A 5-step Claude Project setup that turns any company's IR filings into a grounded research analyst. Locked to your uploaded docs, no speculation, no hallucinations. Copy-paste project instructions included.
First, The Disclaimer
This is not financial advice. This is the workflow I personally use to actually understand a stock before I buy it. It replaces the equivalent of an hour of Googling and squinting at 10-Ks with a focused analyst that only knows what you upload to it.
If you want zero chance of outside info slipping in, use NotebookLM instead. Claude Projects is an easier workflow for me, but the principle is the same either way: ground the model in the company's own filings, not the internet.
One Claude Project per ticker. Name the project the ticker symbol so you can find it later. Project Instructions are where you lock the analyst to your uploaded docs only. Use this exact text:
Why The Lockdown Matters
Without these instructions, Claude will happily mix the company's own filings with vague training-data memories and current news, which gives you a confident-sounding but unreliable answer. The lock keeps it grounded.
Before you dump filings, give the analyst context for the world this company lives in. Run this prompt as your very first chat in the project:
Save the output as a project doc titled Industry Overview so every future chat in this project pulls from it as standing context.
Go to the company's Investor Relations page and grab the last 5 years of:
• Annual reports (10-K)
• Quarterly earnings (10-Q)
• Earnings call transcripts
• Investor day slide decks
• Your own research notes if you have them
If you cannot find the IR page, every US public company files at SEC EDGAR for free. Search by ticker.
Drop everything into the project. Yes, all of it. Claude Projects can hold the full document library, and the more raw source material your analyst has, the sharper the analysis gets.
Naming Convention
Name files with the year and type, like AAPL-2024-10K.pdf or AAPL-2024-Q3-earnings-call.txt. Claude will cite filenames in answers, which makes its citations actually verifiable.
The single most useful pair of outputs you will ever get from this setup. Run both prompts, save both memos.
Six months from now, you will want to reread these two memos. The bear case especially. Half the time, the answer to "should I still hold this?" is in the bear memo you wrote when you first bought.
Every time a new earnings report drops, add it to the same project and run this prompt:
The Compounding
After 4 quarters, this project has a longitudinal record nobody on a stock-tip account is going to match. You start spotting drift, narrative shifts, and quiet pivots months before the market headlines catch up.
The Real Win
Most retail investors are buying tickers they heard on TikTok. The 1% who outperform do real research. This is how you do real research without paying $200/mo for a terminal.
The Only AI Masterclass You Need
If this guide helped, but you’re looking to go deeper, I got you!! My 30-Day Challenge takes you from saving AI tips you never use to actually building with AI, step-by-step.
I show you exactly how I automated two e-commerce brands, my social media, and most of my personal life, then hand you the agents, workflows & systems to do the same. I’m teaching you every single thing I know with one lesson and one build a day.
Join the AI Masterclass →© 2026 Mariah Brunner. All rights reserved.