AI Agent Skill-AI-Powered Investment Research Methodology

AI-Powered Investment Research

Complete Methodology — Validated on 8 Years of PDD Earnings Calls

Transform raw earnings call transcripts and financial data into structured, actionable research reports using a battle-tested four-phase AI framework.

Overview

Most investment research is shallow — a quick glance at P/E ratios and analyst ratings. This methodology teaches your AI assistant to do real fundamental research: extracting primary sources, cross-referencing management claims against actual results quarter by quarter, and producing multi-document research reports that professional analysts would recognize.

The Four-Phase Framework

Phase 1 — Data Collection: PDF extraction, transcript scraping, stock price aggregation, financial statement parsing. All with reusable Python scripts.

Phase 2 — Quantitative Analysis: Rolling P/E calculation engine with EPS alignment, data validation, and CSV/Excel output.

Phase 3 — Qualitative Analysis (AI Core Value): Quarter-by-quarter deep reading, cross-period pattern recognition, management credibility tracking, contradiction mining.

Phase 4 — Synthesis: Multi-document report generation, valuation framework, risk checklist, monitoring agenda.

Core Principles

Primary sources over everything — never let AI summarize analyst reports as "research".

Management credibility is an independent variable — track every promise against delivery.

Quantitative + qualitative cross-verification — numbers must match narrative.

Time series beats snapshots — a single quarter misleads; longitudinal comparison reveals truth.

Honesty about uncertainty — every conclusion includes "what could go wrong."

What's Included

SKILL.md — Complete AI workflow with all four phases and core principles.

PDD Case Study — Full real-world example (8,447 lines of raw input → 848 lines structured report).

Methodology Reference — Data source guides, naming conventions, and script references.

Requirements

• VS Code with GitHub Copilot Chat

• Python 3.7+ with pdfplumber, pandas

• Access to earnings call transcripts (Seeking Alpha, company IR, etc.)