Kitces & Carl Ep 156: How Are You Getting Value Out Of ChatGPT As A Financial Advisor?

***Editor's Note: Can't get enough of Kitces & Carl? Neither can we, which is why we've released it as a podcast as well! Check it out on all the usual podcast platforms, including Apple Podcasts (iTunes), Spotify, and Stitcher.

Show Notes:

• K&C 154: The Limits Where Technology Can't Scale Financial Planning Advice
• K&C 155: How Do You Know When Are You Adding Value Versus Just Trying To Justify Your Fees?
• K&C 35: Kitces & Carl Ep 35: Favorite Questions To Ask A Prospective Client To Build Trust In The First Meeting
• 50 Fires Podcast
• A.I. Chatbots Defeated Doctors at Diagnosing Illness

Kitces & Carl Transcript

Michael: Well, hello there, Carl.

Carl: Michael Kitces, how are you?

Michael: I'm doing well. I'm doing well. I'm excited to dive into today's episode. It's like I've been a little excited waiting for this because it's kind of a there's a reverse reveal moment.

Carl: Like a meta episode.

Michael: And the reveal was like the topic today is I want to talk a little bit about ChatGPT and AI and where this is showing up in advisory firms. I think we're pretty clear we're a year or two into this thing. We haven't all been vanished off the career map. We still seem to have jobs. It's not imminently replacing us, but we're all trying to figure out how to use it. And kind of the big reveal on that vein was so for those who listen to our last episode around you are you really adding value or just justifying your fees and engagement fallacy, we did that episode because Carl typed into ChatGPT, "What are four good potential episodes that we could cover on Kitces and Carl?" And that was one of ChatGPT's four suggestions. And so we did it. And it was a pretty good suggestion, as were the others that I don't think we're going to tell anyone because we really might do them in the other...some other good ones in there.

Carl: Right. Right.

Michael: But I forget what the full prompt is, Carl, that you had you had typed at the time. But it wasn't a lot more than we publish this podcast called Kitces and Carl for financial advisors. Please look it up on the Internet so you know what it is. ChatGPT, you're our executive producer. Please come up with four topics for us. And it did. I think it's come up with four punchy topics that would be a little edgy. And it did. So it gave us things like, "Are you really adding value or is it all just fluff that justifying your fees?" So with that as the fun sort of queue up of well, one way we're using ChatGPT is it's now feeding us show ideas, how are you using ChatGPT and AI? Where is it starting to show up for you?

How To Use ChatGPT As A Creative Partner [02:40]

Carl: Yeah. And look, maybe we'll just use that example as a way to sort of give advisors practical tools on how to use it. But I just want to start by saying it has changed my life. Like I am, the last four weeks, I am more productive than I can think of any time in my life and enjoying it more. And here's why. And I think I'm so fascinated by how profoundly uncurious most humans still are about using this tool. And I think it's largely because maybe people think, and it's even hard to describe, maybe people think you just said. Now, the reason it's really fascinating to me to use last episode as an example – the reason that worked is because I've been using ChatGPT to help me with a bunch of other projects. I think of it as a really great thinking partner. It helps me clarify my thinking. I don't go in and say, "Write a book for me." I say, "I'm working on a book about risk. Here's 18 articles I've written about it. Here's the transcripts from 20 behavior radio episodes. Tell me what you see. What lines do you like? Pull out some interesting quotes." And then that starts a whole discussion, it'll pull out something. So in this situation where we...and then I want to spend at least 10 minutes of our conversation on very specific applications for advisors because ChatGPT has become – and I don't care if it's Claude, ChatGPT, Perplexity, whatever tool you want to use. I've just been using...

Michael: ChatGPT is the proxy for AI the way that Google is the proxy for search.

Carl: Yeah. And it's become...or like Kleenex is the proxy for tissue, just to use an analog example. But ChatGPT is actually what I use. I use the pro account and I also use Claude and I also use Perplexity for different things. And it's the one app, ChatGPT the app, the desktop app and on my phone that I use every day almost the whole day. If I have an email I want to respond to I go into Chat, I copy the email and I say, and I literally just dictate, I just hit the microphone button on the Apple computer and I just dictate, "Hey, this is an email from Michael. Michael's asking about this. Here's a little bit of background. I really like working with him. We've been doing this, this and this. Here's the email he asked. Can you help me think through... In fact, will you craft an email using my language that says this, this and this?" So I'm not letting Chat just say, "Write a response to Michael."

