Univé rolls out ChatGPT Enterprise across its workforce

Univé, one of the Netherlands' largest cooperative insurers, serves millions of members across insurance, mortgages, financial services and risk prevention. In a customer story published by OpenAI, the company describes how it built an AI-ready workforce around ChatGPT Enterprise by combining leadership sessions, governance designed into the rollout from day one, and employee-led experimentation, rather than treating the deployment as a routine IT rollout.

The transformation started with leadership rather than with the tool itself. Instead of treating AI as an IT initiative, Univé brought its entire management community together for dedicated AI leadership sessions that skipped product demonstrations in favor of a harder question: how work itself would change, and what role managers would play in enabling that shift. Managers moved from approving individual AI initiatives to creating the conditions for their teams to innovate responsibly, with Univé framing the philosophy as building more builders rather than just building more solutions.

Univé reasoned that adoption at scale would only happen if employees trusted the platform, so governance was designed into the rollout from day one rather than added afterward. The framework combines enterprise authentication, permission inheritance from Univé's existing systems, so ChatGPT never gives an employee access beyond what they are already authorized to see, plus privacy assessments, security reviews, stated responsible AI principles, continuous monitoring and clear human accountability. With those guardrails established early, governance became what let employees experiment freely rather than a barrier that slowed them down.

With direction from leadership and confidence from governance, employees drove the actual adoption. Rather than requiring a business case for every idea, Univé gave staff permission, time and structure to redesign their own work; across the company, employees collectively put in hundreds of hours a week rebuilding workflows with ChatGPT Enterprise, building custom GPTs, testing Workspace Agents and sharing what worked with colleagues. ChatGPT Enterprise is now used across claims, underwriting, finance, HR, legal, IT, customer service and management, and employees have built about 1,500 custom GPTs to solve their own workflow problems rather than waiting on centralized development projects.

The clearest example is pet insurance claims. A Workspace Agent assembles the claim file, reviews veterinary invoices, checks policy conditions, flags missing information and anomalies, and prepares a traceable recommendation before a human claims handler starts their assessment, cutting preparation time from hours to minutes. The claims handler still makes and owns every final decision; the agent prepares the case, it does not decide it. Underwriting works in a similar way: before an underwriter logs in, a Workspace Agent has already reviewed the incoming queue, pulled information from approved enterprise sources, flagged missing documentation and risk indicators, and highlighted which cases need priority attention, so underwriters spend less time assembling files and more time on the judgment calls that need a person.

Univé reports 97% of ChatGPT Enterprise licences activated, 85% of licensed users active every week, and an average of 40 prompts per active user each week. Today's use cases mostly support employees on individual tasks; Univé says it is now looking at a further step through Workspace Agents, agentic workflows that would proactively prepare recurring work across approved systems before employees start their day, with every new application still run through the same governance framework.

Key facts

97% of Univé's ChatGPT Enterprise licences are activated, and 85% of licensed users are active every week, averaging 40 prompts per active user each week.
Employees have built about 1,500 custom GPTs for their own workflows, and ChatGPT Enterprise now touches claims, underwriting, finance, HR, legal, IT, customer service and management.
Governance, including enterprise authentication, permission inheritance from existing systems, privacy assessments and continuous monitoring, was built into the rollout from day one rather than added afterward.
A Workspace Agent assembles pet insurance claim files, reviewing invoices and policy conditions and flagging anomalies, before a human handler decides, cutting prep time from hours to minutes.
For underwriting, a Workspace Agent pre-reviews the incoming queue and flags missing documents and risk indicators before the underwriter logs in each day.

Why it matters

This is a large-scale, governance-first enterprise AI rollout at a regulated financial-services company, not a small pilot: Univé reports 97% of its ChatGPT Enterprise licences activated and 85% of licensed users active every week. OpenAI presents Univé as a template for how a deployment gets there: leadership sessions before broad tool access, governance designed in from day one rather than added afterward, and employees given explicit permission and time to rebuild their own workflows rather than waiting on a centralized IT project. It also shows agentic workflows doing production work, not just chat: a Workspace Agent prepares insurance claim files and underwriting queues before a human ever opens the case.

Who it affects

Univé's own staff across claims, underwriting, finance, HR, legal, IT, customer service and management, whose daily work is being restructured around AI-prepared files and queues. Univé's members, who rely on the insurer for insurance, mortgages, financial services and risk prevention, and whose pet insurance claims and underwriting cases move faster as a result. Other regulated enterprises facing the same trust and compliance questions before letting staff experiment with AI. And OpenAI, which publishes this as a customer story showcasing ChatGPT Enterprise and Workspace Agents inside a regulated industry.

How to use it

There is no product to buy or price to weigh here since Univé already licenses ChatGPT Enterprise; this is a playbook for adoption. Univé's own stated lessons for other organizations: treat AI as an organizational capability rather than an IT project; design governance into the rollout from day one instead of adding it afterward, letting permissions simply inherit from what an employee is already authorized to see; invest in leadership sessions as much as in the tool itself; give employees explicit permission, time and structure to redesign their own work instead of demanding a business case for every idea; and measure success by sustained weekly usage and employee capability rather than short-term productivity alone.

How solid is it

This is a customer story written and published by OpenAI, the vendor, rather than independent reporting, so read the framing accordingly. The three pull-quotes in the piece are not attributed to any named person, title or role. No date is given for when the rollout began or when the 97% and 85% figures were measured, no total headcount is given for Univé's workforce beyond serving "millions of members," and no licence count, contract value or underlying model version is disclosed. The adoption and usage figures come only from Univé and OpenAI, with no independent audit mentioned.

Risks and caveats

The efficiency claims are categorical rather than measured: the shift from "hours" to "minutes" for claims prep carries no numeric before-and-after figure. Governance is described only in general terms, such as privacy assessments, security reviews and continuous monitoring, with human accountability stated as a principle rather than tied to a specific audit or override process. As with any vendor-published customer story, it has little incentive to mention rollout friction or failed use cases, so the reported adoption numbers are best treated as self-reported rather than independently audited.

“Most organisations try to scale AI by building more solutions. We chose to scale AI by creating more builders.”

— Univé