How One Investment Firm Is Building AI Analysts

Key Takeaways

• Phoenix built custom AI screening named Himilco.
• AI tools cut stock research time from months to weeks.
• Workflow automation links AI with databases via n8n, LangGraph.
• Human judgment remains essential for intuition and future forecasting.
• AI extracted pricing power insights from Softcat documents.

Summary

Phoenix Asset Management showcased how it is embedding AI agents into its equity research workflow, creating custom tools such as the Himilco screening system and an AI‑driven version of its DREAM evaluation model. The firm leverages large‑language models, workflow platforms like n8n and LangGraph, and data‑search tools such as Supabase to automate document ingestion, comparative analysis, and scenario testing. These AI assistants accelerate knowledge ramp‑up, expand data coverage, and have already cut research cycles from months to weeks, while still relying on human judgment for intuition and forward‑looking assessments. The initiative positions Phoenix as a front‑runner among boutique managers in AI‑enhanced investing.

Pulse Analysis

The rise of generative AI is reshaping how boutique asset managers source ideas and conduct due diligence. Phoenix Asset Management’s experiment illustrates a pragmatic path: rather than replacing analysts, the firm equips each researcher with a personal AI assistant that can ingest earnings calls, regulatory filings, and market commentary in seconds. By integrating large‑language models with orchestration tools like n8n and LangGraph, Phoenix creates repeatable pipelines that transform raw data into actionable insights, a capability that previously required weeks of manual effort.

Concrete outcomes underscore the value proposition. An AI‑driven screening engine, dubbed Himilco, surfaces qualitative traits beyond simple valuation metrics, flagging opportunities such as TrustPilot’s divergent adoption rates across the UK and US. In the gambling sector, the system quantified parlay margin differentials and identified FanDuel’s revenue advantage. For Softcat, the AI compiled a multi‑source knowledge base and distilled pricing‑power dynamics that informed the firm’s investment thesis. These examples demonstrate how AI expands analytical bandwidth, improves memory retention, and enables rapid task switching—attributes that human analysts alone struggle to achieve.

Nevertheless, Phoenix’s approach acknowledges the limits of automation. While AI excels at data aggregation and pattern recognition, nuanced judgment about future market dynamics, regulatory risk, and macro‑economic shifts remains a human forte. The hybrid model—AI handling repetitive, data‑heavy tasks and analysts focusing on strategic interpretation—offers a scalable template for the broader industry. As more firms adopt similar architectures, we can expect a compression of research cycles, heightened competition for alpha, and a redefinition of analyst skill sets toward AI‑orchestration and critical thinking.