New Way Now: Benifex gives benefits management a brainpower boost with Google AI

*Featured in this video:* James Fairhead, EVP of Product Engineering at Benifex *Executive summary:* Benifex, the global leader in employee benefits technology, is helping customers make smarter benefits decisions with Project Brain — a multi-agent system that acts as an intelligent consultant to help HR teams design competitive, engaging, and compliant benefits programs for employees around the world. Using Vertex AI, Google’s Gemini models, and AlloyDB, Project Brain draws on Benifex’s 20+ years of global benefits data and expertise — including consultant guides, competitive benchmarking data, and detailed benefit scheme information — to deliver trustworthy, data-driven insights that inform benefits strategies. Now, HR teams can simply ask questions using natural language and instantly receive expert advice on any country’s specific benefit regulations and market trends. The result: Faster market understanding, more confident decisions, and the ability to build competitive benefit packages in a fraction of the time. *Challenge:* To develop competitive benefits schemes, HR teams first need to fully understand a market’s benefits landscape — a costly, manual process that involves sifting through vast amounts of complex benefits information. Recognizing its wealth of expert intelligence contained answers that could help teams move faster, Benifex wanted to bring this valuable data to life and put it directly at their disposal. *Solution:* Benifex created a multi-agent system built on Google Cloud, designed to act as an intelligent consultant for HR teams, codenamed “Project Brain.” At its core, Vertex AI orchestrates multiple specialized agents, powered by Google’s Gemini models and trained on Benifex’s vast library of consultant guides on benefit norms, competitive benchmarking data, and in-depth benefit scheme information. These complex documents — some hundreds of pages long — are broken down into searchable chunks and stored as vector embeddings in AlloyDB, e