Artificial Intelligence budget management and spending reality in public institutions
McKinsey experts state that despite falling token costs, public institutions are facing a new budget reality due to increased usage and agentic processes.
According to a recent research published by McKinsey, artificial intelligence unit costs are dropping rapidly while public institutions' AI expenditures are expected to increase. Experts point out that this situation requires effective cost management and new financial operations.
The Contradiction in Artificial Intelligence Costs
While artificial intelligence prices are falling rapidly, the total bills of public institutions are expected to increase. Unit token costs have decreased tenfold in the past year and about a thousandfold in three years.
Despite falling costs, increasing usage rates, agentic processes, and the transition to consumption-based contracts are driving up spending amounts.
Similarities with the Cloud Computing Process
Public sector leaders need to see how artificial intelligence usage translates into value and mission impact. Experts compare this process to the early-stage cloud services migration.
Both processes involve variable consumption, rapid experimentation, and complex pricing layers. The lessons learned by institutions from cloud expenditures also apply to artificial intelligence.
From Pilot Projects to Workflow Design
It is of great importance not to get stuck in hundreds of disconnected pilot projects and to avoid unscalable efforts.
When usage ceases to be a budgetary margin of error and becomes a major item, end-to-end workflows must be redesigned.
FinOps and Accountability
A clear owner must be assigned for artificial intelligence costs who can track token consumption and automate FinOps practices.
Accountability should not rest solely on a central artificial intelligence manager, but rather on service leaders supported by shared safeguards.