Michael: So how does it know your language? What do you do?

Carl: So I've fed it a bunch. So you can take, excuse me, let's circle back to that. So first, the last episode, which is really interesting to me because Chat gave us a couple of ideas and we're like, "I don't like that one. How about that one? Could you give us a little bit more there?" We told it about the audience. We said these are forward-thinking financial advisors and financial planners who are interested in being pushed. They also love getting practical examples and tips, tactics to use, and they want their thinking pushed a bit, we know the audience pretty well. And then I also love telling Chat that it's really good at something, like, "You're the best executive producer in the world. Give us a couple of ideas." I find Chat, it's almost a little bit...

Michael: You're very nice to it.

Carl: Partially, it's Pascal's wager. Like, if this never happens, I want...

Michael: If it turns out that ChatGPT becomes sentient, you want it to remember you well.

Carl: I want it to remember me as nice, yeah. I actually think it operates better but I also like walking through the world where I treat everybody nice. So Chat gives us this idea. We have this conversation but guess what wasn't in this idea? Right? This is the unique interplay between creativity and humans and AI, engagement fallacy was not anything that Chat said. Right? And neither was utilization. That came from you and I brainstorming and bouncing ideas back and forth. And so I think now you can start to frame it's just a thinking partner. It's a creative partner. It's a, "I have a friend, for instance, who oh, okay." So that's how we used it for the last episode.

Using ChatGPT’s Summarizing Capabilities To Discover What’s (Really) Worth Focusing On [07:16]

Michael: So I want to shift a moment in a moment around how we use this advisors because we're not necessarily in the creative creator business that you are, but I actually do want to come back for a moment. You said at the beginning you've been more productive over the past four weeks in using this. So what do you mean? What's productive to you? What do you mean by that?

Carl: Yeah. Okay. Let me give you a couple of concrete examples. So my new framing is anytime I get stuck on anything, I go ask Chat. So it could be as stuck as like, "I don't know how to reply to this email." It could be as stuck as we're considering – based on our last conversation last episode, are you really adding value or are you just justifying fees? We're considering removing something from our service offering. What should we think of? We have this thing that we've always done and we're not sure we... Oh, how about this? We talked specifically about experiment design in the last episode around questions you could ask. So you could feed it. The survey questions you're, "Hey, we're going to ask our clients some questions. Here's our goal. We're trying to decide whether clients value something. We've got these five things that we're wondering about. Here's the questions we were going to send on a survey. You're the best survey data analyst in the planet, Chat. Would you please tell us of these five questions, do you like them? Can you think of any better ones?" That's a great use case.

Michael: Interesting.

Carl: Here's another productivity thing.

Michael: Here's my survey questions. Can you come up with better versions?

Carl: Can you see anything I should do differently?

Michael: Because I'm trying to measure this.

Carl: Yeah. Do you think I'm going to get what I want? You're a data scientist, Chat. You tell it this things. I had it help me write my outline for a 50 Fires episode. And so I said in my mind, I was like, "Who would be my dream executive producer?" It's Ira Glass. So I said you were the executive producer for Ira Glass's "This American Life" and Krista Tippett's "On Being" and you've helped produce the Rich Roll podcast and the Huberman show. That's who you are. Here's my guest. Here's what I was thinking of asking them. And I just literally am talking. Here's what I was thinking of asking them. I'm really interested in this and this and this. This is why we want them on the show. Here's the 12 normal questions we ask in 50 Fires. Take those 12 questions, take my ideas and do your magic. Help me design a show. So there's and then let me give you one last example.

Michael: And so it gets you to an output faster than you literally sitting there with a white pad.

Carl: Yeah. Not even. It's not even like I just can't even. Now, I'm literally like I can't imagine my life. Now let me be clear, the first answer is not normally the one I want. Like I'll say, "Oh, that's interesting. I really like this." I was most interested in asking...we just had Kara Swisher on it. I was most interested in asking Kara about what interesting conversations has she had with tech masters of the universe about money. What do you think about that? But here's one. Take this show. Go to YouTube. Does this show up on YouTube?

Michael: Yes.

Carl: Go to YouTube. Over on the right-hand side of YouTube there's a place for the transcript and it has the timestamped transcripts on most YouTube shows. Literally start at the top and just copy that whole thing. You don't have to clean it up. Don't do anything. Go to Chat, paste.

