In May the Pentagon stated that half of its cited 3.5 million employees were using AI tools. The US Army promoted AI use among its personnel via the Ask Sage platform, which runs models including Gemini, Llama, and ChatGPT. The Army had an annual allotment of 100 million tokens for Ask Sage; that allotment was depleted by mid‑June.
What happened in practice
The initiative was supported by internal nudges: reminder emails to inactive users, leaderboards for the most active staff, and automatic top‑ups intended to meet high demand. Despite these measures, there appears to have been no adequate modelling of the financial impact of a policy that essentially encouraged widespread AI use.
As a result of advertising "unlimited" usage and the rapid depletion, staff began receiving rationing messages around mid‑June asking them to curb consumption. What had been budgeted for a year was used in roughly thirty days.
Why this matters
The episode highlights a gap between efforts to drive technology adoption and the realities of budget forecasting. The Army presented AI uptake as evidence of institutional transformation, but the token exhaustion shows that cost planning did not keep pace with adoption and usage patterns.
Open questions
Officials have not confirmed whether funding for the token pool will be extended beyond October. That uncertainty could affect the pace of further AI deployments and how personnel integrate these tools into their workflows.
Conclusion
This is more than a logistical hiccup: it illustrates both the power of internal incentives and the consequences of insufficient budgetary planning. The Army's move from promising "unlimited" access to asking users to limit themselves is a cautionary example for other government and corporate rollout programs.
