Joel Bruckenstein: The disruptive state of AI and wealth management

Veteran wealth technology consultant Joel Bruckenstein says the financial advisory industry is undergoing its most dramatic technological shift in decades, as artificial intelligence tools move from curiosity to necessity at a pace that has surprised even tech observers like himself.

In a conversation with Action! magazine, Bruckenstein argues the technology promises to fundamentally alter the economics of wealth management by automating the majority of administrative work that consumes so much of an advisors’ normal day.

Yet Bruckenstein cautions realizing these benefits requires financial professionals to develop a new literacy around AI tools, learning not just how to use them but how to evaluate their reliability and explain their outputs to regulators. He also notes it will potentially eliminate entire job categories, and the speed of its evolution means today's cutting-edge applications can become obsolete within weeks.

The following is an edited transcript of the conversation.

Action!: Let’s start with AI. Is adoption getting ahead of best practice here?

Joel Bruckenstein: I wouldn't say that. I would put it a little bit differently. I think there's a lot of concern about regulatory oversight, general best practices, cybersecurity, all the usual concerns. Whenever there's something new in technology, advisors tend to be a little more conservative and maybe they don't educate themselves the way they should.

People don't realize for instance that using a Large Language Model (LLM) that’s free on the Internet is different from something that's custom-built for advisors with a lot of guardrails around it. We need to educate advisors about how to safely use AI and help them understand what the actual use cases are. Because advisors are like, ‘Okay, AI is great. What's it gonna do for me?’

There’s a lot of talk about AI and operational efficiency. That's certainly part of it, but it's not all of it. It's not going to replace advisors. What it's going to do is make advisors a lot more efficient, free up a lot of their time, so they can do less back-office tasks and grunt work and spend much more time in front of clients. The statistic that generally gets kicked around in the industry is that advisors spend no more than 40% of their time, and usually less, on client-facing tasks. We'd like to see that a lot higher.

Action!: When you think about the introduction of AI tools into the RIA space, the adjectives that people use are incredible, rapid, mushrooming … You’ve seen these types of tech trends before. What’s your critical take on it?

Bruckenstein: I think everybody was skeptical at the beginning, including me. I don't think anybody envisioned just how fast and how rapidly things would change. We were joking about it with a bunch of tech people earlier. Whatever we're talking about right now, probably tomorrow, someone's gonna be obsolete already. Literally, that's how fast it's changing. So it's very hard to predict the trajectory of it and where it's going to go, but we do know it's going to be transformative, and it's going to critically change the way advisors serve their clients. That we do know.

Action!: What's changing from the perspective of advisors then?

Bruckenstein: The fastest adoption is with AI notetakers, because it's low-hanging fruit. If you've done your due diligence and bought one that was made for this industry, it's already got some guardrails around it. They've gone through the compliance checks. It doesn't really interfere with anything, and because it's not, quote, unquote, giving advice, which is what the former United States Securities and Exchange Commission (SEC) was so anal about, it is considered safe. This SEC is a lot more libertarian. They are trying, from everything I hear, to foster innovation, and they understand that AI can deliver better results, potentially for clients, at a lower cost. So, as an advocate for advisors, as an advocate for their end users, I buy into that, and I think that's why it's going to be transformative.

Action!: Where are the gaps? In any kind of software development, there’s an evolutionary cycle. It's still early innings.

Bruckenstein: If there's a gap, it's a lack of understanding on part of some of the new entrants into the space about how regulated this industry is; about privacy concerns, security concerns, vetting concerns; all the things that go into building a new product. Because a lot of new companies, when they start out, they don't think so much about those things. They think more about their marketing budget than their cybersecurity budget and their compliance budget.

But overall, I'm really pleased with the way things are going, because AI is influencing so many things that advisors don't see behind the scenes. Every company in this space is now using AI in some way, shape, or form, in encoding and development. So that means better products, quicker to market.

You still need a human. It's not going to get you 100% of the way. But it can get you 50%, 60%, or even 80% of the way there, and then you bring in the human to do the more complex stuff that AI doesn't have a full understanding of yet, or the capabilities to do yet. Everybody benefits from that. [Generative AI platform] Claude came out with a financial services version. It didn't exist two months ago. And next week or next month, we'll be talking about things that didn't exist today, that will exist by time this article actually gets read. I can go to Claude right now and get a full, operational, fully vetted Monte Carlo simulation model. It used to take people months to build those things. Now I can just get it in an instant.

Action!: Do you expect RIAs to start vibe coding and building their own tools?

