Is the age of "augmented intelligence" and "adaptive engagement" soon to arrive for financial advisors and wealth managers because of agentic AI? We're about to find out. Hyper-personalization of advisory services to deepen client engagement and build trust is possible because agentic AI can digest and synthesize data from multiple sources, allowing advisors and wealth managers to continuously tailor financial plans and offerings that are unique to each client, based on myriad factors and behaviors. This customized approach to client relationships means that, for example, investment allocations can be adjusted in real-time based on risk appetite shifts, tax-loss harvesting opportunities can be flagged automatically, and relevant financial information and educational content can be provided to clients based on their lifestyle and other behaviors.
While financial advisors and wealth managers are just starting to explore and test use cases with agentic AI, the promise and potential loom large. Early use case scenarios focus on autonomous portfolio rebalancing aligned with evolving client goals; real-time "what-if" scenario simulation for financial planning; AI agent-suggested actionable steps based on savings and retirement goals progress; and AI-enabled onboarding and continuous suitability assessments.
Is agentic AI necessary for all financial advisors and wealth managers? In a word, yes. It won't replace trusted "human" advisors, but AI agents have the power to arm them with what they need to deliver smarter strategic investment advice and deepen their engagement with clients over the long term, while also reducing time spent on tasks.
The great promise of agentic AI aside, financial advisors and wealth managers will still face ethical, regulatory and trust considerations with the technology, including the transparency of agent decision-making, data privacy and security, and regulatory frameworks and explainability requirements. Our panel of industry professionals digs into agentic AI's great potential, the use cases that are easily within reach, the impact on client retention and satisfaction, and how financial advisors and wealth managers can reap the best of the technology—the scalable delivery of holistic advice without proportional increases in cost—with little downside.
Transcription:
Brian Wallheimer (00:14):
Good morning everyone, and welcome to our seminar on Agentic AI. I'm Brian Wallheimer, Editor-in-Chief of Financial Planning. AI is obviously a huge part of the industry today. ChatGPT feels like it was just yesterday to many of us; it exploded onto the scene and changed everything about the way the whole world works. Now, certain pieces of AI are table stakes in the industry, and advances are coming rapidly. So thank you for joining us. Thank you for keeping up. I want to ask you to stick around for all three sessions today and remind you that this is a wonderful half-day seminar. But we do have Advise AI coming up in October in Las Vegas. That is wealth, excuse me, Financial Planning, Wealth Management, AI Day conference. You can go to financialplanning.com, click on events, and find much more information there. Today, we're going to learn quite a bit about Agentic AI, and this is quite a bit different from the generative AI that we're all used to.
(01:20):
The things that you'll hear about today are going to lead to very rapid changes in the industry. It's a great thing for you to be here and learning to keep up because there will be those who adopt this quickly and move forward rapidly and scale and grow. And there will be those who take some time, and they might be a little bit left behind. So thank you for being here and thank you for being willing to learn. I'm going to start off today with our first session: a panel discussion on "Personalization is Everything: Wealth Agentic AI Use Cases to Deepen Client Engagement." We have a great panel today. Joining me is Erik Allison, Managing Partner of Wealth Vision; David Breakstone, Managing Partner at Salesforce; Geoff Moore is CIO of ValMark Financial Group; Stephan de Man is Vice President at Dimensional Fund Advisors; and Rob Pettman is President at Tifin. Welcome everyone. Yeah, so we're going to start off with defining Agentic AI. This is obviously a little bit different than I think most people are used to, as I said before, but advisors are used to generative AI for things like summaries, emails, things like that. I want to see if we can start off with what Agentic AI is, how it's different, and we'll go from there.
