Public recently announced AI Agents that could monitor your portfolio and execute trades based on your instructions. You can use plain English to tell agents what to do, like buy options if the VIX goes above a certain level. If you have excess cash in your portfolio, you can have an agent sweep it into a money market or something similar.

I think these are all cool features… but I also think it’s unnecessary. Perhaps dangerous.

As a buy and hold investor, I’m not monitoring the markets and making informed trades based on what I see. I’m also not sophisticated enough to know what to monitor and what to do when I see something worth acting on.

“But Jim, can’t you just learn?”

Yes, I can. But as a father of four, there are other things I’d rather be doing instead. I’d rather spend time with my kids, my lovely wife, exercise, or pursue my hobbies.

I think AI Agents can be powerful but I’m not comfortable with them buying anything for me, whether it’s yogurt or shares of VOO. I’m not sure how anyone would be comfortable given what happens with flash crashes (but they obviously still do it with high frequency trading).

This brings up a bigger question, and one you will have to answer for yourself, do you really need AI in your finances at all?

My experiences below are with ChatGPT on the ChatGPT Plus plan, which costs $20 per month. I’m not sure how it compares with the free in terms of the quality of its responses, but I suspect it’s not terribly different than the free tier for these purposes.

Do You Need AI? No.

Personal finance is quite simple and the advice for most Americans is the same:

1. Save more of your money,
2. Invest it wisely, and,
3. Have the patience to wait.

It’s why most personal finance advice fits on an index card.

It’s not that we don’t know what to do, it’s that life and our lizard brains get in the way. Derek Sivers famously said “If more information was the answer, then we’d all be billionaires with perfect abs.”

We know what we should be doing but we aren’t doing it. It’s not lack of knowledge, it’s lack of resources or discipline. We aren’t saving enough because we aren’t making enough or we’re spending too much. We don’t need AI to tell us that, we know it already.

But are there areas where a model could make our lives a little bit easier?

Is It Useful? Yes, But…

Do we need to give AI models an excessive amount of personal data?

This comes down to personal comfort. Some people are comfortable putting their home address and phone number into plaintext emails. Some people are not. Where you fall on that spectrum can determine whether you need AI.

While wouldn’t use an AI Agent to trade stocks, there are ways to use AI models to help identify blind spots.

I’m married, have four kids, and a couple of credit cards where almost all of our spending occurs. I quickly review the statements each month but I don’t study them closely (I use transaction notifications to notify me of potential fraud). For example, I’m not confirming that someone didn’t add a little extra to a tip on a bill if it total doesn’t look strangely too large.

But, by throwing a list of transactions into an LLM, I’m able to quickly ask it questions about my spending, trends, and anything that looks out of the ordinary.

And since you can easily download your transactions and obfuscate your personal information, it’s easy to drop it into a model without giving up too much personal information.

So I did.

NOTE: If you want to maximize the value of asking AI, you need to provide a lot of information not captured in a list of transactions. If you start working with a financial advisor, they should be asking you a LOT of questions to understand your overall situation and your goals.

If you give an LLM your situation, it still doesn’t know what your goals and that’s something you need to provide. Or get it to ask you. In the example conversations below, I offered nothing except our investment portfolios (though ChatGPT does know roughly my age). If you want to get the most out of it, you need to give it more than a list of transactions and stock holdings.

What Did It Find?

I had ChatGPT look at my credit card spending for the year and give me insight into where it was all going. It matched what we knew, a lot of our spending was on travel, but the totals were a minor shock to see aggregated into a 12 month period.

But ChatGPT didn’t have all the available information and so once you add that in, the numbers aren’t so shocking.

For example, for New Year’s Eve we often go on a group trip. We rent a large house for a few days and that charge lands earlier in the year. We pay it and our friends pay us back. As a result, our travel budget looks enormous but really isn’t.

ChatGPT did a good job of identifying some larger recurring charges we paid (our local fencing club, our Peloton membership) but those were charges we already know and happily pay.

As for advice, ChatGPT said that it would be a waste of time looking at those small charges when we have such big charges in categories like travel. If we put “guardrails” on our travel spending, we would benefit more than if we cut back on streaming subscriptions. But I don’t want to put guardrails on that.

Even when I explained the house rental charge, the same advice still applied because our travel budget covers six people.

Did I really need AI to tell me that? No, but it did give me permission to not be so nit-picky about our streaming subscriptions. (I still try to juggle them anyway)

I still had it identify a list of recurring subscriptions we are paying for to see if there were any that were obvious cuts. We pay for Disney, Spotify, Netflix, and Amazon Prime – and we use them all. I did think it was funny that through bundling we are getting Hulu twice… but it wasn’t something we could cut on its own.

It’s Great for Asking General Questions

One area where I find AI is useful is for asking general questions with a smidge of added context given the history of our conversation.

