The only AI case study you need to read this quarter
Most enterprise AI rollouts follow a depressingly familiar pattern: IT picks a vendor, procurement negotiates licenses, employees get accounts, adoption flatlines at 20%, and executives wonder why their "AI transformation" never materialized.
Univé, one of the Netherlands' largest cooperative insurers, ran a completely different playbook. They hit 97% license activation and 85% weekly active usage of ChatGPT Enterprise across 1,500+ employees. That's not a pilot program—that's organizational transformation at scale.
What makes this case study worth your time isn't the vendor (though OpenAI will happily quote these numbers). It's that Univé's approach inverts almost every conventional assumption about how to deploy AI in a regulated enterprise.
Leadership as culture-builders, not approvers
Univé made a counterintuitive bet: they invested in leadership transformation before rolling out tools.
Instead of treating AI as an IT initiative, they convened their entire management community for dedicated AI leadership sessions. But these weren't product demos or vendor pitches. They were structured conversations about how work itself would fundamentally change—and what role managers would play in enabling that shift.
The result: managers stopped seeing themselves as approvers of AI initiatives and started creating conditions for responsible innovation across their teams. Yous van Halder, Univé's Director of Data & AI, frames it perfectly: "Most organisations try to scale AI by building more solutions. We chose to scale AI by creating more builders."
This is the hardest part to copy and the most important. You can replicate Univé's governance framework or technology choices. You can't fake leadership conviction that AI capability matters more than AI solutions.
Governance as accelerator, not gatekeeper
Here's where Univé's approach gets genuinely interesting for anyone working in a regulated industry.
They recognized that adoption at scale requires trust—and trust requires governance designed into the rollout from day one, not bolted on afterward. Enterprise authentication, connector permission inheritance that mirrors existing access controls, privacy assessments, security reviews, responsible AI principles, continuous monitoring, and clear human accountability all shipped together.
The critical insight: strong guardrails enable experimentation. When employees trust that the platform respects existing security boundaries and that permissions automatically inherit from underlying enterprise systems, they feel safe testing new approaches.
Governance became an accelerator for innovation, not a compliance checkbox. This completely inverts the typical enterprise AI narrative where security and legal teams are positioned as obstacles to overcome.
Every AI vendor will tell you their platform is "enterprise-ready." Univé's lesson is that enterprise-readiness is mostly an organizational design problem, not a technical feature.
Employees as builders, not users
With leadership providing direction and governance providing confidence, Univé made their third unconventional move: they gave employees permission, structure, and dedicated time to rethink their own work.
No detailed business cases required for every experiment. No centralized prioritization committees. Just clear boundaries, trusted tools, and organizational permission to redesign workflows.
The numbers tell the story: employees collectively spend hundreds of hours weekly redesigning work with ChatGPT Enterprise. They've created approximately 1,500 custom GPTs to solve internal challenges. AI now supports knowledge work across virtually every function—claims, underwriting, finance, HR, legal, IT, customer service, management.
This is the "creating more builders" philosophy in action. Univé's competitive advantage isn't that they use AI—it's that thousands of employees are learning to reinvent their own work every week.
What this looks like in production
The pet insurance claims workflow shows the pattern clearly. A Workspace Agent assembles the claim file, reviews veterinary invoices, checks policy conditions, identifies missing information, highlights anomalies, and prepares a traceable recommendation before the claims handler begins their assessment.
Work that previously took hours to prepare is ready for decision in minutes. Claims professionals start with a well-prepared case and focus on applying expertise. Critically, the trained claims professional remains fully accountable for every final decision. AI prepares; people decide.
Underwriting follows similar logic. Before an underwriter logs in, a Workspace Agent reviews the incoming queue, combines information from approved enterprise sources, identifies missing documentation, flags risk indicators, and highlights priority cases. The underwriter begins with structured context instead of spending valuable time searching and assembling files.
These aren't cherry-picked demos—they're production workflows processing real customer cases under regulatory accountability.
The metrics that actually matter
Univé's results reveal what successful enterprise AI adoption looks like when you measure the right things:
- 97% of ChatGPT Enterprise licenses activated
- 85% of licensed users active every week
- 40 prompts per active user weekly on average
- ~1,500 custom GPTs created by employees
- Claims preparation time reduced from hours to minutes while maintaining human accountability
- Adoption spans virtually every knowledge-based function
- Internal conversation shifted from "Should we use AI?" to "What should we build next?"
Notice what Univé doesn't lead with: productivity percentages, cost savings, headcount reduction. They measure sustained adoption and employee capability—because those are the leading indicators of organizational transformation, not lagging outputs.
The fact that employees average 40 prompts per active user weekly tells you AI has become part of everyday work, not a special-occasion tool.
What they're building next
Univé is now exploring agentic workflows through Workspace Agents—AI that proactively prepares recurring work across approved enterprise systems before employees start their day.
This is the natural evolution: today's use cases support employees during individual tasks. Tomorrow's agents will bring together information, surface relevant context, and create evidence-based starting points continuously.
Every new capability continues to be evaluated within their governance framework. Security, accountability, and responsible AI remain central to adoption as complexity increases.
The ambition isn't broader AI adoption for its own sake—it's a new operating model where AI becomes integral to how work gets done. As van Halder puts it: "AI will not replace your employees. But employees who learn to build with AI will redefine what your organisation is capable of."
Lessons for the rest of us
Univé's playbook won't map directly to your org—cooperative insurers in the Netherlands face different constraints than SaaS companies in San Francisco. But several principles generalize:
-
Treat AI as organizational capability, not IT implementation. The technology is table stakes. Culture and capability are the hard parts.
-
Design governance to enable, not prevent. Strong guardrails let people experiment responsibly at scale.
-
Invest in leaders as much as technology. Leadership creates the conditions for transformation. Tools don't.
-
Give employees time to experiment, not just access. Permission without capacity is performative.
-
Measure capability and sustained adoption, not productivity theater. Leading indicators matter more than quarterly cost savings.
The insight that will age best: "Most organisations try to scale AI by building more solutions. We chose to scale AI by creating more builders."
That's the whole game.