John Barrios · Yale School of Management Claude Code for Accounting Research

Claude Code for Accounting Research

A hands-on PhD course on agentic AI for empirical accounting research — five build modules and a closing seminar. It starts at first install, ends at a verified WRDS-to-Stata pipeline, and closes with a discussion of what these tools mean for the profession.

6 modules · 5 labs · 41 skills · 1 verification protocol

Each of the five build modules runs about three hours — a short lecture, a live demo, an extended lab, a debrief — and each ends with something you keep: a reproducible pipeline, a publication-style figure, a text-extraction skill, a regression-ready Compustat extract on your own WRDS account. Speed is only half of each module; the other half is what makes the speed usable in a paper — verification, paper trails, credential hygiene, and the measurement-validity judgment no agent supplies. Module 6 is different: a discussion seminar on AI, the production of knowledge, and the academic research market, with a position memo instead of a lab.

The course assumes no Claude Code experience. It does assume a laptop you control, a working WRDS account, and the willingness to run code live.

The six modules

Module Session The question it answers Materials
1 Foundations What kind of tool is this, and how do I stay in control? Lab · Slides
2 The Empirical Loop Can I go from idea to verified figure, reproducibly? Lab · Slides
3 EDGAR Text → Data, + Skills Can I build a text measure I’d defend to a referee? Lab · Slides
4 WRDS MCP at Scale Can Claude work my licensed data, safely, on my account? Lab · Slides
5 Writing, Safety, Verification What do I delegate, and how do I make my work auditable? Lab · Slides
6 Seminar: AI & the Research Market What will I still be paid for? Slides

Beyond the modules

The Skills catalog has 41 agent skills for accounting research — econometrics, data work, writing, review — each with a download and a one-line install command. The Tools guide compares Claude Code with Claude Desktop, Cursor, and Codex, with a decision table for picking the right one per task. The Prompt gallery collects every reusable prompt the course teaches, and the Verification protocol is the one-page checklist the whole course drills. Resources lists other useful tools and readings, grouped by the module where each is most relevant.

Taught by John Barrios, Yale School of Management. Course materials draw on cited sources credited on the About page; all student-facing content is original.