In 1998, a hedge fund called Long-Term Capital Management nearly broke the financial system. Its partners included two Nobel laureates and some of the finest quantitative minds on Wall Street. Their models were rigorous, their math was correct, and the Federal Reserve Bank of New York ended up organizing a rescue.
The math was not the problem. The problem was that the models assumed the future would resemble the data, and nobody in the room could imagine what it felt like when it didn’t.
I think about this whenever someone tells me they are using AI to make money decisions. Not because AI is wrong. Because it is so often right, and being right was never the scarce thing.
Information was already free
Ask almost any experienced investor what to do, and the answers are not mysterious. Save more than you think you need. Diversify. Keep an emergency reserve. Stay invested. Don’t panic when prices fall.
None of this is secret. It has been printed in books for decades. If knowing were enough, most people would be wealthy.
Money is one of the few fields where the gap between knowing and doing is the whole game. The investor who sold in a crash did not lack information. He had more of it than anyone in history, delivered every minute. What he lacked was the ability to sit still while it hurt.
AI is the next step in this story. It makes answers cheap, fast and fluent. That is a wonderful thing, and it also means the one input that was never scarce just became even less scarce.
What becomes valuable instead
When answers cost nothing, three other things get expensive.
The first is the question. “Which is the best mutual fund?” is a question about a product. “What is this money for, and when will I need it?” is a question about your life. AI will answer either one with the same polish. Only one of them is worth asking.
The second is doubt. A wrong number and a right number look identical on a screen. Suppose a tool tells you a strategy returned 12.4% over five years, and you later find that three of those years never happened, that they were back calculated from how the strategy would have done. The explanation sounds reasonable. But a strategy that never existed has an enviable record, because it never had to be held through anything.
The third is temperament, and no tool supplies it.
A mirror that talks back
Here is the part that worries us most.
Ask AI, “Why is gold a good investment right now?” and you will get a thoughtful list of reasons. Ask, “Why is gold a poor investment right now?” and you will get an equally thoughtful list. Both will sound balanced. Both will feel like research.
Most of us do not come to a conversation neutral. We arrive with a leaning, and a system that is fluent, instant and eager to be helpful can hand that leaning back to us, polished and footnoted. You walk away feeling more certain, not more informed. Those are very different things, and the difference gets paid for in money.
The physicist Richard Feynman said the first principle is not to fool yourself, and that you are the easiest person to fool. For most of history, fooling yourself took effort. You had to find the right book, the right friend, the right newsletter. Now it takes one sentence typed in the right direction.
Feeling informed is not being informed
There is a particular sensation that arrives after twenty minutes of good AI explanations. You understand derivatives, or small caps, or a new category of fund. It feels like understanding.
But real understanding of an investment includes knowing how it behaves in its worst year, and how you behave in that year. Neither of those fits in a clear paragraph. The fluency of the explanation is borrowed confidence, and borrowed confidence tends to show up as a larger position than your actual knowledge deserves.
The same goes for scenarios. Being able to run twenty projections in an hour feels like control. It is closer to the opposite. Uncertainty does not shrink because you modeled it more times. It just gets better formatted.
The plan you can keep
A man tells AI he is forty, has a stable income, a ten year horizon and a moderately aggressive appetite for risk. He means it. He would tell you he can live with a 25% fall.
Then the market falls 10%, and he finds out what he actually meant.
Nothing in his prompt was dishonest. He simply did not know himself, and no one can describe a feeling they have not yet felt. This is why the best financial plan is rarely the optimal one on the screen. It is the one you can still hold at three in the morning, in the second year of a bad market, when everyone around you has a persuasive reason to sell.
A slightly worse plan you keep beats a perfect plan you abandon. The math has no way of knowing which one you are.
How I would use it
Use AI for the parts of money that have right answers: the arithmetic, the definitions, the tax mechanics, the comparison of fees. It is superb at this, and it saves hours.
Be careful using it for the parts that don’t have right answers: what you want, what you fear, how much risk you can bear, whether to prepay the loan or build the retirement pool or keep more for family. These are not calculation problems. They are priority problems, and a calculation will happily dress a priority up as a result.
A few habits help more than any checklist.
Write down what you believe before you ask. If you cannot say what you expected to hear, you will not notice when you hear it.
Ask for the opposite case, then take it seriously for a full minute before dismissing it.
Decide in advance what would make you change your mind, and what you will do if this falls by a quarter. Decide while calm. You will not be calm when it happens.
And if you notice that you are rewording the question until the answer matches the one you wanted, stop. That is not research. That is shopping for permission.
The thing that doesn’t change
Every generation gets a new tool that promises to make money decisions smarter, and each one does make us more informed. The returns people actually earn, though, still depend on whether they stayed put when it was uncomfortable.
AI is a remarkable tool, and it will make all of us better informed. The question is whether we become better investors, and that has never been decided by what we know. It is decided by what we do when knowing is not enough.
