OpenAI publishes a case study on Univé, a Dutch insurance company that integrated ChatGPT Enterprise into its operations. Three pillars: structured governance, leadership from the top, innovation driven by employees.

For anyone bringing AI into their organization, the piece is worth reading for the constraints it reveals. A regulated organization encounters compliance requirements, training needs, role definitions, and limits on what data can be fed to the model. Structure comes before the brilliant use case.

Just a few weeks ago OpenAI formalized its enterprise platform. The Univé case is the first concrete testimony of a large, regulated company using it in production, rather than in a demo.

The piece is published on OpenAI’s blog, so it should be read with the vendor filter in mind. What remains useful is the organizational architecture, rather than promotional metrics.

If your organization is at the same stage, the course on bringing it to the team starts from the same principle: if only one person can make the process work, you’ve built a dependency, not a method.

In detail

Univé is a Dutch insurance group with a broad customer base. The case study published by OpenAI on its blog presents it as an example of large-scale adoption of ChatGPT Enterprise, the version of the assistant designed for organizations with governance, compliance, and data security constraints.

The product has been on the market for some time. What the case study describes is the organizational journey: defining who can use the tool and for which tasks, establishing rules on what data can enter prompts, training people on judgment about when to trust and when not to, rather than on technical usage.

For an insurance company, the regulatory framework is strict. Customer data is sensitive. Decision-making processes must be traceable. AI adoption cannot start with a single employee secretly testing the tool: it requires a level of structure that decides upfront what is permissible.

The three pillars described have practical meaning. Structured governance means written policies: what can enter prompts, what cannot, who approves use cases. Leadership from the top means the mandate comes from above, not from individual enthusiasts. Without top-level support, adoption remains episodic. Bottom-up innovation means the best use cases are found by people doing the work, not by those observing it from outside.

The case emphasizes a point our course on bringing AI to the team addresses directly: adoption is an organizational problem before it is a technical one. Governance comes before tools. Real innovation emerges from employees finding use cases in their daily tasks, not from the digital team imposing solutions from above.

This is also the limitation of the case study. Published on OpenAI’s blog, it tells a success story. We don’t know what didn’t work, which use cases were discarded, how much time the journey took. The piece doesn’t report numbers that would allow us to assess real impact: how many employees use it, how often, on which tasks, with what measurable return.

Still, it’s a signal. Enterprise case studies on AI are beginning to focus on governance, roles, and training, not just productivity. For those working in a regulated function, the question has shifted from “if” to “how”, and the “how” involves written rules before granting access.