The Setup That Creates the Problem
The most common approach to enterprise AI rollout in state government right now is straightforward. Central IT absorbs the cost of AI tokens during the adoption phase. Agencies get access to the tools at no direct charge. And the funding conversation gets deferred to a future fiscal year when adoption is more mature.
This is not an unreasonable strategy. Getting agencies comfortable with AI tools before introducing cost complexity is sensible change management. The problem is that deferred does not mean solved.
States like Massachusetts have been explicit about their timeline. Central IT is covering AI token costs through FY27, with agency chargebacks beginning in FY28. That gives the state roughly one fiscal year to build a tracking and cost recovery model that agencies will understand and accept.
Other states are less far along. Several state IT leaders have shared in recent conversations that they have not yet determined how agencies will pay for token usage at all. The tools are live. The costs are running. The billing model does not exist yet.
Two Risks That Grow Every Month You Wait
Delaying the financial infrastructure for AI cost recovery creates two distinct risks, and both compound over time.
The first is budget drain. AI token costs scale quickly with adoption. If central IT is absorbing those costs without a recovery mechanism, the IT fund is being depleted by a service that was never appropriated for that purpose. By the time leadership decides to act on cost recovery, the backlog may already be significant.
The second is Shadow AI. When agencies find that the enterprise AI path is slow or administratively burdensome, they look for alternatives. Consumer AI tools are accessible, affordable, and require no procurement process. Agencies that go this route are not trying to create problems. They are trying to get work done.
But Shadow AI means usage that central IT cannot see, costs that cannot be tracked, and compliance risks that cannot be managed. The governance model that justifies the entire CCoE investment breaks down when agencies are buying around it.
Why Showback Comes Before Chargeback
The solution is not to flip the chargeback switch immediately. In most state environments, moving too fast on cost recovery before agencies have visibility into their own usage creates exactly the political resistance that kills adoption.
The right approach is Consumption Showback: giving agencies a transparent, real-time view of their AI token usage before any billing begins. Let them see what they are consuming, what it costs, and what the trajectory looks like. Let their finance teams start building it into future appropriations requests.
When the chargeback eventually activates, it lands on a number that agencies already knew was coming. The conversation shifts from a surprise bill to a planned transition.
Showback also creates a data record that states will need later. Historical usage data supports rate setting, budget modeling, and agency-level forecasting in ways that retroactive estimates never can.
What SOFTRAX + BluLogix Built for This Problem
Consumption Showback at enterprise scale, across dozens of agencies and multiple AI platforms, requires automated financial infrastructure. It cannot be done with spreadsheets.
Our platform ingests AI token usage data, maps it to agency cost centers and organizational hierarchies, applies rate logic, and delivers real-time showback dashboards that agency finance teams can access directly. Central IT gets a complete view of enterprise AI consumption. Agencies get the transparency they need to plan and budget responsibly.
When the state is ready to move from showback to chargeback, that transition requires a configuration change, not a system rebuild. The data, the hierarchy, and the rate logic are already in place.
The same infrastructure handles cloud chargeback alongside AI token recovery, giving states a unified financial sub-ledger for all variable IT costs rather than separate manual processes for each service type.
The Window Is Closing
The AI mandate is not going away. The token costs are already running. Every month without a tracking infrastructure is another month of usage data that will never be recoverable.
States that act now will have the data and the infrastructure they need when the chargeback conversation arrives. States that wait will be building a billing model retroactively, without historical data, on a timeline driven by budget pressure rather than good planning.
Consumption Showback is how states close the AI funding gap on their own terms, before it becomes a crisis.
David Mink will be at the NASCIO Midyear Conference in Philadelphia, April 26-29. If your state is navigating the AI funding question, connect with him there or reach out to the BluLogix team directly.