Bruckenstein: I've talked about this with a lot of people in the industry, and the way I think about it is, look, since I started in the business, some advisors have tried to build software for their own firm. And some of them tried to commercialize it. First of all, it was a big investment that maybe took years, and ... certainly less than 5% of those things that got built for an individual advisory firm ever really got traction. We always say advisors are like snowflakes. No two are the same. Or, if you met one advisor, you've met one advisor. So there's going to be a firm that has a need that they think one of their vendors should build, and it's just not economical for the vendor to do it. They'll do it themselves with this technology. And that's fantastic. And there'll be apps like that in a few years in every advisory firm. Some of them will be very complex, some of them will be very super simple. They’ll be everywhere.

But it's not going to change the fundamental need that if you want to maintain something at an enterprise scale, it will need constant reinvestment. There's constant security checks. People think you build it and it's done. That's just the first check you write. It's ongoing checks, forever. That said, do you want to take that on? It's very difficult to run a successful RIA firm and a successful technology firm at the same time. I know a few friends who have tried to do it, and they're not getting much sleep. Let's put it that way.

I should add, the trend is swinging back towards instead of trying to assemble a tech stack yourself, let the professionals handle it. I do think advisors want to simplify their tech stacks. They don't want 25 tech vendors. They want to rationalize what they're doing, minimize the amount of interactions they have. Integrations are still much better than they were, but not as good as they could be. And a lot of advisors still get caught in that web of when something's not working, one side says it's the other side's fault, and vice versa.

Action!: It’s been said that software will eat the world. What will AI eat in wealth management? It’s been suggested it will be CRM providers.

Bruckenstein: Somebody said that to me just the other day. Nope, wrong. The way I think about it, the definition of CRM, and the design of CRM is going to change. We may call it something else. It is going to have a lot of the same functionality, but it'll do it differently. For example, there are a lot of complicated front-end interfaces right now in CRM. That's slowly going to go away because you're just going to be able to tell AI what you want, and you won't need to interact with that interface. It'll still be doing the same stuff, but it will be doing it better, faster, more efficiently, and we may not call it CRM.

The same thing is happening in portfolio construction. You just tell the AI agent, I want you to build a portfolio, starting with the Russell 1000, that has a value tilt and quality tilt. I want it to be 75 stocks, blah, blah, blah, blah, blah. And in minutes, it'll do it. Is it replacing a functionality? No, it's using the same functionality. It's just making it easier to do. I wouldn't even call that putting it in a new wrapper. It's taking a new approach to it. It's taking the wrapper away and exposing the guts.

You don't need a risk tolerance questionnaire. Just talk. You can have a conversation that actually makes the risk tolerance part of onboarding a little more natural as opposed to how we do it now. So whatever your process is, if it gets you to the results you need in a more comfortable way for the advisor and the client, that's a win. We don't know what that's going to look like in a few years. It will be different than today, but it will still capture the data that's necessary to build an appropriate portfolio for the client.

Action!: There are concerns AI promotes some bad practices; for instance, you came to a meeting expecting a peer conversation but found only notetaker agents in attendance.

Bruckenstein: Sometimes, if you have a meeting with 10 people, five are only tangential to the conversation, and they're there for one specific thing. Why should I sit through an hour of a discussion when only five minutes of it is relevant to me? And maybe I'm not even going to speak. I need to hear what the other people say. I can just then record the whole thing and go to AI and say, ‘This is what's relevant to me, tell me what was relevant in that hour that I need to know?’ You just saved 55 minutes. Yeah, it's uncomfortable, because AI technology is going to disrupt us in ways that we haven't even imagined yet. But it doesn't need to be a bad thing. It's just different.

Action!: End clients will probably take the same approach. Talk to my AI agent.

Bruckenstein: And that's fine, if they like it, but I don't think that's coming that soon, because so much of what advisors do is behavioral. Every advisor will tell you that the clients want the client interaction. They want to be able to talk to their advisors. That's how you build trust. I think there may be some shorter meetings or things that can be pushed to AI. Someday, yeah. But not now.

Action!: You touched on portfolio construction. There was a recent discussion about how using AI in spreadsheets was causing hallucinations in some of the calculations, so maybe stay away from that right now.

Bruckenstein: You just made a comment that makes me angry. If you're talking to a large GPT, that's true. But again, there are dedicated apps out there today and all they do is Excel calculations, and some do it better than “The Best Excel Master in the Universe.” So again, it's about using the right tools. When we make generalities, that's when we fall into problems. And if we don't understand, if we're using the wrong tool for the job, that's when we have problems. When you make a blanket statement like that, I'm not criticizing you. That's what 99% of the people in this room do. But that's wrong. I was at a conference a couple of weeks ago, and an attorney asked a compliance question of ChatGPT, knowing he was going to get a hallucination by the way he asked the question. My answer was, ‘The problem's not with the GPT. It's with you.’ Because you use the wrong tool, and you ask the question the wrong way. You framed it in the wrong way.

