Insights from NASTD Spring Conference 2026: The Unfunded AI Problem in State IT 

In many states today, central IT is covering the cost of AI tokens. This approach makes sense in the early stages. It removes friction and allows agencies to experiment without immediate budget constraints.

But this model is temporary.

As usage grows, those costs will need to be allocated back to agencies. And when that shift happens, the lack of visibility into usage becomes a serious problem.

Agencies will be asked to pay for something they haven’t been tracking. Finance teams will be expected to bill for consumption they can’t clearly explain. And central IT will be caught in the middle, trying to justify numbers that lack transparency.

This is where the risk begins.

Without a clear understanding of usage, billing becomes reactive. Questions arise. Disputes increase. And in some cases, agencies may look for ways around the system entirely, adopting unsanctioned tools to avoid unclear or unpredictable costs.

The states getting ahead of this challenge are not waiting for chargebacks to begin. They are focusing on showback first.

Showback creates visibility without immediate financial impact. It allows agencies to see what they are using, how costs are calculated, and where consumption is increasing. It builds familiarity before accountability is enforced.

More importantly, it sets expectations.

By the time costs are formally allocated, agencies already understand the model. There are fewer surprises, fewer disputes, and a stronger foundation for financial governance.

AI adoption is not slowing down. But without a funding strategy that includes transparency, it introduces new financial and operational risks.

The unfunded AI problem isn’t a future concern. It’s already here. And the states that address it early will be in a much stronger position to scale AI responsibly.

Want to read more about the real conversations happening in state IT right now?

Read this exclusive blog post from our Director of Public Sector here.