Before anything else: this post is about how I use AI to research and evaluate publicly traded companies as part of my own process. It is not financial advice, not a recommendation to buy or sell anything, and not a substitute for doing your own research or consulting a licensed financial advisor. What I share here is a personal AI investment thesis building process that works for me, not a system I am suggesting you replicate without your own judgment.
With that said, building an AI investment thesis has genuinely changed how I approach stock research, mostly by making the research phase faster and more structured than it used to be.
What an AI investment thesis actually is
An AI investment thesis is simply a structured, AI-assisted argument for why a specific company is worth owning at its current price. It answers three core questions: what does this company do and how does it make money, why is it likely to be worth more in the future than it is today, and what would have to be true for that view to be wrong.
That last question is the one most retail investors skip, and it is the one where AI assistance is most valuable, because it forces a structured consideration of the bear case rather than just building a list of reasons you like the company.
The AI investment thesis process I use combines three different tools at different stages, each chosen for what it does best.
Stage one: initial research with Perplexity
When I first look at a company I have not followed before, my starting point is Perplexity rather than Claude or ChatGPT. The reason is that Perplexity searches the web in real time and cites its sources, which matters for investment research where I need current information rather than AI trained knowledge that may be months old.
I ask Perplexity to give me a concise overview of the company’s business model, its main revenue streams, its competitive position in its industry, and any significant recent developments. I also ask it to summarize the main bull and bear arguments analysts currently make about the stock.
This stage takes about 15 minutes and gives me a factual foundation to work from before going deeper into the AI investment thesis building process. It also tells me quickly whether the company operates in an industry I understand well enough to evaluate, which is itself an important filter.
Stage two: financial statement analysis with Claude
Once I have a basic understanding of the business, I move to Claude for the financial analysis stage. I download the company’s most recent annual report or 10-K filing and paste the relevant financial sections into Claude with a structured prompt.
The prompt I use asks Claude to analyze the income statement, balance sheet, and cash flow statement together and identify the following: revenue growth trend over the past three to five years, gross and operating margin trends, free cash flow generation relative to reported earnings, debt levels relative to earnings and cash flow, and any significant one-time items that distort the underlying picture.
What Claude does here that would take me much longer manually is connecting the three statements to tell a coherent financial story. Revenue growing while margins are compressing tells a different story than revenue growing with stable or expanding margins. Claude surfaces those connections faster than I can when reading tables of numbers.
I also ask Claude specifically to flag anything in the financial statements that looks unusual, inconsistent between periods, or that warrants a closer look. This has caught things I would have missed on a quick read, including accounting policy changes that affected reported earnings comparability between years.
Stage three: building the AI investment thesis structure
After the financial analysis, I use Claude as a thinking partner to structure the actual AI investment thesis. I give it everything I have gathered so far, the business overview, the financial picture, my initial impressions, and ask it to help me build a structured thesis using the following framework.
The bull case: what are the two or three strongest reasons this company could be worth significantly more in three to five years. This includes industry tailwinds, competitive advantages, management quality signals, and underappreciated growth opportunities.
The bear case: what are the two or three strongest reasons this AI investment thesis could be wrong. This includes competitive threats, regulatory risk, execution risk, valuation risk, and any structural challenges to the business model.
The key assumptions: what specific things would have to be true for the bull case to play out. This is where I force myself to be explicit about what I am actually betting on rather than leaving it vague.
The monitoring signals: what metrics or events would tell me the thesis is playing out as expected, and what would tell me it is not and I should reconsider my position.
Claude does not build the AI investment thesis for me. It helps me stress test and structure the one I am developing. The difference matters because a thesis you did not build yourself is not one you actually understand deeply enough to hold through volatility.
Stage four: valuation work with ChatGPT
For valuation work I tend to use ChatGPT, primarily because it handles the back and forth of financial modeling conversations well and I can iterate quickly on assumptions.
I give it the company’s current financial metrics and ask it to help me think through what the stock is currently pricing in. What growth rate does the current price imply over the next five years at a reasonable discount rate. What multiple is the market applying to earnings or free cash flow relative to historical averages and peers. Is the current valuation pricing in optimism, pessimism, or something close to historical norms.
This is not a precise exercise. Valuation in investing is more art than science, and I am not trying to arrive at a single precise fair value number. I am trying to understand what the market currently believes about the company and whether I have a different, well-reasoned view that forms the final piece of my AI investment thesis.
If I do not have a differentiated view from the consensus, that is itself useful information. It suggests I do not have an edge on this particular company at this particular time, which is a legitimate reason not to invest even in a company I think is good.
What an AI investment thesis cannot do
Being clear about the limits of this process is important. An AI investment thesis cannot tell you whether a stock will go up. AI cannot access real-time earnings calls, recent management interviews, or insider sentiment signals without current data being provided. It cannot replace the judgment that comes from deep industry knowledge built over years.
Building an AI investment thesis makes the research process faster and more structured. It does not make investment decisions for you, and it should not. The judgment layer, deciding whether the thesis is compelling enough to act on and how much conviction you have, remains entirely human.
The honest benefit I have seen
The clearest benefit I have gotten from building an AI investment thesis for every company I seriously consider is that it makes me more disciplined about the bear case. Before I used AI as a thinking partner for this, I would naturally spend more time building the bull case than stress testing it. Having Claude explicitly help me construct the strongest possible bear case has made me more honest about risks I was inclined to minimize.
That discipline has improved my process more than any specific tool or technique.
A note on information currency
Investment research requires current information, and this is where general purpose AI assistants have a real limitation. Claude and ChatGPT have training data cutoffs and do not know about earnings releases, management changes, regulatory developments, or market events that happened after their cutoff dates. Always verify any factual claims about a company’s current situation through current sources, including the company’s own investor relations page, SEC filings, and reputable financial news sources.
Perplexity’s real-time search makes it better for current information at the start of the AI investment thesis process, but even there you should verify important facts directly with primary sources before making any investment decision.
Do you use AI in your investment research process? I would love to hear what is working for you in the comments below.