Deutscher Maklerverbund's multi-agent AI platform triggers system actions in 70% of broker conversations

Overview

DEMV Deutscher Maklerverbund, a German insurance broker network serving more than 7,000 offices and 16,000 users, built Professional Works X (PWX), a multi-agent AI platform using Azure OpenAI Service, Microsoft Foundry, and Azure API Management. Each PWX agent manages a specific task area -- drafting communications, handling policy and claims operations -- allowing brokers to interact via chat or voice and submit claims, manage policies, and capture customer information directly through AI-powered processes, such as creating a claim from an uploaded photo of a police accident report. After a roadshow launch, weekly active PWX users grew about 65%, from around 850 to 1,400, with the platform handling 700-800 chats daily (peaking around 1,000) and about 70% of conversations triggering at least one system action such as retrieving data or preparing documentation.

The challenge

Insurance brokers faced rising expectations for faster, more proactive service within one of Europe's most tightly regulated financial environments, and DEMV's feature-rich 'Professional Works' platform, supporting more than 7,000 offices and 16,000 users, grew increasingly complex as adoption increased.

The solution

DEMV built Professional Works X (PWX) using Azure OpenAI Service, Microsoft Foundry, and Azure API Management, creating a coordinated multi-agent platform where each agent manages a specific task area, using a hybrid cloud approach that kept core systems of record in place while AI orchestration, reasoning, and interaction were built in Azure.

Reported business value

After a roadshow launch, weekly active PWX users grew about 65%, from around 850 to 1,400, with the platform supporting 700-800 chats per day (peaking around 1,000), and about 70% of conversations triggering at least one system action, freeing brokers for higher-value work.