Dutch Insurer Unive Rolls Out ChatGPT Enterprise, and Staff Build 1,500 Custom GPTs

Dutch Insurer Univé Rolls Out ChatGPT Enterprise, and Staff Build 1,500 Custom GPTs

One of the Netherlands' largest cooperative insurers has quietly become one of Europe's more instructive experiments in enterprise AI, not because of a single flagship system, but because it handed the tools to nearly everyone and watched what its own employees built. According to a case study OpenAI published on July 31, 2026, Univé has rolled out ChatGPT Enterprise across virtually every business function, and its workforce has responded by creating roughly 1,500 custom GPTs to reshape their own daily work.

The headline numbers point to unusually deep adoption. OpenAI says 97% of Univé's ChatGPT Enterprise licenses have been activated, and 85% of licensed users are active every week. Those active users average 40 prompts each per week, a figure that suggests the tool has moved past novelty and into routine. For a mid-market financial-services company, sustained weekly engagement at that level is the metric that matters more than any pilot demo.

The most striking operational result comes from an unglamorous corner of the business: pet insurance claims. Univé describes a Workspace Agent that assembles the claim file, reviews veterinary invoices, checks policy conditions, flags missing information and anomalies, and prepares a traceable recommendation before a human handler begins the assessment. Work that once took hours to prepare can now be ready for a decision in minutes. Crucially, the company stresses that the trained claims professional remains fully accountable for every final decision. As Univé frames it, AI prepares the work; people make the decision.

The strategy: build builders, not solutions

What separates Univé's approach from a typical software deployment is that leadership treated it as an organizational change rather than an IT project. The company brought its entire management community together for dedicated AI leadership sessions that, per the case study, focused less on product demonstrations and more on how work itself would change.

“Most organisations try to scale AI by building more solutions. We chose to scale AI by creating more builders,” said Yous van Halder, Univé's Director of Data & AI, in the OpenAI case study.

That philosophy is the throughline. Rather than requiring a detailed business case for every idea, Univé gave employees permission, dedicated time, and a governance structure and let them redesign their own work. The company says staff collectively spend hundreds of hours every week building custom GPTs, experimenting with Workspace Agents, and sharing successful approaches with colleagues. The 1,500 GPTs now span claims, underwriting, finance, HR, legal, IT, customer service, and management.

“Our competitive advantage is not that we use AI. It is that thousands of employees are learning how to reinvent their own work every single week,” van Halder said.

Governance, notably, was designed into the rollout from day one rather than bolted on afterward. Univé says enterprise authentication, connector permission inheritance, privacy assessments, security reviews, continuous monitoring, and clear human accountability were in place at launch. A key safeguard: permissions always follow the underlying enterprise systems, so an AI tool cannot reach data an employee is not already authorized to see. The company describes governance as an accelerator for experimentation rather than a barrier to it.

Underwriting is the second high-value use case. Before an underwriter logs in, a Workspace Agent reviews the incoming work queue, combines information from approved enterprise sources, identifies missing documentation, flags risk indicators, and highlights priority cases. The underwriter arrives to a queue that is already structured, with the relevant context and evidence attached, and can spend time on judgment rather than assembly.

Why It Matters

Univé is a useful data point in a larger shift: enterprise AI moving from isolated pilots to production across the workforce. The insurance sector in particular has been an early mover, and industry surveys in 2026 have pointed to OpenAI's models dominating enterprise AI stacks as insurers push from experimentation into deployment. Univé's adoption numbers, if they hold, are the kind of proof point that CIOs elsewhere will cite when arguing for broad, rather than narrow, rollouts.

More significant is what the case reveals about the “citizen developer” trend. Custom GPTs lower the barrier to building bespoke tools so far that a claims specialist or a legal analyst, with no engineering background, can encode their own workflow into a reusable assistant. Univé's 1,500 GPTs represent 1,500 small automations that never had to queue behind a central IT backlog or wait for scarce specialist resources. That is a structurally different model of software creation inside a company, and it shifts internal conversations, as Univé puts it, from “Should we use AI?” to “What should we build next?”

For a regulated industry that runs on accountability, the insistence that humans retain final authority over every claim and underwriting decision is not a footnote. It is the condition that makes the rest defensible to regulators, members, and auditors. The model Univé is testing, AI does the preparation and a licensed professional owns the outcome, is likely to become the template other insurers reach for.

What to Watch

The open question is whether 1,500 employee-built GPTs become a durable asset or a governance headache. Sprawl, duplication, version drift, and inconsistent quality are real risks when tool-building is democratized; the value of the guardrails Univé built early will be tested as the library grows.

The bigger bet is agentic workflows. Univé says today's prompting will evolve into Workspace Agents that proactively prepare recurring work across approved enterprise systems before employees even start their day. That moves AI from an on-demand assistant to an always-on colleague, and it raises the accountability stakes accordingly. Watch whether Univé can extend its human-in-the-loop discipline as the agents take on more of the workflow, and whether the productivity gains from claims and underwriting translate into measurable outcomes for its members. The company insists any competitive advantage should follow from its cooperative purpose, not the other way around. Whether that holds as the automation deepens is the story's next chapter.