Univé’s 1,500 custom GPTs slash claims prep from hours to minutes

Most companies treat AI like an IT project. Univé, one of the Netherlands’ largest cooperative insurers, treated it like a wholesale rewiring of how 3,000 knowledge workers think about their jobs. The numbers tell a stark story: 97% of ChatGPT Enterprise licenses are activated, 85% of employees use it weekly, and staff have built roughly 1,500 custom GPTs without a centralized development team dictating what gets built.

This wasn’t a free-for-all. Univé’s leadership started by dragging its entire management community into dedicated sessions that weren’t product demos—they were exercises in reimagining accountability. The governance came baked-in from day one, not bolted on after a scandal. Enterprise authentication, privacy assessments, and a strict rule that AI permissions can’t exceed what an employee already has access to created guardrails that made experimentation safe.

With leadership setting direction and governance providing cover, employees got something rare: permission, structure, and dedicated time to redesign their own workflows. The result is a workforce that spends hundreds of collective hours each week tearing apart old processes. A claims handler no longer burns hours assembling a pet insurance file—a Workspace Agent now reviews invoices, checks policy conditions, flags anomalies, and serves up a traceable recommendation in minutes. The human still makes the call, but they start with a finished brief instead of a pile of paperwork.

Underwriting follows the same rhythm. Before an underwriter logs in, an agent has already triaged the queue, flagged missing documents, and highlighted red flags. This is what Univé’s leadership means when they say their competitive advantage isn’t the tool itself. It’s that thousands of employees are learning to reinvent their work every single week, building institutional muscle that centralized AI roadmaps rarely deliver.