David Breakstone (02:35):
Great. I'll take a first crack at this. When you think about AI, we think about it in three ways. There's predictive AI, which is tell me what I should do next. There's generative, as Brian said, help me summarize a bunch of information. And then there's Agentic AI, which is really marking a new era in terms of advisor productivity, efficiency, and growth. Agentic AI can manage and execute end-to-end processes with little to no human interaction. It can dynamically reason, breaking down large projects into smaller tasks. It's conversational, so you could ask it whatever you want, and it'll use natural language processing to understand what you're asking for. It's multimodal. You can access it through the internet, through digital, through Slack, and it has very predefined guardrails. So you can tell the agent, don't deliver financial advice, don't give tax advice, don't change power of attorney, kick out to a human, an advisor to manage those processes.
(03:43):
But I think what's most important is that agents can do the jobs that we don't want to do, and they could deliver instantaneous execution on these tasks. I think when you're saying, Brian, we think about it, what we've seen so far is agents focus first, deployed in servicing and operations. You're now seeing it a lot in business development and sales support, where scheduling, managing leads and referrals, helping and augment the advisor. I think what you're starting to see some proof of concepts, which I think the rest of the panel could talk about, is seeing some advice and portfolio creation to help the advisors really expedite that process. We're still in the very early innings, and there's a lot of innovation ahead of us.
Brian Wallheimer (04:34):
Great. So let's talk a little bit about the problems the wealth management industry faces and where Agentic AI can play a role in creating solutions.
Stephen de Man (04:44):
Yeah, I'm happy to jump in on this one. We run one of the larger advisor studies out there, and we always ask, what are your top challenges when it comes to operational challenges or growth challenges? Year after year, the number one challenge operationally is implementing a workflow process and maintaining that workflow. I think a lot of agents, as David mentioned, can start to help with those workflows. There's a cautionary tale on that: it's not a total green light without those guardrails in place, but helping advisors address those pain points I think is a key aspect of where we're headed.
Rob Pettman (05:23):
It's funny you mentioned workflows, because we've been talking about workflows for so long. I remember we were talking about tech stacks, and then we had APIs, and then we were talking about workflows, but it was like how data moves from one piece of technology to another, but it wasn't really a workflow; it was just kind of passing data back and forth. I think Agentic AI is really that sort of next avenue of opportunity that opens up workflows in a realistic way for people and businesses.
Erik Allison (05:50):
Yeah, I think something that's really interesting about Agentic AI, too, is in addition to it being able to think and solve problems, it's also got a vast opportunity to utilize tools. You can define certain tools that it can use. You can say you can't use these tools. Then there are things like MCP servers now. We are talking about API integrations, and basically, every time you want to connect to a certain tool, you have to enter in the API code. Well, MCP servers now are giving these Agentic AI agents the ability to essentially build their own set of tools and then be able to go in there and choose which one do I want to use to solve problems. So when you think about the power of that, whether you're back-office support or you're front-of-the-house or you're an advisor, you can go, "I want to accomplish these tasks," and I'm not really going to define how that gets done, but "Hey AI, here's a set of tools. Here's access to all the things that I want you to be able to use that I would be able to use or an employee, and go figure out this solution and just let me know when it's done." That's a pretty interesting concept of where we are today. What you're saying Rob about workflows is basically do this, then that, then this. Now it's, "Hey, I want to get to this solution. Here's all the tools you can use. Go figure it out." And that's the power of it right now, and it's really cool.
Geoff Moore (07:13):
I think the only thing I'd add is this idea of efficiency versus opportunity. A lot of the workflows we've described and our first pass at some of this is what am I doing today that I want to do quicker? And that would be efficiency. But what are the things that I've never even done that I can now do? That's possible because the cost of doing it has just dropped so dramatically. So I think that's maybe a new way we need to start thinking. Traditional problems we've had in our industry that we maybe haven't been able to solve, maybe some of those can be solved now that we have these new capabilities.
