Univé Hit 97% Activation on ChatGPT Enterprise Licenses. Insurance Companies Aren't Supposed to Move That Fast.

Univé, one of the Netherlands’ largest cooperative insurers, achieved 97% activation of its issued ChatGPT Enterprise licenses, with 85% of licensed users active weekly — the kind of adoption curve regulated, risk-averse industries like insurance rarely produce on a first company-wide rollout. OpenAI’s case study on the deployment credits three ingredients: leadership commitment, responsible governance built in from the start, and employee-led innovation, with the explicit goal of making every employee able to use AI safely and effectively rather than restricting access to a pilot group. Univé’s next phase is Workspace Agents designed to proactively prepare recurring work — gathering information and surfacing context across approved enterprise systems before an employee’s day starts.

The number that matters here isn’t the license count, it’s the weekly-active rate. Plenty of enterprise rollouts hit high initial activation and then quietly stall into occasional use; 85% weekly engagement six months into a company-wide deployment is evidence of behavior change, not a stalled pilot with a good press release.

For a sense of where sustained enterprise trust in AI can eventually go, Rakuten’s deployment of Claude is a useful contrast point. Rakuten’s General Manager of AI for Business, Yusuke Kaji, reports agents now closing issues roughly ten times faster across every domain the company tracks, driven by what he calls “taste alignment” — the model re-checking its own assumptions and returning to first principles without being prompted. His quote captures the shift in kind, not just degree: “We tested Fable, and we love its capability for self-reflection and self-verification. Compared with previous models, it understands its mistake before I point it out at 2 a.m. — so that I can sleep.” Univé’s 97%/85% is the adoption-and-governance stage; Rakuten’s ten-times-faster is what that adoption compounds into once an organization starts delegating whole tasks instead of narrow ones.

Read together, the two cases sketch a rough maturity curve for enterprise AI: get activation and governance right first — the unglamorous part — before the compounding gains from delegating bigger units of work become available at all.