FREE NEWSLETTER
Go to main Forum page »
To try to keep up with the 21st century, I recently started to use ChatGPT.
I have an excellent financial advisor, who is a Certified Financial Planner, and for over 40 years, I have also made a lot of investment decisions on a “do-it-yourself” basis. I have read the Humble Dollar for years as well.
An early request to ChatGPT was to develop a training plan for me to help me revive a long-neglected hobby. I was very pleased with the analysis and the specific suggestions that ChatGPT provided to me.
As a result of that successful experience, this week I asked ChatGPT to review and evaluate our IRAs, including our Roth IRAs, our other tax-deferred accounts, our taxable accounts, and a real estate investment. I entered the necessary information, and ChatGPT generated a very thorough report. I provided more information to clarify a few issues, such as for example, confirmation that investments in the taxable brokerage account had embedded capital gains. ChatGPT modified its analysis and conclusions, as appropriate, based on the additional information. It became an interactive dialogue between ChatGPT and me, which yielded a more complete review and very pertinent suggestions.
ChatGPT’s conclusions and advice were clear and well-reasoned. They were very similar to both what our financial advisor has told us and what I had viewed as next steps to implement. I had tentatively planned to take some of the steps that were recommended, and now I definitely will do so.
ChatGPT even graded our overall portfolio as an “A- to an A.” That validation of our financial advisor’s suggestions and my DIY efforts was reassuring.
I still struggle to keep up with the 21st century in other ways, but now am more comfortable with using ChatGPT. Of course, I will continue to read the Humble Dollar for its insights and information.
I came across this article from Alan Roth who tested AI for investment advice. He is one of my favorite investment writers.
Analyzing the Analysis: How Do AI Portfolio Recommendations Hold Up? – Articles – Advisor Perspectives
It’s been a minute but my early forays performing financial analyses with Chat GPT were very frustrating and seemed more focused on the CYA small print. But things change. Subsequently, I found Gemini to be more useful, but it is ‘eager to please’ and it is easy to go down the rabbit hole. Additionally, I frequently find myself reminding it of meaningful facts from earlier in the dialogue. I have not yet spent much time using Claude. Next on the list.
Most recently I started evaluating Roth conversions, customized to my specific portfolio construction and tax situation. With Gemini I arrived at what seems to be a good plan. Before acting on this, I plan on testing this two ways. Firstly, utilizing Claude and if that checks out, then subscribing to Boldin (formerly known as NewRetirement) which is financial modeling and projection platform to triple check the analysis.
Which will have to wait for summer travels to be completed 😊
I used Boldin a few years ago with a trial, downloaded the information then cancelled the membership. Once we both claim Social Security I will start an actual membership and run the numbers with those inputs to see where we stand. I realized that with Boldin’s Monte Carlo I no longer need to pay for a financial advisor to do the work.
I have asked Gemini for our portfolio analysis. I found it to be insightful and it validated my portfolio decisions.
Even though I do no use a financial Advisor I am going tomorrow for a meeting with one to get a portfolio risk analysis.
I am looking forward to see how they compare, agree, it disagree
I have used AI a lot in the past year doing various research on medical, financial and other topics. It is an extremely useful tool. One issue I have noted is that it tends to take what you give it and not ask questions, at least not many. When I see its result and follow up with clarifying data, it modifies it results accordingly.
A few months ago, I strained my lower back hitting golf balls. AI gave me a good assessment quickly, but my Ortho PT examined me thoroughly to rule out some things and then reached a similar conclusion. I felt much better by her diagnosis because I knew that she pinpointed my problem with a thorough examination.
AI is a great tool if you give it all the information, but good luck if you fail to give it all the information.
Oh, and AI does like to acknowledge my brilliance in our dialogues. Lol
I’ve used Gemini to research an auto immune condition I’ve been experiencing. It actually suggested a relationship to other auto immune problems I’ve had in the past. Lots to talk about with my rheumatologist. You can definitely go down a rabbit hole with these AI tools!
My experience with Chat and Gemini mirrored Gary’s. While I’m certainly aware of AI models being a bit sycophantic, I largely neutralize that by starting conversations with something like this:
“Act as a completely objective, fee-only financial analyst. Do not flatter my choices or validate my ideas just to be polite. Your job is to look for blind spots, misalignments, and hidden risks based strictly on the data I provide.”
Chat and GPT have been very helpful to me on a range of financial topics, in particular, positioning me to be ready to take RMDs in the most beneficial way possible. I’ve spent hours and hours asking follow-up questions to responses Chat or Gemini gave me.
They are not perfect. And you can’t take what they say at face value. But they are only going to get better. And they are pretty darn good right now, depending on how you tee up conversations.
If nothing else, they help focus the mind on critical issues that provide the best possible blueprint for a discussion with a human CFP or other professional.
If I were in the business of providing financial advice, I would be keeping a VERY close eye on AI models – and how younger people in particular were engaging with them.
Gary, just be mindful. I find most AI models sycophantic. In my opinion Gemini seems to be the worst, with ChatGPT holding the middle ground and Claude the most forthright, though I haven’t meaningfully tested Grok. I’d rather ask Claude for an opinion, as it normally doesn’t try to frame its reply to please the user to the same extent as the other two.
I’ve only used Gemini, and frankly, the way it remembers my information and engages with me on such a personal level, I find it a bit creepy. It is a bit sycophantic, now that you mention it. So yes, creepy sycophantic.
If Gemini ever answers with: “I’m sorry, Dan, I’m afraid I can’t do that.” Be worried. Be very worried.
I’m laughing my butt of, Mark, because I once cursed my first smart phone, and that’s exactly how she answered me.
Good point. In this situation, I viewed its comments and attitude as straightforward, analytic, and constructive, positive without becoming sycophantic.
I’ve used it too and found it very helpful. However I’ve also found it to be wrong sometimes, and the only way to know this was to already be pretty well informed myself. So I’m careful not to have undue confidence in its answers for something really important.
I remember a financial advisor recently saying lots of clients were coming with what AI had told them which makes for a good conversation starter. (Might have been Adam Grossman on The Long View actually.) I’ve also found this to be true with other professionals as well.