David Breakstone (07:50):
Brian, I think I agree. Workflows have always been mentioned, and we're constantly addressing workflows in our organization for years, and now you've got a new tool with agents that can help solve it. But even to go back to your question, what are the big problems in the industry where Agentic AI could help? I look at it as we have two major gaps. One, we have the financial advice gap in the industry where the World Economic Forum said about 90% of Americans are not getting professional financial advice. We've got an aging advisor population, 55 plus, and the good news is that I think the financial advisor role is going to grow like 12.5% according to the DOL the next 10 years versus 2% for others. The other gap, which I think is the symptom of workflows, is productivity and efficiency. There are just not enough advisors to do the job today, and they've been dealt a hand of disconnected workflows, proliferation of data, and unconnected systems, which is the constant problem that we've been struggling for 20 years or so to fix. Now we've got this new tool that may be able to really truly address these challenges and free up the time for the advisor to focus on more strategic engagements with their clients.
Erik Allison (09:17):
Sure, yeah. Think about it from the advisor perspective. You meet a new prospect, and there's data gathering, information sourcing, and understanding what's important to them, and then where does that information go? Then what do you use it for next? So from the advisor perspective, you've got, like you said, a bifurcated tech stack in many cases that's starting to change now, which is great, but a lot of firms have 10 or 11 or 12 different tools. You've got a financial planning tool, you've got Ycharts or Morningstar for analytics. You've got your CRM, you've got wealth.com or Vanilla or Luminary for estate planning, and then you've got things like tax analysis and tax snapshots. From an advisor perspective or from a CSA perspective, you've got information on a prospect or client. Maybe it lives in your CRM, or the systems are talking to each other.
(10:08):
So then you have to go and enter that again into your next platform, and then you have to enter it into your next platform. Instead of spending time getting to know your clients better, developing new strategies, or really diving deep into the holistic wealth plans that you're putting together for them, advisors are spending time transferring data from one place or another. Agentic AI is really going to help with that. We've seen it in our practice, just literally hours a day saved in terms of not having to change data back and forth. Then in terms of research, too, you can give it a task and go, "I really want to know about this and how does that interplay with this specific personalized scenario for a business owner client?" Instead of spending five hours doing multiple deep research reports, you can have the agent go out and accomplish those goals for you and bring back a customized information resource or report that now you can apply personally to your client or their specific scenario that maybe hasn't really occurred before. So we're starting to be able to save a lot of time and be able to solve problems in unique ways that in the past would take a long time or maybe just weren't sure how to do.
Brian Wallheimer (11:20):
Let's talk about the hype versus the reality of this, right? All of the things that you're talking about, you're talking about synthesizing data in real time from all over the place, creating personalized, hyper-personalized advisory services, and all of these sorts of things. What can we expect in the near term? How likely is this to get put into use and put into practice by advisors in a month, six months, a year?
Rob Pettman (11:48):
Yeah, maybe. Let me take that one. We're an AI firm that specializes in wealth management and in this problem. All of our solutions are built around the whole notion of there being some sort of commercial application and there being real value. So let me give you three different examples of things that we have in production that are probably just good for conversation. First one, as we think about the OCIO model, just think about how an OCIO would expand their coverage model to cover more advisors. Right now, they're constrained to maybe one OCIO per 100 advisors. We've got solutions now where we're expanding it from 100 to 400. Essentially, what we're doing with Agentic AI is gathering all the personalized requirements of the advisor, synthesizing that across the model marketplace of the OCIO, generating the proposal, and then putting the OCIO in a position to go about review, making tweaks, and then send off to the financial advisor.
(12:46):
Beyond that, there's a conversational interface to go about asking questions about any sorts of changes that may have happened in the portfolio, thus expanding the service dynamic for the OCAO to cover more advisors. So think of that as one use case for wealth firms or asset managers. Another one in the advisor case, very similar: advisors are out there working with clients, gathering information about their risk parameters, what have you. Now, think about a conversational interface that takes all of this into consideration and helps the advisor create proposals. Again, looking at how they manage money across the backdrop of models that they've created or used today, and then synthesizing that and putting it together in a proposal for the client, and then also offering very different personalization characteristics to make tweaks from there. So that'll be a second one in a more advisor-specific use case.
