From Unsearchable documents to Instant Investment Intelligence

Years of Intelligence. None of It Searchable.

A firm's Investment Strategy Group (ISG) was sitting on years of valuable research it couldn't access. PDF archives held unstructured analysis — insightful but unsearchable. A data warehouse consolidated structured data from multiple systems — but getting answers meant building custom BI reports that required specialized skills and time nobody had.

The intelligence existed. The ability to use it didn't.

Built Around How ISG Actually Works

We started by mapping the real workflow — interviewing ISG users about their daily operations and documenting the questions they asked most. Not what the systems could theoretically answer, but what analysts actually needed to know and how they needed to ask it.

We migrated PDF archives from shared drives into the firm's Snowflake environment, making years of unstructured research accessible for the first time. Snowflake's semantic models, Cortex Search, and Cortex Analyst mine and contextualize both structured and unstructured data simultaneously. LLMs transform the output into natural language responses through a Streamlit interface that lets users ask questions conversationally — no SQL, no BI expertise, no specialized skills required. Continuous monitoring and tuning keeps responses reliable and contextually appropriate across different query types.

Questions That Took Hours. Now Answered in Seconds.

ISG users now query years of research and live portfolio data through a single conversational interface. Questions that used to require manual searches across multiple disconnected systems return answers in seconds.

But the bigger shift isn't speed. It's access. The solution surfaces intelligence that was previously too time-consuming to retrieve — analysis that existed but never influenced a decision because nobody could find it fast enough.

That's the difference between data that's stored and data that's actually used.