Governance as an Accelerator for Innovation

Univé’s successful AI integration relies on the principle that governance should enable, not restrict, experimentation. By embedding security, privacy, and accountability into the rollout from day one, the company built the trust necessary for widespread adoption. Key governance pillars include enterprise authentication, strict connector permission inheritance, and continuous monitoring. These guardrails ensure that AI tools cannot access data beyond what an employee is already authorized to see, allowing staff to experiment with custom GPTs and agents without compromising organizational security.

Shifting from IT Deployment to Organizational Transformation

Rather than treating AI as a top-down IT initiative, Univé engaged its entire management community in leadership sessions focused on rethinking work processes. This approach moved managers from simple tool approval to active enablement of team-led innovation. By providing employees with structure, time, and permission to redesign their own workflows, the company fostered a culture where staff members—not just centralized developers—build solutions. This has resulted in over 1,500 custom GPTs created by employees to address specific internal challenges across departments like claims, underwriting, and HR.

Agentic Workflows and Human-in-the-Loop Decision Making

Univé is moving beyond simple task automation toward agentic workflows that proactively prepare work. For example, in pet insurance claims, a Workspace Agent now assembles files, reviews invoices, and identifies missing information before a handler even opens the case. This reduces preparation time from hours to minutes. Crucially, this model maintains a strict human-in-the-loop requirement: the AI prepares the evidence and structures the context, but the trained professional remains fully accountable for the final decision. This allows employees to focus their expertise on judgment-heavy tasks rather than administrative data gathering.