Watchtower: Research Automation for Hedge Fund Management

This case study shows the client problem, what we built, and the result. It explains how we turned expert knowledge into a working AI product.

Replacing expensive legacy tooling with a custom AI research stack.

Max Berry - TracyChat — Dec 2025

Mark Farrington was spending $24,000 per year on clunky research software and still losing 13+ hours every week gathering and synthesizing data manually. Existing tools were rigid, expensive, and delivered limited intelligence.

We built Watchtower, a custom research automation platform that replaced Mark's legacy stack. The system includes a skeleton key scraper for directory sites and other sources, an intelligent synthesis agent for processing raw data, and a daily digest engine that delivers decision-ready insights by email.

The architecture runs on a modular scraping engine where Mark defines targets and patterns via a visual interface. Collected data flows through LLM-powered analysis that identifies relevance, extracts key points, and generates executive summaries. The digest builder aggregates findings across sources on a configurable schedule.

Mark needed a team that could extract his research methodology and translate it into AI that worked the way he thinks. He was not looking for generic automation. He needed a system that understood what signals mattered in his corner of the market and could monitor sources continuously without breaking.

92% cost reduction: research spend dropped from $24,000/year to $1,200/year, with 13 hours/week reclaimed and launch completed in 12 weeks.

Watchtower now delivers stronger intelligence through consistent coverage across more sources than manual workflows ever allowed.

Next: Emily Pearlman