How Franklin Templeton Scales Investment Analysis with Agent Bricks

Franklin Templeton’s distribution model depends on its analysts delivering timely, accurate investment commentary to financial professionals navigating an increasingly complex product landscape. The firm offers mutual funds, ETFs, digital assets, and alternative investments — a universe that has expanded significantly and grown harder to cover with a fixed team. Seven analysts were responsible for producing manually authored, deeply researched notes for distribution teams and their financial advisor clients, leaving most of the product catalogue without current, tailored analysis.

The coverage gap was structural. Standard dashboards and shared libraries couldn’t adapt to each financial professional’s portfolio context, and early experiments with general-purpose foundation models fell short in a different way: they could generate text, but the outputs lacked grounding in Franklin Templeton’s proprietary scoring models and internal documents. For a regulated investment firm, ungrounded AI commentary was not a viable product. The team needed a governed, explainable framework that could blend proprietary data with analyst expertise and produce outputs that held up to compliance standards.

The solution, called SIGNALS, was built by Head of Portfolio Selection Mark Nigro and Lead Data Scientist Colin Zimmerman using Databricks Agent Bricks. They unified Franklin Templeton’s proprietary fund scoring algorithms and critical unstructured documents — prospectuses, fact sheets, portfolio manager commentary — inside Unity Catalog and the Databricks lakehouse. On top of this governed data layer, Zimmerman purpose-built several agents: a custom LLM trained on 200 analyst-authored notes, and two extraction agents that parse PDF fund documents into structured, machine-readable formats. Agent Bricks’ evaluation loop allowed the team to iteratively tune outputs for length, clarity, and compliance requirements until commentary consistently matched analyst-level quality.

The results were concrete. Coverage expanded from 200 manually supported products to hundreds of funds and ETFs. Each analyst saves more than two hours per week — reclaiming more than 15 collective hours weekly across the team. Within the first months of the broader rollout to more than 300 salespeople, field teams reported identifying $15 million in new product opportunities that would previously have taken days to surface. Financial professionals now receive instant replacement options and portfolio comparisons backed by data their teams can trust.

Franklin Templeton’s next step is extending SIGNALS’ AI capabilities into their web-based Portfolio Analytics Tool, making governed AI-generated insights accessible to a broader audience of financial professionals. The architecture — proprietary algorithms plus unstructured enterprise documents, all grounded in Unity Catalog — is designed to be replicated across geographies, channels, and product types as the firm’s AI ambitions scale.