Proving that enterprise AI can thrive in strictly regulated environments, industrial insurance broker MRH Trowe partnered with AllCloud to launch a secure, cost-effective conversational AI platform on AWS Bedrock. The solution centralizes institutional knowledge for approximately 800 employees, empowers teams to build specialized domain agents via self-service, and sharply lowers ongoing token and licensing costs, all within full compliance of European data protection regulations.
When a business acquires over 35 insurance brokerages within a few short years, it gains more than new colleagues, it inherits distributed domain knowledge, fragmented tools, and significant data silos. To support this rapid expansion technologically, MRH Trowe partnered with AllCloud to build a scalable, enterprise-wide AI platform on AWS Bedrock in just four months. The platform consolidates expertise across all business units in one central location, keeps all data within the EU, and meets the strict requirements of the EU AI Act, GDPR, and DORA.
The result: Significantly lower operational costs, massive time savings in daily consulting, and an enterprise platform that gives newly acquired brokerages immediate access to cutting-edge AI.
MRH Trowe is one of Germany’s largest owner-managed industrial insurance brokers. Headquartered in Frankfurt, the company serves mid-market enterprises, industrial clients, and high-net-worth individuals across risk management, insurance solutions, and financial advisory services. Driven by a disciplined buy-and-build strategy, the group is growing rapidly, uniting newly acquired brokerages under the umbrella of One MRH Trowe.
MRH Trowe had already launched an internal corporate chatbot based on Azure and OpenAI models back in 2023. However, what was missing were clear governance structures, reliable cost controls, and a long-term AI strategy backed by an experienced partner to continuously manage and evolve the solution.
With every new acquisition under their expansion strategy, structural challenges intensified: valuable specialist knowledge, such as niche coverage for yachts or custom policy structures,remained isolated within individual teams or regional offices rather than being accessible across the entire enterprise.
At the same time, regulatory pressures mounted. European frameworks like DORA, GDPR, and the EU AI Act require strict oversight when handling highly sensitive M&A, client, and HR data. When Dr. Malte Polley assumed responsibility for the company’s AI strategy in late 2025, the priority was clear: replace the temporary setup with a well-architected, scalable, and fully controllable enterprise architecture.
Working side by side with AllCloud, MRH Trowe developed a custom, enterprise-wide conversational AI platform. Rather than opting for an expensive off-the-shelf SaaS product, the insurance broker chose the open-source conversational AI framework LibreChat, hosted on a highly scalable AWS cloud infrastructure. The backend leverages Amazon ECS (Fargate) with auto-scaling alongside Amazon Aurora PostgreSQL for managing user profiles, conversations, and agent configurations.
At the core of the architecture is AWS Bedrock, giving MRH Trowe flexible access to around 160 different large language models without requiring separate subscriptions for each. Four dedicated Knowledge Bases powered by Amazon Titan Embeddings natively process company data and unstructured documents such as PDFs, Excel spreadsheets, and Confluence pages.
The security and governance architecture integrates Entra ID with Role-Based Access Control (RBAC), ensuring employees only access data repositories approved for their specific department. Fixed per-capita token budgets provide reliable cost caps. Through an integrated Agent Creator, business teams can build custom AI assistants without any coding knowledge, while an internal data & AI community fosters knowledge sharing across departments through hands-on workshops and hackathons.
The new platform’s spring 2026 rollout transformed operations, driving significant efficiency gains across MRH Trowe.The impact is particularly evident in daily sales and advisory workflows: when preparing complex annual client reviews, a specialized AI agent analyzes existing contracts to automatically identify coverage gaps or cross-selling opportunities. Preparation time plummeted from 12 hours to just 30 minutes. Similarly, evaluating unstructured PDF policies, the dominant format in the insurance sector, dropped from 4 hours to under 30 minutes, making deep domain expertise instantly accessible to all advisors.
The project also set new financial and operational benchmarks. By migrating to AWS, ongoing per-capita costs dropped from $25 to just $9 per month, capping total platform costs at a stable $5,000 monthly even as usage scaled to 800 active users. Operationally, leveraging the AWS Cloud Development Kit (CDK) reduced feature deployment times from 25 minutes to under 5 minutes, enabling a remarkably fast time-to-market with the entire transformation completed in just four months, from initial scoping in December 2025 to go-live in late April 2026.
“As a client, AllCloud put me in the comfortable position of not having to make complicated structural decisions because everything was already so well prepared. All we had to do was nudge things slightly left or right, the strategic direction was crystal clear from day one.” – Dr. Malte Polley, Team Lead Data Analytics & AI, MRH Trowe