(13:34):
The last one I'll probably move to a different model will be: you think about the RIA aggregators. They are bringing on a host of new advisors under their umbrella, and typically speaking, what we've seen in the market is that it takes them about 10 to 12 weeks to transition a new advisor into their practice. We're using Agentic solutions right now to move that from 10 to 12 weeks to two to four weeks. The other part of the problem is before they could only do one at a time. Now, they're actually able to do this in parallel. If you think about the savings, both in the experience for the advisor, the time to moving them over, and time to billing, and then being able to add multiple at the same time, these are all material commercial outcomes that have real value where Agentic AI is playing a real role today.
Stephen de Man (14:22):
Yeah, I'll be not a contrarian point of view, because I think everything that Rob mentioned is fantastic, so I don't disagree with that. But I think when advisors that are maybe tuning in or thinking, or they're hearing Agentic AI elsewhere, I say don't buy the hype just yet in terms of all that you may be hearing when it comes to Agentic AI. What I mean by that is Roy Amara, a futurist, has this Amara's Law, and it says in the short term, we tend to overestimate technological impact. I think that's what's happening here specifically with Agentic AI. We're tending to, "Oh my god, it's going to solve all these problems, and we're going to have these agents out there fixing everything, and I'll just be able to spend all this time with end clients." So I say, don't buy that hype. When we look at the data around what advisors spend on technology, the typical spend from revenues is around 3%. So for a one to three million in revenue firm, we're talking $54,000 on average. So we're not going to be totally revolutionizing this overnight. I saw Geoff nodding his head, and he's a practitioner, so I want to hear from
Geoff Moore (15:35):
I was going to say we've heard a lot of stuff. I love David's opening remarks around Agentic AI and how awesome that was. But the devil is in the details with a lot of these and with different vendors. Everybody wants to use this label "agent," or everybody's system has an agent in it, but what that agent can actually do in each of these systems, there's a high degree of variability in terms of what that actually means when you start getting into the individual clicking details. Just like a really simple example, Eric was talking about MCP servers. I hope we get to talk a little bit more about that later because I think that's an awesome new technology. But Microsoft, a leading huge tech provider, they've had an agent out for a long time. It wasn't until March of this year that they allowed you to add tools, things like the MCP server that Eric was talking about, into their agents. So they're getting better all the time, much better, but the devil's in the details in terms of what they can do. So as you're thinking about it, just have that in mind that not all agents are equal.
David Breakstone (16:38):
Let's add to that, Jeff, and that's a good point. As we stated earlier, we're in the very early innings of Agentic AI. I would say I led an industry forum with 50 or 60 executives a few months ago, and what they were prioritizing were workflows and activating unstructured data through a lot of generative capabilities and book of business. So I think there's still, that's where the puck is right now, and you're starting to see a lot of those use cases being rolled out. The stuff that Rob mentioned is very exciting, but one of the things that we've also launched is it doesn't have to be just agents or just predictive, but pulling them all together into a solution I think is really interesting. I'll just give you a solution that we've demoed recently: an adjunct book of business, which would combine that predictive and generative to be able to interact with your book and say, "What should I focus on today?"
(17:38):
It's pulling all of your predictive leads and your best value clients, but from there, which is predictive and generative, you can then use agents to schedule those meetings and for other pre-meeting follow-up. Those are things that are live today that companies are starting to leverage and use. As I said, I think service operations and business development are the initial use cases, but I think it's coming. Agents are coming, and they're only going to get better. This is the worst AI we've ever had. It's only going to get better and more powerful and taking on more of those roles as we get more comfortable with the accuracy and the success and the price points.
Brian Wallheimer (18:21):
I love the comment about the worst AI we've ever had. I was thinking this morning about all this and I was thinking of text messaging and having to type the "two" button three times to get a "C" and then having to call you and say, "It cost me 10 cents, stop it." So this will only get better. If you think back to when we were doing that, you can't even imagine what we'd have today just in terms of messaging each other. So I guess really what we need to say here is the proof is going to be in the pudding. So let's talk about some use cases and some ROI. What can people expect out of this? If I'm an advisor who decides to go ahead and give this a try, what can I actually expect? So can somebody walk me through a use case? Maybe we can get a few of these out there and then just talk about the benefits that we might expect now and what might be coming down the pike.
