AI for Finance

Where to actually learn Claude, Claude Code, and Codex for finance work — models, DCFs, variance bridges, forecasts, reporting packs.

32 candidates went in. 12 came out. Most "AI for finance course" results are content farms, affiliate listicles, or prompt libraries in a course costume.

Nothing here requires you to write code.

Start free — first-party

The best material here is written by the people who build the models, and none of it costs anything.

• Anthropic Academy — 22 free courses. Path that matters: Claude Code 101 → Claude Code in Action → Agent Skills. Zero finance content; you bring the domain.

• anthropics/financial-services — the repo behind the "10 finance agents" headlines, and the best free thing here. 252 markdown files to 17 Python: the skills are written instructions, not code. 3-statement-model, dcf-model, lbo-model, comps-analysis, audit-xls each ship a SKILL.md that reads like a desk manual, with reference notes on formula conventions and pulling from SEC filings. Worth reading for the methodology even if you never install one.

• Codex finance recipes — ten step-by-step workflows: DCF workbook, cash-flow low point, budget-vs-actuals, formula auditing, scenarios.
• OpenAI Academy — How finance teams use Codex — a monthly business review end to end, with dummy data and the exact prompts.

Then pay — finance-native

• CFI — AI Financial Modeling with Claude in Excel — 90 min, Tim Vipond. Spends as long auditing AI-built models as building them. Assumes you can already model.
• Udemy — Complete OpenAI Codex Masterclass — 11.5h, with finance, forecasting and analytics modules. The most technical thing on this page; skip the plugin chapters if you never want to touch a terminal. Wait for a sale.

Watch it fail — free

• Wall Street Prep — Ranking the best AI tools for financial modeling — publishes the prompt, the scorecard, the hallucinated line items and hardcodes.
• A CFO stress-tests Claude for Excel — statements, revenue, headcount, circular references, weak error controls.
• CFI on a three-statement model — rebuilds a stripped forecast, checks every formula against a benchmark.

India

No India-based course teaches Claude Code or Codex for finance. The market has packaged chat-window Claude — one logo in a 30-tool sweep. Useful for a CA doing tax research; not the same skill as pointing an agent at a workbook. Use these to get productive cheaply, get the agentic depth from the free list above.

• CA Monk — AI for Finance Level 1 · Free — 15h, CA Shivam Palan (ex-EY). Well made, and openly the front end of a paid Level 2.
• Thinking Bridge — AI & ChatGPT for Finance MasterClass · ₹1,999 — best value per rupee. PDF extraction, statement analysis, audit docs, tax research, Excel automation. Ignore the "30+ tools" framing.
• CFTE — GenAI for Finance Professionals, India cohort · ₹26,460 — literacy and a certificate, not hands-on modelling. Next cohort 10 Aug 2026.

Off-target despite topping every search: IIM Calcutta × TalentSprint FinTech, IIT Guwahati AI in Fintech, NSE Academy Financial Analytics & AI. Real institutions teaching classical ML, not LLM agents.

Cut, and why

• Developer-first — the Anthropic Cookbook's financial-applications notebook and write-ups on skill internals. Both excellent, both assume you read Python. (The financial-services repo is in, because its skills are markdown.)
• Marketing, not teaching — Anthropic's six financial-services skills (enterprise waitlist) and the Claude for Financial Services page.
• Right programme, wrong subject — Columbia × Wall Street Prep's AI certificate has no substantive Claude Code or Codex content.
• Priced like corporate training — Financial Edge's live one-day Claude modelling course, ₹72,590.
• Funnels in costume — paywalled "60 prompts" newsletters, ₹2,500 CFO prompt packs by email, content farms ranking themselves.
• Course mills — 53 minutes of generic prompting; Claude listed beside Bard and Bing as a logo.
• Hype demos — "Wall Street–level model in 10 minutes," no audit of the output, Skool link in the description.

How this list was built

1. Wide sweep across course platforms, vendor academies, YouTube, and newsletters. 32 candidates.
2. Verified each one at source — instructor, length, price, curriculum. Several didn't match their own search snippet.
3. Adversarial pass by a second model. Handed the full list to OpenAI's Codex with instructions to attack it: does the resource exist as described, is the Claude/Codex angle substantive or decorative, is this an SEO farm or a lead magnet. It cut 17 of 27 and surfaced 5 first-party resources the first pass had missed.
4. Cost filter. Anything priced like corporate training got dropped.
5. Separate India sweep, every rupee price read off the course page rather than a listicle.
6. Ranked by signal per hour — free and first-party first, then paid and finance-native, then people breaking the tools in public.

Last reviewed July 2026. If something here has rotted, tell me.