Action!: Firms, though, are being asked to understand what’s driving their outputs. That requires a level of technical expertise. A GAI tool is not a word processor.

Bruckenstein: Do I need to understand entirely how a word processor works? No. But I generally have a good understanding of how to use it. There is a level of knowledge you need to be able to explain to a regulator, what is this portfolio management tool that you've created? How did it come to these calculations? I know the answer to that question. Again, it comes back to what tool are we talking about? If you're talking about a free LLM, you are absolutely correct. I would not recommend that any advisor use those models for anything business-related, maybe for things like brainstorming, getting directions or asking it to plan a vacation.

I would recommend using a deterministic model-based LLM. It's based on data from a provider. They know where every data point comes from. It doesn't go out to the Internet and ask questions. Every answer comes from their data that is fully vetted. If you use Perplexity’s paid model, it will give you citations about everything. You can look at those citations and vet them: This came from the Federal Reserve Bank of New York. Probably good data. This came from XYZ capital, I’ve never heard of them. Probably not good data.

So we need to educate advisors to understand how to vet the models, because it's not rocket science. The ex-SEC Commissioner made it sound like it's a black box. LLMs to some extent, are. Most of them are not black boxes. You can say, Okay, that's a great answer. Where did you get this data from?

If I want to know something about compliance, and I want to use Chat GPT or Gemini, I'm not just asking them a question. I will ask, ‘Could you please go to the SEC website and find me some citations about this? I need to know more about this particular topic.’ It's not going to hallucinate, because I'm telling it where to go to get the answer. And if it can't get the answer there, it'll come back and say, either it doesn't exist there, or I couldn't find it there, or something like that. It's a totally different way of prompting. So if you know what you do, there’s very little chance of a hallucination.

Action!: Is this why T3 launched the AI University?

Bruckenstein: If an advisor is educated about AI and asks the right questions, they’ll be a great consumer of this product. And if they ask all the wrong questions, you probably don't even want them as a client until they learn. The latest Schwab survey said over half of advisors are already using AI and another 20 percent plan to use it over the next three years. This is something you need to do hands on. Use it with caution at the beginning, until you feel you understand it. Take some courses online, talk to people, colleagues who know what they're doing. But, yes, you need to learn it, and you need to use it, and you can't wait.

Action!: Are advisors expected to become programmers?

Bruckenstein: You learned how to use Microsoft Word and Excel. You learned how to use the Internet. This is no different. I have a friend who's 75 years old. He started using AI for creative ideas. Now he is holding hour-long conversations every day with Chat GPT, and he finds it very useful for what he's doing. At the beginning, he didn't know what he was doing. But he has an inquisitive mind. He said, ‘This is where the world's going. I gotta get on top of this.’

Action!: There are plenty of firms that rely on one person, a superuser, to understand all of the tech they use, and that’s the extent of their engagement with it.

Bruckenstein: AI is going to democratize technology. You won't need to go to that guy to ask him to run a custom report. AI will do it for you. That model is outdated already. A lot of people just don't know it yet. Some firms don't need a receptionist anymore. Some may not need as many administrative assistants. Same for people who reconcile within firms. If there's a person doing it, it'll take a fraction of the time it takes now. If you have three CFAs on staff, you'll probably need one. All of that manual work on the tax side, particularly the tax reporting side, is slowly going away. All the work is not going to 100% go away, but 80% of it will go away, and what's left will be higher value stuff that humans are still better at.

Action!: Where does that leave young people trying to get into the industry

Bruckenstein: The way an advisory firm functions is rapidly changing, the kind of skills they need are rapidly changing. Universities are a little behind the curve in supplying the skill set that today's graduates need to succeed in advisory firms. Having said that, there's still an advisor shortage in the industry. So there's room for new advisors. I think every advisory firm of scale, the regional and the national ones, they're all going to have data scientists. A lot of them are starting to hire PhDs in AI. Because to your point, maybe they want to build some of their own stuff. But they just want to strategize about how they take advantage of AI? What's the low-hanging fruit? And it may be different for different types of firms. So are there still jobs available in this industry? Hell, yeah. Are we hiring more people to reconcile portfolios? No. So if that's what you studied in school, you got a problem. But if you happen to be fluent in AI and data science, I can think of 100 firms that will hire you tomorrow.