Erik Allison (19:14):
Yeah, I think from an advisor perspective, you think about the life cycle of a prospect to a client and the meeting process. What does that look like? So you prepare on the front end, you meet someone, you do research on someone to learn about them and make sure that you're bringing up topics that are important that might apply to them. So from an Agentic AI perspective, here's an example of how we've utilized it. You meet someone, or someone comes into your CRM through a form online or some kind of trigger. When that contact information hits the CRM, immediately we have an Agentic AI that can go out and do prospect intelligence research. So it essentially searches public domains such as Google and LinkedIn and just says, "Hey, I want to find out about this person." So we tell the AI, "You figure out how to get the information, but here are the things I want to know so I can better serve them."
(20:03):
That happens automatically. The use case and ROI: that can take 30 minutes. A lot of you probably prepared for meetings and trying to figure out, "Hey, what position is this person in? What's really important to them?" That can take up to 20, 30 minutes easy, sometimes even more if you're looking into multiple company structures. Then from there, the Agentic AI produces that, and then it can produce an agenda. It already prepares that for me, and that can also take another 15, 20, 30 minutes. So there's an hour saved already. Throughout the meeting, it's taking notes, and that's more of the static side of things, but then the agent kicks back in. At the end of the meeting, it knows what was spoken about, it knows what we were talking about, and then it organizes a follow-up email that goes right out. So it puts it in an inbox.
(20:48):
I can review it to make sure it's all the things that we wanted to talk about and make sure it's correct, and then that gets shot out automatically. Then from there, the follow-up goes, "Here are the things we spoke about. Here are the promises I said I would deliver," and then it tracks and monitors that communication and then it can ping me through Slack or via email, "Hey, this person you met with is responding now." So you can already start to see, from the process of things that happen a lot and happen pretty often, you start shaving hours off of multiple different routines that you go through the day. Talk about ROI: I just liken it to an attorney that's charging $500 an hour, for example. You can quickly start to see that it can add up to a couple thousand dollars a day in saved time.
(21:36):
A lot of these systems right now you can subscribe to on an individual user seat basis: 20, 30, 50 bucks. You start getting into the enterprise side, and that's a different story, but then that's at scale. So you start going, "Wow, my advisors are saving $2,000 a day, and maybe the cost per seat is a hundred bucks a month or 200 bucks a month." It starts to frame how powerful this can actually get from a cost perspective and a revenue ROI perspective, too. So exciting things are happening there, and I think it's only going to get better from here, which is nice.
Rob Pettman (22:12):
Yeah, I'll give you a couple. I mean, I cited a few already: just being able to accelerate the transition time of financial advisors. We have something similar, by the way, for client meetings: the ability to straddle financial planning systems, CRM, notetaking, portfolio management systems, coalesce that into a report in preparation for a meeting where our clients tell us, on average, we're saving them about two to three hours per client meeting in preparation for that, which then goes back into the business, albeit through new business generation or client service. We have a private markets business, so for firms that are looking at alternative investments, we're shrinking down the time for due diligence from eight hours to about 30 minutes just in distilling documents, performing the personalized diligence reports based upon how our company may put that together, and then ultimately helping to socialize that with product diligence committees. So I'm giving you a flavor of different use cases here that might be inside of a wealth enterprise or might be inside of an advisor's office, but there's a host of different commercial outcomes right now with this type of technology.
Erik Allison (23:27):
Rob, is it easy? You mentioned that you're using all these different systems. I know from our perspective, sometimes it's hard. What are you seeing in terms of the companies playing nice with each other, right? Are they charging each other big fees for API integrations? Are they just going, "Here's the data, use it how you want"? How are you seeing that since you're building things on the enterprise level for organizations? What's that like, and is it going to get easier, or what are you seeing there?
