How to do investment research using AI
To do investment research using AI, run a consistent process: use AI to gather the filings and financials, assess business quality and the moat, build a conservative valuation, stress-test the bear case, and check the price against a margin of safety. AI's value is consistency — applying the same rigorous checklist to every company — but every output has to be traced to a real source before you trust it.
Good investment research is a process, not a hunch. The reason AI is genuinely useful here isn't that it's smart — it's that it's consistent. Doing investment research with AI means running the same disciplined checklist on every company, with AI handling the gathering and the arithmetic, so your judgment goes to the questions that actually decide the outcome.
This is the framework we use, and how AI fits each stage.
Stage 1 — Understand the business
Before any numbers, you need to know how the company actually makes money. Ask AI to read the latest 10-K and explain the revenue model, the customers, the cost structure, and the competitive landscape in plain language. The test of a good summary: could you explain this business to a friend in two minutes? If not, it may be too complicated to value with confidence — and "too hard" is a legitimate conclusion.
Stage 2 — Judge the quality of the business
Now assess durability. This is where most of the long-term return is won or lost. Have AI gather the evidence for a moat:
- Return on invested capital over 5–10 years (is it consistently above the cost of capital?)
- Gross and operating margin trends through a downturn
- Market share stability without ruinous discounting
AI can pull a decade of these figures in seconds. Your job is to interpret them — a single great year is luck; a durable pattern is a moat.
Stage 3 — Value it conservatively
Ask AI to build a discounted cash flow on a skeptic's assumptions, and a reverse-DCF to reveal what the current price already implies. The output should be a range, not a single figure. Precision to three decimals on a number built from assumptions is how investors fool themselves — the honest answer is a band of plausible values.
Stage 4 — Run the margin-of-safety test
Compare the price to the conservative end of that value range. The margin of safety is the discount of price to value — and the discount you require should scale with how hard the business is to forecast. Predictable, wide-moat company? A smaller cushion is fine. Cyclical or opaque? Demand much more, or pass.
Stage 5 — Attack your own thesis
The most important AI prompt in research is adversarial: "What would make this a bad investment?" Use AI to build the bear case, identify the assumptions most likely to be wrong, and surface risks buried in the filings — covenant terms, customer concentration, accounting choices. Research that only confirms what you hoped isn't research.
The guardrails that matter
Three rules keep AI honest:
- Trace every number to a source. AI can misread or invent a figure. If it isn't in the filing, it isn't real.
- Demand reasoning, not verdicts. An answer you can't inspect is an answer you can't trust.
- Keep the decision human. AI ends one step before the buy. The judgment — and the responsibility — stay with you.
Doing it at scale
Hand-prompting a chatbot through all five stages works for one company. The power of AI research shows up across many. Claremont Street runs this exact process as a 167-point framework on any ticker — filings read, quality scored, conservative value built, margin of safety tested, bear case included — so every company gets the same rigor, every time.
FAQ
How do you do investment research using AI?
Run a consistent process: use AI to gather filings and financials, assess the moat and quality, build a conservative valuation, test the bear case, and check the price against a margin of safety — verifying every figure against a real source.
What's the biggest risk of using AI for research?
Outsourcing judgment. AI can gather and analyze, but acting on a confident answer you can't inspect — or a number you haven't verified — is how AI research goes wrong.
Can AI replace a financial analyst?
Not for judgment. It replaces the slow gathering and arithmetic an analyst does, which frees human analysis for the parts that decide outcomes — durability and price.
How is AI research different from a stock screener?
A screener filters on metrics; AI reads the actual filings and reasons about the business, getting you closer to understanding a company rather than just sorting it.
Should AI research change how I value a company?
It should make you more disciplined, not more confident. Use AI to build conservative, range-based valuations and to expose weak assumptions — not to manufacture precision.
This analysis is for informational and educational purposes only and is not investment advice. Claremont Street is not a registered investment advisor. Do your own research.
Related reading:
Patient, AI-native investing — built for the long term.
Run any stock through the 167-point framework. Free to start, no card required.