After I provided our portfolio, I can ask how it’s impacted by higher interest rates.

“How do rising Treasury Yields impact me?”

I have a folder with a screenshot of my Ally Invest portfolio and a Quickbooks download of my Vanguard portfolio, I drop it into a temporary chat and asked a simple question – “This is my portfolio, how might it be impacted by higher treasury yields?”

First, all my bond holdings fall in value because the price of a bond goes down as interest rates go up. The longer the term of the bond, the more it falls as rates go up because you’re locked into the bond until maturity. Fortunately, my bond funds all have relatively short average rates of maturity so the impact is minimal. Also, it is a small percentage of our entire portfolio so the impact is minimal.

Our stocks face a bigger impact from higher yields. A higher Treasury yield means stocks are less attractive because the risk free rate of return as gone up. If I got 3% last year from Treasuries risk free and now I can get 4% this year, I need stocks are less attractive. Experts like to phrase this as a headwind, since the yield itself doesn’t lower stock valuations, they just create conditions in which stocks are less attractive and in lower demand.

Where AI shines, mostly because humans won’t do this level of analysis for free, is breaking down the portfolio stock by stock. For example, it told me that our shares of AFLAC and Berkshire Hathaway would benefit from higher Treasury yields because insurance companies tend to do better in these periods. And all of my utility companies would fare less well given how much money they borrow.

It also does a “stress test,” explaining how a 1% increase in Treasury yield would likely affect the portfolio’s value. Here’s what it said:

“Actually Jim, there’s something else…”

Everything up until now were things I already knew. I know the relationship between interest rates and bonds and stocks. It was fun to see the deeper dive into the portfolio and learn that I had more exposure to AFLAC than I realized because I own it directly and through S&P 500 index funds.

But here’s where it identified a potential blind spot – I asked generally about Treasury yields but that question lacked sophistication. There are several Treasury yields, from 2 years to 30 years, and each one tells a different story.

For my stocks, ChatGPT said I needed to look at the 10-year real yield, not just the nominal yield. The nominal yield is what is quoted on CNBC. The real yield is the nominal yield minus expected inflation. You can see this reflected in the 10-year inflation-protected Treasury (TIPS).

The difference between the 10-Year and the 10-Year TIPS is what the market believes will be the rate of inflation for the next ten years. As of 9/8/2026, the 10 Year was at 4.79% while the 10 Year TIPS was 2.425%, implying the market assumes the annual inflation rate over the next ten years will be 2.365%.

That’s what you look at for how rates impact stock valuations. A higher rate there means your stocks face those pesky headwinds because investors need a higher rate of return to justify investing. If the risk free rate goes up, you need a higher return to take that risk.

In that case, ChatGPT highlighted a minor blind spot worth knowing.

Also, ChatGPT offered this up:

And because roughly 88% of your combined portfolio is equities, I’d personally give the 10-year real yield more attention than the nominal yield when trying to understand day-to-day valuation pressure on your portfolio.

It also cautioned:

One important addition: don’t interpret either yield mechanically. If the real 10Y jumps 30 bp and stocks don’t fall, that can mean earnings/growth expectations are improving enough to compensate. Yield + reason for the move + equity valuation together are much more informative than any single Treasury number.

In other words, stock prices move for a variety of reasons! Shocker!

“If I’m worried about interest rates, what should I do?”

I am not worried about interest rates, but I was curious what ChatGPT would suggest and it did offer up a few good ideas. It offered up a bunch of standard advice, like reduce equity exposure since it is at 88%, somewhat high given my age and plans.

It did make one good suggestion, a Treasury ladder based on our spending needs, which is a classic retirement cashflow planning strategy:

There’s also a potentially attractive strategy available given the size of your portfolio: match several years of expected withdrawals/spending with a Treasury ladder. For example, if you need $X annually from the portfolio, you could hold several years of that requirement in T-bills/notes maturing sequentially. That makes the question of whether the 10-year goes to 4%, 5%, or 6% much less consequential—you don’t have to sell stocks during a bad market to fund spending. Treasury ladders are a standard way of structuring fixed-income cash flows.

From here, you can tell ChatGPT what you want and it’ll build you that ladder with step by step instructions.

Keep Asking Questions!

The benefit of of AI and LLMs is that you can and should keep asking questions. If it doesn’t make sense, keep asking. The model will keep answering and it’s built to be very complimentary and kind, so you’ll never feel stupid doing it.

You want to keep asking questions also because LLMs can get things wrong. They can hallucinate. They can tell you things it thinks you want to hear. If things don’t make sense, keep asking. There’s no need to feel embarrassed about not knowing because the LLM doesn’t care.

Whereas you might just nod when a financial advisor says something you don’t quite get because you don’t want to seem ignorant, no need for that with LLMs – ask away!