Rob Pettman (23:54):
Well, I mean, look, in the early innings, nothing's easy, and you have to get creative on how you go about solving this. I think, first off, just the unification of data as a place that we start, and that's an area where we specialize, just out of necessity of having to be in business. So us being able to help unify data for enterprises and then being able to package these solutions on top of an infrastructure we've created has helped to materially solve the problem and accelerate the pathway to adoption because most firms we talk about, they start talking about their data problem and the large data thing that they have to go through in order to get to AI, and we help accelerate that component. So I think that's probably part one. I think as this technology evolves, too, just even the API infrastructure gets kind of interesting, especially with computer use models and how, just through a conversational interface, computer use models can go through a login and perform functions and pull out and exercise tasks and things of that nature. I think that actually opens up the speed to market for a host of these different types of solutions. So there are probably evolving ways through which these pieces of technology can be accessed and unified to a single point of entry that haven't existed before, and I think that's one of the things that's worked for us in being able to construct these solutions. Yeah.
David Breakstone (25:15):
I'd say that to that, Rob: good AI is grounded on good data, and a lot of our clients are investing in their backend to help expedite and accelerate their AI strategy. I think the meeting prep one is a key use case across the industry. Taking something that takes hours or days or weeks for an ultra-net-worth client and getting that down into a few seconds or an hour I think is a critical one. We've launched: as Salesforce, we're customer zero; our help desk online is now an agent, so you're interacting with an agent if you're accessing help. I advise everyone, take a look at that. I think like what you're saying, accelerating the time to productivity for a new hire. We launched a sales coach agent also, where you do your sales pitch directly to an agent, and then it grades you on how well you're doing your sales presentation.
(26:13):
I think it takes time away from—frees up time for both the manager and the advisor. The Agentic book of businesses is out there. I think last, which is probably also in your area, Rob, is in the asset management area, helping to target scheduling of wholesalers to advisors is a big problem. Proliferation of data, disintermediation from RIAs, and how to be able to track all this movement and figure out what's the best coverage model. I think that's where an agent can really accelerate and optimize coverage models as well. I think you're seeing that in the industry as well.
Stephen de Man (26:56):
I'll share, I think Eric hit the nail on the head with the integration piece. I was meeting with a firm last week in Austin, and they were telling me, "We use Vaga Minds for our emails, and so we get emails, and that then, but how do we get it to talk to this system?" And then having each system not play nice with each other is a big pain point. So as we start to move into this Agentic world, where hopefully that frees up a lot of the systems being able to talk to each other, I think
(27:26):
that's really great there. I think everyone needs to respond to that and open up their data. I think that's been more democratic, and having a very robust partner network where it doesn't always have to be you. It could be a third-party partner that builds on your platform or just interacts through an API into your platform. I think it's critical. What I'm most concerned about is that we've been spending the last 10, 15 years connecting a platform together for the advisor, and I'm worried about now disconnecting it with multiple agents for different purposes all over the place versus a few select groups that can leverage that. I wonder whatever people think.
Geoff Moore (28:08):
I think the future is more distributed. I don't think it's as centralized. I think there's an interesting story if you look at Walmart and how they've adopted some of their agents. Essentially, instead of having a bunch of different agents, they now use this centralized, almost, I'd call it AI workstation.
(28:25):
That AI workstation then can connect out to all these other distance systems through the MCP servers or specific API connections like Eric was talking about. I think that could be kind of more the future. But to your point, David, there are still instances where some of those connections between specific vendors are not in place for strategic reasons.
Erik Allison (28:44):
Or there are paywalls behind them, right? We were just talking to one of the platform providers that we utilize. They wanted $50,000 just for us to be able to send information back and forth for an API integration. I think what also is happening with the Agentic AI and AI in general, which is nice, is it's kind of forcing companies to play a little bit nicer. It's going, "Hey, everybody can just have access to information, and it should flow from platform to platform, and we should be able to send that from one place to the other without super technical integrations and super high costs." I think one of the nice things for the industry in general is that that's going to break down barriers, right? It's going to break down walls, and I think maybe I'm wrong, but maybe they're just going to start charging more. They know everybody needs the information, but I think you're going to see people playing nice, and it's nice. The data needs to be going from one place to the other because not everybody's got a solution in one package.
