FREE NEWSLETTER
Go to main Forum page »
If you’re inclined to create a spreadsheet — and I know that some of you are and some of you aren’t — you can put that spreadsheet directly into an AI program and ask it to analyze what you own. What happens next might surprise you.
Setting Up Your Spreadsheet
Your spreadsheet doesn’t need to be complicated. At minimum, include these columns:
That’s it. Four columns. If you want to go further, you can add current price, total value, and gain or loss — but the four basics will get you started.
What to Ask
Once you’ve pasted your spreadsheet into an AI program, ask it two specific questions:
What Will Likely Happen
The AI will almost certainly respond with textbook answers based on traditional portfolio theory. It may tell you that you have too much in one sector, not enough international exposure, or that your bond allocation seems light. It may suggest moving certain holdings from one account type to another.
Here is the important part: you do not have to follow any of it.
The AI doesn’t know your age, your income sources, your risk tolerance, your family situation, or your investment philosophy. It is working from general principles. Think of it as a starting point for a conversation, not a prescription.
What Happened to Mine
When I put my own portfolio into an AI program, it flagged several things that looked unusual by conventional standards — some sector concentration.
By textbook measures, these looked like problems. But each one had a specific reason behind it that the AI couldn’t know from the spreadsheet alone.
When I explained my reasoning, the AI engaged with it thoughtfully — pushing back where appropriate, affirming where the logic held, and asking questions that helped me think more clearly about positions I’d been holding on autopilot.
That back-and-forth was genuinely useful. Not because the AI told me what to do, but because explaining my reasoning to something that asked good questions forced me to be honest about which positions I owned with conviction and which ones I was simply carrying out of inertia.
Where It Actually Helped Most
The more valuable conversation happened around asset location — meaning which investments belonged in which types of accounts.
The general principle is straightforward: your highest-growth assets belong in your Roth IRA, where every dollar of appreciation is permanently tax-free. Income-generating holdings — dividend funds, REITs, bond funds — are better suited to traditional IRAs, where the tax deferral softens the impact of regular distributions.
When I asked the AI to evaluate my asset location specifically, it identified a holding in my Roth that didn’t belong there — an income-oriented fund that was better suited to a traditional IRA. I made the change, replacing it with a broad market growth fund. It was a simple swap, but one I probably wouldn’t have made without the nudge.
The Right Way to Think About It
AI is not a financial advisor. It doesn’t know your full picture. But it can act as a knowledgeable second set of eyes — one that will flag things worth thinking about, ask questions you hadn’t considered, and point out where you’ve deviated from conventional wisdom, intentionally or accidentally.
The key word is intentionally. If the AI flags something in your portfolio and your response is “I know, and here’s why I own it that way” — that’s a healthy outcome. You’ve confirmed your thinking. If your response is “huh, I’m not sure why I own it that way” — that’s valuable too.
The goal isn’t to have a portfolio that satisfies an algorithm. The goal is to have a portfolio you understand, that fits your life, and that you can hold through whatever the market delivers.
AI can help you get there. But the judgment still has to be yours.
I have played with Claude AI with my portfolio and, like your experience, the value came from identifying areas where I had gone on autopilot. The answer you get depends on the question you ask. I worked with each different account type separately. One day I did our combined Roth holdings; another day I did our combined Traditional IRA holdings, etc.
I provided our ages and our goal for that account and uploaded a spreadsheet of my portfolio holdings (ticker and balance for each holding). There was nothing in what I provided that identified us, although I guess AI could link them to my sign in??
The first account I loaded consisted of mutual funds. I asked for a critique of the portfolio without giving criteria. The response was almost exclusively focused on fees. For example, I had a highly rated active fund in a specialty area that has consistently out performed its peers. The recommendation was to exchange it for an index fund because of the high fees. I asked Claude to reconsider, looking back at past performance and Morningstar analysis. It changed the recommendation to “continue to hold.” As I loaded in other accounts, I made sure to ask for the evaluation to include fees and performance.
There were other areas where I questioned the recommendations. I had other instances where the recommendations changed, as well as some where Claude stuck to its guns. The value was in the dialog. The lesson is that you need to know why you are holding a specific fund or stock and push back if you think AI is missing something.
I was very impressed with the detail of the recommendations and the downloadable spreadsheets it provided.
I was dismayed when I found multiple math errors. When I challenged the errors, they were corrected but my first thought was how can I trust the big picture ideas if it can’t add 2+2. A comment in Gary Klotz’s recent post ChatGPT’s Portfolio Advice – HumbleDollar referred to an outside article evaluating AI for portfolio analysis. The author made it clear that AI programs are not calculators and they are likely to make math errors. I realized that I have a spreadsheet to do the math. What I’m looking for is a dialog about the big picture issues. Claude gave me that.
The biggest eye opener was in a Vanguard IRA account that I have neglected. I had accumulated the money prior to retirement in the STAR fund and have just left it there. It has grown nicely, but Claude pointed out that I could save a substantial amount in fees each year if I invested in the component funds rather than the blended fund. It specifically recommended a portfolio of funds to replicate the STAR holdings without having to buy them all. I felt that the first portfolio Claude recommended was missing the small/midcap space. It went back to the STAR holdings and concurred. Then it produced a new portfolio and a transition plan for me to move my investments.
A future project is to load in all my accounts together to look for overlap across the various accounts as well as to get asset location recommendations for different account types.
Be aware that several well known financial writers have tried AI on fictional accounts and have noted that when it comes to math they found errors, and AI itself has said it is trained on large language, ie on information, not calculations.
Which AI tools are the best to use? I’m concerned that my financial data would be scraped and used somehow by [who knows what or whom?].
WD, thanks for another very useful post.
I’ve asked AI to compare funds and received useful analysis. You have taken it several steps further; I was thinking of Morningstar’s X-Ray tool as I read. I may try this out.