Michael: So I can't just tell Chat like, "Go to this website and read the transcript."

Carl: Yeah, so you copy and paste that. You can upload PDFs. So if you have a book you've written or you have client, this is another example. If I had a client education like a flyer or a pamphlet that I handed out like my flagship piece as a PDF, I would completely for sure tomorrow, I would upload that as a PDF and say, "Hey, you're our marketing and education expert. Here's what we're trying to communicate here. We doing a good job? What do you see that we can improve on? Anything you see we could take out?" I'd ask it specific questions. But if you go to the show, copy the transcript, paste the transcript into Chat and say, "Michael and Carl talk forever. Could you give me 2 or 3? Give me the 2 or 3 most valuable points that I need to know from this." I did that with a 2.5-hour episode of Huberman. And this is what I said to it. I said, "Pretend like you are the head of my... You are my chief of staff. And the business depends on me not missing anything from this 2.5-hour episode, anything important. It depends. It depends on me. If you miss it, the business will fail."

Michael: High stakes.

Carl: "I'm the CEO. I don't have time to listen to the whole 2.5 hours. Tell me what I need to know." It was incredible. In 12 minutes I feel like I knew what I needed to know. And it saved me 2.5 hours. Those are some examples.

How Financial Advisors Can Actually Use ChatGPT [12:44]

Michael: So now translate this for us to the advisor world. What do we do with this? Because, again, you live in a creative realm where I totally get it as creative thinking partner. So what are my advisor use cases? How do you fit this into our advisor world?

Carl: Yes. So let's just pick some. Any email you're stuck on, how to reply. That's an easy one.

Michael: So I copy, paste the client's email to me in the ChatGPT?

Carl: And again, let's just be clear here, take away all the identifying information. Just copy and paste.

Michael: Not the email and not the signature line?

Carl: Say to it, "I just got this email from a client. I'm unsure how to respond. I'm thinking of responding." I'd literally talk this into it. But if you type, that's great. I just hit the dictate button on my Apple computer and the dictate thing. I don't edit it. I don't even wait until the dictation is done before I hit enter.

Michael: Meaning your prompt is just a jumble of real-time talking. You're not finally crafting it.

Carl: Yes. This is natural language computing, right? This is where we all hoped we'd get. So I would say I got this email from a client. Let's just pick an example. This client relationship is kind of sensitive. I'm never exactly sure what they mean. But I really like them and I want to make sure I give them what they need. I'd love you to just look at this email real quick. Summarize it for me. Tell me what you see. What do you think they're asking for? Get that back. You're like, oh, okay, that's good. Hey, here's what I think the reply should be. Here's how I would have replied. Tell it. Based on what you suggested and based on what I suggested, we craft a reply for me. Now, I'm not suggesting you copy and paste that reply. Although I'm doing that a lot, because it's now my reply. I've now worked with... Chat just is my editor. But you just copy, paste that as a draft, that will save you anything you're stuck on. Right? You know those emails where you're like ah...

Michael: Yeah. Yeah.

Carl: That's not just a simple reply, one minute, yes, no. It's going to take a lot of thought. Any of those. That's one example. I think another example is what we already talked about, the show. Any book you want to read. Right? Anything that you need a summary of.

Michael: Oh, give me a summary.

Carl: Yeah I just did that. For some research I'm doing on a book about risk, I said, "Make me a list of the 12 most important books about this kind of risk." It's a specific kind of risk. It's the kind of risk you feel, related to imposter syndrome, it's the kind of risk you feel when you go to put something in the world. Right? Whether that's a business, a new first meeting, a sketch, a book, a poem, that risk is very intimate and I was trying to figure out what it's called. And I said, "Do you have any books to suggest?" It gave me a list of 12 books. I said, "Hey, that's a lot of books. Out of those, which ones should I have use? Should I just work with you for an hour back and forth summarizing, and which ones should I read?" And it gave me two to read and said, "Let's work together on these 10." So then I started a new Chat. And I actually I said to it, "Which one would you start with?" What’s the most important one to start with? Because the framework will help the others.

Michael: Right.

