Building a Proprietary AI Research Engine at ValueSquare

The Challenge ValueSquare, a prominent Belgian value investing fund, relies on deep fundamental analysis to uncover undervalued companies across Europe. Their edge lies in reading what others ignore—footnotes, risk factors, and management commentary. However, as they sought to scale their coverage using Generative AI, they hit a technical wall. Feeding raw PDFs into Large Language Models (LLMs) resulted in hallucinated data and lost context, while building an internal OCR pipeline to convert thousands of reports was distracting their engineering team from the core investment logic.

The Solution ValueSquare integrated FinancialReports’ European Markdown Feed directly into their internal research stack.

1. RAG-Ready Data: Instead of wrestling with PDFs, ValueSquare ingested our pre-processed Markdown files. This format preserves document structure (headers, tables, lists) while stripping away visual noise, making it the perfect input for Retrieval-Augmented Generation (RAG) pipelines.
2. Pan-European Scope: The feed covered the entire European market, allowing their models to screen for value signals across diverse jurisdictions without worrying about varied reporting formats.
3. Automated Ingestion: By connecting to our API, they automated the flow of new filings directly into their vector database, ensuring their AI models always reason over the freshest data.

The Result ValueSquare successfully deployed a proprietary "AI Analyst" that pre-screens thousands of companies for specific value criteria. The system automatically summarizes risks, extracts non-GAAP metrics, and highlights discrepancies in management tone. By outsourcing the data structuring to FinancialReports, ValueSquare accelerated their AI roadmap by months and secured a sustainable informational advantage.