RBC BD Simplifies Client Review Preparation

Overview

RBC Brewin Dolphin built a custom Databricks App orchestrating GenAI pipelines with Llama 3.1/3.3 and RAG to automate about 90% of client annual review meeting pack preparation, unifying data across five systems and saving an estimated 4,700 hours annually while cutting administrative costs by roughly 50% in some offices.

The challenge

Preparing an advisor for the annual client review meeting was a manual, time-intensive process. Data lived across five disparate systems, including CRM tools, trading platforms, SharePoint, and locally stored advisor documents, with no consistent format for compiling data, and this challenge was compounded across 14,000 annual reviews.

The solution

RBC Brewin Dolphin used Delta Lake to centralize structured and unstructured data with bronze, silver and gold layers, Unity Catalog for row-level security integrated with SSO, a custom Databricks App built with Dash for advisors, and Lakeflow Jobs orchestrating GenAI pipelines running 30 to 40 tailored LLM prompts per session (primarily Llama 3.1/3.3), plus AI Search and retrieval-augmented generation to produce contextual, advisor-ready outputs.

Reported business value

By automating approximately 90% of meeting pack preparation, Databricks automation saves 20 minutes per meeting pack, adding up to 4,700 hours saved annually, and is expected to reduce administrative costs by 50% for client reviews across 33 offices.