Carl: Gave me the one to start with. I went into a new chat and said, "Okay, will you work with me? Teach me everything I need to know." And I was planning on spending like an hour. I was planning on this being like half an hour, 30 to 60 minutes maybe two or three times. Let it bounce around in my head. I do this sometimes. Now I've done it a little bit in the car, audio, because Chat has an audio function. You can give Chat a whole book and say, "Teach me this book by quizzing me." I'm doing a bunch of avalanche training. I'm having Chat tutor and teach me everything I need to know in all the material. I uploaded 250 pages of a PDF.

Michael: And you have it quiz you, it's making up quiz questions.

Carl: Yes. I'm reading it still because it's really important. But it's creating the flashcards you used in college.

Michael: Yeah.

Carl: So those are some. And then back, like any marketing proposal, how about this? A first meeting brainstorm. Sit down with Chat and say, "Here's how we run our first meetings." You could download an episode that we've done here. Go download the transcript of an episode that Michael and I did on first meetings. Say, "Here's how we do first meetings. Here's what those crazy nut cases, Kitces and Carl say about first meetings. Here's where I'm hung up. I don't want anybody to cry in my office. I'm a financial person with a calculator. I don't know. Everybody's telling me I need to be emotionally intelligent." You can literally say that and then say, "Can you just work with me and help me understand how I can make my first meeting better?" And then you could say, "Generate an agenda for me with the exact questions in order that you want me to ask." And then just look at it.

I'm not suggesting that you're going to go walk in the room and do it. Although, I think if you did, you would probably be better than 99% of the advisors out there. But just use it as a template. So I just read, this is in "The New York Times," they did an experiment where they had Chat diagnose. I can't remember the number of patients. So that they knew the patient. They already knew the diagnosis. And they had Chat diagnose. They had doctors with AI diagnose and they had doctors. Chat did a better job than all of them. Then next would be the doctors with AI and then the doctors by themselves. So I'm just suggesting, start thinking of it as a tool.

I've told all my kids, "Please use this all the time. Start just using it as a thinking partner. Crisp, fast, speed things up a bit." Hey, tell me what I need to know. Look, you can't listen to all the podcasts you want to anyway. The ones that are really important to, listen to the whole thing. That's what it said to me about books. It was like, "Look, most of these books we can work together and it's going to take you six hours to read. It will take you an hour with me and you'll know it better." But then it said, "But some of them you're going to really want to notice the cadence and the way the author writes, the way they build the case, and those would be important to read." I get fired up about this because I'm telling you the future is already here. It's just unevenly distributed. That's the saying. Right? And your value is not going away. It's only getting better. Your presence, your ability to ask questions, your ability to be there, you're only getting more valuable. And the ability to quickly get through things as a thinking partner, as an editor, as a marketing assistant, has just gone through the roof and it's 20 bucks a month. What else? Sorry for ranting.

Using ChatGPT As An Iterative Partner To Get Better Responses [20:28]

Michael: So I'm struck by some of what you're saying here. I feel like a lot of the concern I still hear from any advisory firms essentially comes down to various versions of data security, private client privacy. And just, I'm struck, almost none of what you're talking about even gets into that. You're not in the... I've uploaded all my clients' private file, private client information, you construct a financial plan for them, even in your email example, you just copy the body of the email not their email address and their picture plate that has their name and contact information. You give me a summary of a book, which book should I read. I'm working on a marketing proposal. I'm brainstorming my meeting process. These aren't these aren't things that even get into client data-related stuff in the first place.

Carl: No, no. Yeah, I know. Yeah, yeah, I agree. I don't I don't think of it that way and I also don't think of it as... This is what I did for a long time because I have a coach that was really trying to get me to use an AI tool, mainly Chat, and I was like, "I've tried it. It doesn't give me a good answer." I think that's what most people do and I think that's a user problem. It's just prompt engineering, like asking the right questions. So once I started doing that...

Michael: “I didn’t sign up for this job to be a prompt engineer.” That's part of my challenge with this. I'm sure it would be amazing if you could be the perfect prompt engineer. But just do I have to accept that my future reality's that we're all going to learn to be prompt engineers?

Carl: No. I think where I got where I realized was prompt engineering is just a fancy word. To me, it's just talking. So I think even what you just said, "Hey, I'm not a prompt engineer. I don't know how to do this." But I know we have a... For you, I was just reading this great...listening this great podcast about somebody who wrote what they consider their best book, data heavy book, in 30 days. And it's mainly because they used AI to help with all the data because you can upload and it will run, it will build, run, it will build, program. It will think through program and run the regression analysis for you.

