Just weeks after the Department of Defense (DOD) boasted that nearly half of its 3.5 million employees were using AI at work, the Army's Combat Capabilities Development Command (DEVCOM) was forced to rein in usage. An email informed staff that the Army CIO pool of AI tokens was exhausted and limits were reimposed, according to documents obtained by WIRED.
Token Overconsumption at the Pentagon
The email, reviewed by WIRED, stated: "Although the Army CIO announced in May 2026 that they were offering unlimited tokens, by mid-June the Army CIO pool was exhausted of tokens and had to re-establish limits." It added that while the Army renewed token usage at "its current levels," it remained unclear "if the Army CIO pool will be renewed after 1 Oct." An Army employee speaking anonymously told WIRED: "Apparently the whole Army burned through the whole year of tokens for just one service."
The Army uses Ask Sage, a multimodal generative AI platform that supports multiple large language models (LLMs), including Alphabet's Gemini, Meta's Llama, and OpenAI's ChatGPT. The platform is "accredited for Controlled Unclassified Information" and also used by the DOD's Chief Digital and AI Office (CDAO) for acquisitions.
Key data points from the source include:
| Metric | Value |
|---|---|
| Annual enterprise token subscription | 100,000,000 tokens |
| Token definition (Ask Sage) | 1 token ≈ 3.7 characters |
| Daily token burn during Operation Epic Fury (38 days) | 20 billion tokens per day |
| Monthly employee allotment | At least 200,000 tokens per month |
Ask Sage Platform and Usage
Ask Sage is used to "power its enterprise LLM workspace" according to the Army's website. Tasks included "reclassifying personnel descriptions, which involves defining and aligning job duties, experience and backgrounds." Employees who signed up but used few tokens received emails encouraging more usage. However, the rapid consumption forced the Army to dial back its ambition.
Pushback and Reliability Concerns
The anonymous Army employee reported that the tools were not particularly useful for their work and were often unreliable. One model even asserted it had completed a task it had not, they said. The employee noted they had "not found the generative AI tools to be particularly useful." This mirrors trends in the private sector: Meta reportedly removed its internal token leaderboard after encouraging "tokenmaxx," and Instagram head Adam Mosseri floated capping token use per engineer. Uber also saw engineers burn through a year's worth of tokens in four months, according to Fortune.
Implications for Enterprise AI Procurement
The Army's experience serves as a case study for enterprise technology leaders. Token-based pricing for AI models can lead to unexpected cost overruns if usage is not governed. The Defense Department's emphasis on AI continues despite the token crunch. On Monday, The Intercept reported that the Pentagon cut staff at the Civilian Protection Center of Excellence while developing an AI tool to speed up civilian casualty assessments. Organizations deploying generative AI at scale must implement usage monitoring, set hard caps, and prepare for budget reconciliation. The Army's scramble to re-establish limits after an "unlimited" promise highlights the gap between vendor marketing and operational reality.
Sources:
