs what cannot be attributed in time.
This is the first in a five-part series, “The Invisible Deficit,” examining revenue leakage in state government cloud and AI billing. The series draws on NASCIO research, state IT strategic plans, and conversations with state technology and finance leaders.
Dave Mink works at the intersection of public sector technology and financial operations · softrax.com
