OneGov AI Deals Expire: Federal Workers Face AI Dependency Test

What happened

The US federal government's OneGov programme, which secured bulk licensing for generative AI tools including ChatGPT, Gemini and Claude across federal agencies, is approaching expiration. According to FedScoop, federal employees who have built these tools into their daily workflows now face a decision point: renew access through new arrangements, switch to alternative platforms, or lose the tools altogether.

The looming deadline turns what began as a procurement and licensing question into an operational one, as agencies and individual staff assess how dependent their work has become on AI assistants that were, until now, provided centrally and at no direct cost to end users.

Why it matters

Bulk-licensing programmes like OneGov were designed to accelerate AI adoption across government by removing procurement friction. Their expiry surfaces a less obvious effect: once employees restructure how they draft, research, summarise or code around an AI tool, that tool stops being optional software and becomes part of the workflow's architecture. Removing it isn't a simple downgrade — it can mean relearning processes, losing accumulated context, or replicating workarounds in a different interface.

For digital transformation leaders, this is a reminder that adoption metrics (how many users logged in, how many queries were run) tell only part of the story. The more consequential measure is behavioral entrenchment — how much a tool has reshaped the way people actually get work done. That has direct implications for how governments and enterprises structure future AI contracts, budget for continuity, and plan migrations when a vendor relationship changes.

The Renascence take

What's unfolding inside federal agencies is a textbook case of default-driven habit formation — and it's rarely discussed in AI procurement conversations, which tend to focus on cost and capability rather than switching cost.

Once a tool is free, centrally provisioned and embedded in daily tasks, employees don't experience it as a discretionary choice — they experience its removal as a loss, even if a comparable alternative exists. This is the same behavioral mechanic behind subscription retention and loyalty-programme lock-in, just relocated into the workplace. Any organisation rolling out AI tools at scale should treat the exit plan as seriously as the launch plan: map which workflows have quietly become AI-dependent, document the reasoning and prompts behind them, and build in tool-agnostic habits before a licensing cliff forces a scramble. The real lesson from OneGov isn't about which chatbot wins — it's that convenience, once normalised, becomes infrastructure, and infrastructure doesn't get renegotiated gracefully at the last minute.