The challenge
A leading mortgage broker network had its knowledge base scattered across multiple systems, with no single place an adviser could go to find what they needed. Keyword search made this worse, returning results that were inaccurate or incomplete often enough that advisers stopped trusting it and fell back on calling a specialist directly.
That created a daily bottleneck. Every routine question about a lending product, a lender’s criteria, or a compliance rule pulled a second person away from their own work just to answer it.
The solution
NexaQuanta built an agentic AI RAG assistant with a natural language conversational interface, searching products, lender criteria, and compliance rules in real time. Rather than adding another system for advisers to learn, it sits inside tools they already use, available as a web widget and directly inside Microsoft Teams, with a Microsoft Copilot frontend.
The system runs on Azure AI Foundry and Azure AI Search, using OpenAI LLMs orchestrated through LangGraph agents, with a React.js admin portal supporting it behind the scenes.
The results
Adviser search time: cut sharply.
● Zero reliance on specialist lookups for queries the assistant now handles directly.
● One conversational interface now covers what used to take several lookups and a call to a specialist.
● Cost reduction, revenue growth, and improved customer satisfaction (NPS), alongside gains in scalability as adviser numbers grow.
What this means going forward
It replaced the fragmented system entirely, inside the tools advisers already use every day. The measure of success was simple: would advisers stop picking up the phone? They did.
