OneGov AI Deal Expiry Tests Federal Workers' Tool Habits

What happened

OneGov, the US federal government's bulk-licensing arrangement covering ChatGPT (OpenAI), Gemini (Google) and Claude (Anthropic), is set to expire next month, according to FedScoop. The deals gave millions of federal employees access to general-purpose AI assistants for day-to-day work, and their looming end has raised questions about what happens to usage patterns, workflows and productivity habits built up during the agreement.

FedScoop reports that the central issue is not simply procurement continuity but behaviour: staff across agencies have integrated these tools into routine tasks, and the expiry forces a decision point — renew, replace, or revert to prior ways of working. The story frames this as a test of how entrenched AI-assisted habits have become inside government, independent of which vendor or contract structure survives.

Why it matters

This is a live case study in habit formation and switching costs — core behavioural-economics territory. Once a tool becomes embedded in someone's daily workflow, the psychological and practical cost of losing it (or being forced to relearn a substitute) often exceeds the nominal value of the license itself. For service and experience leaders, the OneGov expiry is a reminder that adoption metrics captured during a "free" or subsidised pilot period rarely reflect what happens when the subsidy disappears.

It also has direct implications for employee experience design. Organisations that roll out AI tools broadly and cheaply, without deliberately shaping how habits form, risk either a disruptive drop-off in productivity when access changes, or quiet lock-in that limits future vendor flexibility — both outcomes with real cost and morale consequences.

The Renascence take

Most coverage of this story will focus on procurement and contract mechanics. The more useful lens is behavioural: what OneGov actually tested was whether subsidised access converts into durable habit, and whether that habit transfers across tools or fixes on one.

The real story here isn't which AI vendor wins the next federal contract — it's that a large, low-friction pilot has quietly become a behavioural experiment in habit lock-in. When access is free and ubiquitous, people don't evaluate tools on merit; they build routines around whatever's in front of them, and routines are far stickier than preference. Any organisation running a similar large-scale AI rollout should be tracking dependency and switching cost from day one, not discovering it only when the contract clock runs out. The lesson for service design is blunt: decide deliberately how much habit you want to create before you hand out the tool, because undoing it later is a change-management problem, not a procurement one.