MRH Trowe Scales Secure AI Agents to 400 Employees

MRH Trowe, one of Germany's leading commercial and industrial insurance brokers, has given approximately 400 employees secure, self-service access to AI agents in its first month of production. The company, which operates primarily in Germany, Switzerland, and Austria, was among the first German insurance brokers to move to an exclusively cloud-based IT infrastructure. AWS is the preferred partner for that strategy, and the use of AI is a stated corporate goal aimed at streamlining internal processes and enhancing efficiency. The rollout demonstrates how a regulated financial services organization can deliver broad, self-service AI capabilities without sacrificing governance or control.

The Challenge: Self-Service AI Without Losing Control

As a commercial insurance broker in a highly regulated industry, MRH Trowe faced growing demand from employees across the business to use generative AI in their daily work. Individual teams were beginning to experiment with AI on their own, creating the risk of fragmented, unmanaged tools and potential exposure of sensitive client and insurance data. The company needed a way to let employees build and use AI agents themselves quickly and without deep technical skills, while keeping everything inside a secure, centrally governed environment. The core tension was clear: balance broad, self-service access to AI with the security, control, and cost transparency the organization required. Basic chat interfaces fell short because they lacked contextual responses grounded in internal data, support for multi-step workflows, institutional data integration, and centralized governance.

A Three-Part Technology Stack

MRH Trowe's solution brings together three technologies, each addressing a specific part of the challenge. Builders use Strands Agents, an open source SDK that creates agents in a few lines of code, handling orchestration and reasoning while scaling from a laptop to production so teams focus on the use case rather than infrastructure. To run agents in production, the company uses Amazon Bedrock AgentCore as a platform to build, connect, and optimize agents at scale with any framework or model. AgentCore was the deciding factor because it natively supports open source frameworks such as Strands Agents, isolates each agent session at the compute and filesystem level, and offers a consumption-based model that provides cost transparency as adoption grows. LibreChat serves as the user-facing layer, giving MRH Trowe a familiar, brandable chat experience with robust user management, token budgets to avoid unexpected costs, multi-model support, conversation management, and a customizable interface.

In Practice: From a Teams Meeting to Minutes in Minutes

The first agent MRH Trowe put into production turns a Microsoft Teams meeting into ready-to-use meeting minutes. In LibreChat, an employee asks in German for a recent meeting with a given participant. The agent finds the meeting on the employee's calendar, retrieves the transcript, and drafts a structured summary including date, participants, agenda, topics, and action items. Two design choices make this safe for a regulated broker. First, every request runs as the signed-in employee. LibreChat authenticates through Microsoft Entra ID and passes the user's identity to the agent server-side, so an agent can only reach that employee's own calendar and transcript. The identity cannot be set from the chat box. Second, processing stays in-region: the agents, model, and data all run in the AWS Europe (Frankfurt) Region, keeping client and meeting data in Germany.

Security Architecture and Deployment

The solution is deployed in a single AWS account inside a virtual private cloud, with employees connecting from the corporate network. Secure connectivity is provided through a transit gateway implementation combined with a zero-trust provider, keeping traffic on a private, dedicated connection rather than the public internet. An internal Application Load Balancer routes requests to the application tier. Amazon Bedrock AgentCore and the foundation models used in this solution are available in the Europe (Frankfurt) Region, though service and model availability varies by AWS Region. The MRH Trowe vision for the project is that every question should first be answered by AI before any human intervention, and repetitive processes should be automated by those who performed them in the past.

Cost and Outlook

The initial production cost of the MRH Trowe deployment was approximately $14 per seat in the first month. The company has a clear path to reduce infrastructure costs by about 40 percent through right-sizing and scheduled scaling, according to the AWS Machine Learning Blog. The consumption-based model offered by Amazon Bedrock AgentCore gives the team cost transparency as adoption grows, which is especially important for a regulated organization that must account for every expense. With the framework now in production at scale, MRH Trowe is positioned to expand agent use cases across its operations while maintaining the security, data residency, and compliance standards required in the German financial sector.

Sources

• AWS Machine Learning BlogHow MRH Trowe enabled secure self-service AI agents in financial services