Michael: Okay.

Carl: So I'm simply saying anytime you get stuck like that, right, the first means is a really great example. That's maybe the edge of the creative example that I would use. When you said financial plans is like, man, if I could, if there's a way to have the data, all the confidential stuff removed, maybe mainly just the client's name, I would think it would be unbelievably fascinating to upload 20 financial plans and just say, "Hey, each of these are going to be different but I'm curious, are we ordering it correctly? Is the is the writing crisp and clear? Is it understandable to a client? How could I make this cleaner? How could I make it simpler? What suggestions would you have on what to eliminate?" That's another maybe big header is anything that you would think you need an editor or creative partner on, and advisors are really, generally as a group we're not particularly good at that and...

Michael: No, marketing is not a gift for most of us.

Carl: Or even any form of written communication. So just having an editor in your pocket for 20 bucks a month, it is at least make it a little better. Do you know what I mean? It may not be...

Michael: Well, I can envision, but you can get pretty far with that... “This is my ideal client profile, what are ways that I can market to them and how would I reach them?”

Carl: Oh my god, that's great.

Michael: Interesting marketing tactics, ideas, inspiration, for...

Carl: I think this is fascinating. Sam Altman, the founder of OpenAI and ChatGPT, when he was asked, "How are you going to make money?" he said, "This is the great thing, we're just gonna ask Chat." I would literally go in there and say, "Here's my ideal client. Here's where I live. Here's how I like to do things. I want to do it locally, or I only want to do it remotely, or I like LinkedIn, or I don't like X, or here they are, what would you do if you were me? Could you give me 20 topics for my podcast that dentists between the age of 35 and 47 would be interested in hearing?" I'm telling you it will be better than what you would get from a marketing agency, especially if you'll be comfortable saying, "I don't like that one," and then you'll see something you really like and you can grab it and go, "Hey, I love that phrase. Where'd you come up with that phrase? Could we do something more around that?" And then any suggestion it gives, I always push it at the end. I always say, "Hey, these are great. We can just run with these but I want to push you one more time, Chat. You're the best first meeting consultant on the planet. If I really wanted to make this really effective for a client, I want them to be seen, my goal is that they're going to walk to the car scratching their heads feeling like they've been seen for the first time financially…” That's exactly what I would say.

Michael: This is literally your prompt? This is literally the problem is how you ask it?

Carl: Yeah.

Michael: It is interesting to me, the one of the first descriptions I heard of ChatGPT when was first coming out is that it's just an intern that's pretty sharp but absolutely needs to be trained. Right? For anyone who's worked with an intern, you can have a really whip smart intern but just there's so many things they don't know because they maybe never done the job or had a job before that you just have to learn and practice being really clear and explicit and giving a lot of feedback of, "That really wasn't correct and this would be better and here's why." But the more iterations you do with your intern, eventually they really learn your process or your style or what you need to do and then it's great but you have to give a lot of feedback of, "Yes, this, not that," until they get it and that working with ChatGPT is functionally the same thing.

Carl: I think that's fair. I think that was especially fair six months ago. I think now it's moved to let's say that intern is your head of marketing. It's that level now, I think, where you can say...or an editor that you would hire or first meeting consultant that doesn't even exist, right, or a podcast executive producer, we all know the feeling of how long it takes to train somebody to be the kind of language you want and you get back these proposals from a company where you're just like, "That's not at all what I said," and you paid 20 grand for it, right, and so I'm just saying you're gonna be investing that time anyway. You could invest it in somebody who never gets sick, doesn't have fancy feelings and...

Michael: Remembers every project you've ever worked on for the rest of time.

Carl: Yeah, and I'm not, again, does it replace a marketing consulting firm? No way. I know some really good marketing people in our industry that are never going to be replaced. But for day-to-day stuff, yeah, I would just keep it open as a thinking partner.

Michael: Very cool.

Carl: Yeah. I'm looking forward to email on this. I'm super curious what people say but email us. Where do we email people us? hello@behaviorgap or where else?

Michael: [email protected].

Carl: Perfect. That's even better, [email protected]. Email Michael, tell him how much you... Yeah, anyway, we love feedback.

Michael: Also, tell us how you're using it. How are you using it that we didn't mention in your advisor world? Share with us.

Carl: Cheers, Michael, awesome.

Michael: Thank you very much, Carl. Appreciate it.

Carl: Yep.