David Bellairian on AI Stock Trading: How to Use Grok or Claude, Some Math, and One Strategy to Invest Like a Professional

My name is David Bellairian, and I run a private investment firm in Los Angeles. I use AI every trading day, and I want to explain what that actually means, because the version most people imagine, where you ask a chatbot what stock to buy and it tells you, is the one version that does not work.

Grok and Claude are two of the best analysts I have ever had access to. They are also terrible oracles. Understanding the difference is the whole game.

What an AI is good at, and what it is not

An AI model does not know what a stock will do tomorrow, and if you ask, it will either hedge or make something up in a confident voice. That is not a flaw you can prompt around. Nobody knows what a stock will do tomorrow.

What the model can do, better and faster than any junior analyst, is read. Give it a 10-K and it will pull out revenue growth, gross margin, free cash flow, debt maturities, and share count changes in seconds. Give it five years of financials and it will compute the ratios, spot the trend, and tell you which numbers moved and why. Give it your own rules and it will apply them to a hundred companies without getting tired or bored or emotionally attached to any of them.

So the first principle is simple. Use AI to analyze, never to predict. The prediction is your job, and honestly, it should be a small part of the job.

How I actually use Grok and Claude

Here is a normal session for me. I paste in a company's latest quarterly and annual filings and ask for a one-page summary of what changed: growth rate, margins, cash generation, balance sheet, and anything unusual in the footnotes. Then I ask it to compute the things I care about and show the math: price to free cash flow, return on invested capital, debt to EBITDA, and how those compare to the company's own five-year history.

Then I ask the questions a skeptic would ask. What would have to be true for this business to be worth 30 percent less than it is today? What is the bear case in three sentences? Which number, if it went the wrong way, would break the thesis? A good model will answer those directly, and the answers are where the real work is.

The one prompt I never use is "should I buy this." The model will give you an answer, and the answer will be worthless, because it does not know your time horizon, your risk tolerance, your other positions, or your tax situation. It is not being lazy. It is being honest about what it cannot know.

The math that matters more than the picks

Most retail investors lose money not because they pick badly but because they size badly. They put too much into one idea, hold a loser until it becomes a catastrophe, and take profits on winners far too early. All three problems are arithmetic, and AI is very good at arithmetic.

Three numbers to have the model compute for you before any trade:

Position size. Decide the most you are willing to lose on one idea as a percentage of your account. I use a small single-digit number. Then divide that dollar amount by the distance between your entry and your stop-loss. That is how many shares you buy. Not more.

Expected value. If you think a trade has a 45 percent chance of gaining 20 percent and a 55 percent chance of losing 8 percent, ask the model to run the math. That trade is positive expected value, because the wins are bigger than the losses. Many trades that feel good are not, and the model will tell you that without flinching.

Recovery math. A 10 percent loss needs an 11 percent gain to get back to even. A 50 percent loss needs 100 percent. This one fact is why I cut losers fast. Ask the model to show you the table once and you will never forget it.

Pick one strategy, and only one

If you are starting out, the worst thing you can do is run three strategies at once. Pick one, write down the rules, and let the AI hold you to them. The one I would hand a beginner is trend-following with a hard stop, because it is simple, it has decades of evidence behind it, and it forces the discipline that matters most.

The rules fit on an index card. Buy quality companies, the kind that pass the financial checks above, only when the stock is already trending up and making new highs on rising volume. Set a stop-loss the day you buy, at a fixed percentage below your entry, and never move it lower. Set a profit target or a trailing stop so winners get room to run but do not round-trip back to zero. Size every position with the math above. Review every open position against these rules weekly, and have the AI do the review so it is done without emotion.

You will miss the bottoms and you will never sell at the top. That is fine. The strategy does not win by being clever. It wins by never letting one bad idea take out the account, and by staying in the ideas that keep working.

Why this is the new way of investing

The old way was to have an edge in information. That edge is gone. A teenager with Grok can read a filing as fast as a Wall Street analyst.

The new way is to have an edge in discipline, and AI is the first tool that gives ordinary investors that edge on demand. It does the reading, it does the math, and it applies your rules without hoping, rationalising, or giving a losing position one more week. What it leaves you with is the only part that ever really mattered: deciding what your rules are, and having the honesty to be held to them.

I would rather be held to them. It is the whole reason I started.

David Bellairian is the founder and CEO of Glencore Associates, a private investment and asset management firm in Los Angeles operating across public equities, physical gold, income real estate, a commercial vehicle fleet, and a core private portfolio. He is also the founder of AIDiscover, an AI-powered marketing and brand discovery agency. More at davidbellairian.com and on his newsletter at davidbellairian.substack.com.

This essay is for education only. It is not investment advice, and nothing here is an offer or solicitation. Glencore Associates does not manage capital for outside investors.

Originally published on David Bellairian's Substack.