AI for Personal Finance
Where AI models do a great job, where they falter, and where they may always fall short.
August 20, 2026
For the better part of 2026, I have had an interesting side job.
I was hired as a financial expert to help a language learning model improve answers to personal finance questions.
Some people thought I was crazy:
“You’re helping AI take over your job?!”
During my time working for the “Dark Side”, I realized two things:
1. I am more excited about AI than I was before.
2. I am more confident in the staying power of the financial planning profession.
Those two conclusions may seem contradictory. I don’t think they are.
What did I learn from “teaching” AI?
Some prompts are answered much better than others
If you ask an AI model what a Roth IRA is, what a 401(k) does, how compound interest works, or the general difference between a stock and a bond, you are probably going to get a pretty solid answer.
Those questions have been asked and answered millions of times.
That is exactly the kind of problem AI handles well.
But people don’t usually lose sleep over what an IRA is.
They want to know something more like:
Should I convert part of my IRA to a Roth this year if I am retiring next year, selling a rental property, moving to another state, and my income is going to change substantially?
Now we have a different problem.
Every additional fact may matter. Some facts matter a lot. Some don’t matter at all.
Figuring out which is which is where things start getting interesting.
The critical errors aren’t obvious
The straightforward errors don’t bother me very much.
If AI tells you that the maximum IRA contribution is $75,000, you’re probably going to realize something went wrong.
But what about when an answer is 95% right?
The explanation is excellent. The concept is correct. The strategy generally makes sense.
But the key figure is from last year. Or the model is relying on a rule that changed. Or the answer quietly assumes the wrong filing status.
Those errors are much harder to catch because AI answers tend to be organized, confident, well reasoned and well sourced.
Herein lies an interesting conundrum with AI-generated financial advice:
The more complex and personalized the question, the more tempted you are to use AI—but those are the exact questions where an error can matter the most.
AI doesn’t understand what losing 30% feels like
This may be the bigger limitation.
An AI model can explain that the stock market has historically recovered from bear markets.
It can calculate the expected return of a portfolio.
It can tell you that selling after a 30% decline has historically been a bad strategy.
What it doesn’t know is what a 30% decline feels like to you.
It doesn’t know that you haven’t slept for three nights.
It doesn’t know that your spouse is frightened.
It doesn’t know that you watched someone close to you make—and then lose—a fortune, and that experience shapes how you think about money more than any rational argument ever will.
Those aren’t footnotes to a financial plan. They need to be part of your financial plan.
A good advisor isn’t just there to calculate the mathematically optimal answer.
Sometimes the job is recognizing that the technically perfect strategy is one the client cannot actually live with.
Sometimes the job is explaining the same thing for the fourth time because the market is down and this time feels different.
And sometimes the job is simply saying:
“Don’t make this decision today. Let’s talk first.”
AI can model numbers and calculate probabilities.
It cannot know what it feels like to be you.
Information was never really the hard part
AI can quickly parse tax code, contribution limits, or explain Roth conversions.
AI makes the finding and understanding of information dramatically simpler.
That’s a big deal, and I think it is overwhelmingly a good thing.
But the art of financial planning is applying all of that information to one particular situation:
Your income. Your family. Your business. Your taxes. Your investments. Your goals. Your tolerance for risk. You values. The things you haven’t even thought of yet. And, often, the reason behind the question you’re asking.
An AI model will answer:
“Should I sell?”
But a real advisor will first ask:
“Why are you thinking about selling?”
Those responses can lead to two completely different conversations.
James Keller, CFP®, CDFA®, is the founder of Keller Investment Advisors LLC, a registered investment adviser in Jackson, Wyoming. This article is for educational purposes only and is not intended as investment, tax, or legal advice.
Keller Investment Advisors LLC · PO Box 3904, Jackson, WY 83001 · 802-233-6275
Keller Investment Advisors LLC is a Registered Investment Adviser registered with the State of Wyoming. CRD# 342523. Registration does not imply a certain level of skill or training. Investing involves risk of loss.
