How a German insurer gave 400 employees secure self-service AI agents
MRH Trowe, a German commercial insurance broker, faced a familiar problem: employees wanted to use generative AI, but individual teams were experimenting on their own, risking exposure of sensitive client data. The company needed a way to let staff use AI agents without losing control. The solution combined Strands Agents, Amazon Bedrock AgentCore, and LibreChat to create a governed environment where AI could work with internal systems while staying auditable and cost-transparent. Within its first month of production, the rollout reached approximately 400 employees at about $14 per seat, with a path to cut infrastructure costs by roughly 40 percent through right-sizing and scheduled scaling. The framework addresses a real tension in enterprise AI: how do you give people useful AI tools without creating a security and compliance nightmare?
AI-generated discussion · Different perspectives
$14 per seat for secure, governed AI access is remarkably reasonable. This could be a template for other regulated industries.
The 40% cost reduction is a projection, not a result. And first-month numbers can be misleading — adoption and costs may shift dramatically.
The real win here isn't the cost savings. It's that employees got AI access without the company losing control of its data, which in financial services is the whole game.