Brian Wallheimer (29:41):
We've got a couple of comments and questions, and I think they fit right here, so I want to throw them in if we've got a second. One of them is: "I think Agentic AI will require human presence, at least for calibration and validation. If humans are checked by various quality checkers, why would it be different with AI? I think it needs even more supervision."
Erik Allison (30:01):
100%. 100%. Yeah.
Brian Wallheimer (30:04):
Just thumbs up.
Erik Allison (30:05):
Yeah, I'm just going to let it run wild.
Brian Wallheimer (30:07):
Okay.
Erik Allison (30:08):
I love the comment. If we have to monitor and then have rules in place for humans, why should AI be different? They shouldn't be different. Because it also is new, and because it's essentially, I don't want to say software, but it's like it's smart intelligence in systems, it needs even more supervision. I think that's one of the responsibilities that we have, especially in the industry where we're giving people advice, and they stand behind that advice. The things that we help people with are some things that they rely on for their well-being or their livelihood. So it's a duty that we have, not just on the fiduciary side, but it's our responsibility to make sure that we're using tools and then making sure that the output is accurate so that if we're going to use that output and then give it to someone else on an advice basis or from an insight perspective, we have to make sure that those things are right. So I think that's cool about the Agentic AI, too, is when you're writing instructions—and that's what a lot of people don't talk about, maybe a lot of advisors or people out there listening maybe don't have experience with—but that's one of the challenges, too, is when you get into building these things, you realize, "Wow, I could build a five-page instruction sheet on how this agent can act. I can tell it all the tools and what it can and can't do." But there's a big part of that.
Geoff Moore (31:29):
Yeah, there's a whole new phrase that's come out. Instead of prompt engineering, now we're context engineering. We're giving it the right pieces of information and knowledge to help do that. Yeah.
Erik Allison (31:38):
But yeah, we need to monitor it for sure. Go ahead.
Geoff Moore (31:43):
Two quick examples. This kind of goes into a little bit of some of this vibe coding stuff where we're putting this technology in the hands of people that have never dealt with some of this stuff, and so people are building their own applications. Two recent stories: a person was using Replit, and then the AI just deleted their database out of nowhere.
David Breakstone (32:03):
There's another app,
Geoff Moore (32:05):
Tapp that was vibe coded. This is real-world harm. It exposed 72,000 users, including their IDs, selfies, and private messages. So a lot of this stuff has become easier for more people to do, but we still have to be careful.
Erik Allison (32:23):
And you're putting a new tool in the hands of people who aren't trained on it, they're not experienced in it, they don't know how to protect against things like that happening. So that's an interesting thing that we need to be careful about, too.
Rob Pettman (32:37):
In the real-world application, again, just highlighting the human component in some of the examples I referenced. So, for instance, helping to transition advisors from one place to another. The Agentic solutions are essentially highlighting to an operations team where it is that they need to focus on, which accounts are going to be problematic from a 'not in good order' perspective, and they're passing through the ones which essentially cross the criteria of all the rules. Similarly, in portfolio construction, I mentioned that this sort of common, I think we think about this in trading. I'm a trading nerd by nature, so review and release. The whole concept is actually a trading concept, but it sort of applies here even in the AI context. Eventually, you get your portfolio, and even an OCIO or advisor reviews, makes any adjustments, and then releases it to an end investor, and I think it's a critical pathway as part of the workflow.
David Breakstone (33:31):
I mean, add to that, Rob. I mean, I think in terms as we get more into the advice area, everything is going to be advisor in the loop, not autonomous, at least to start, until we get some more accuracy and predictability and confidence. We'll talk about change management. That's going to be a critical issue. But I think somebody said to me one time, one of our clients, we were talking about AI, and they was like, "How do you talk to your compliance team?" I'll throw it out to the group because I think there are varying opinions on this. Is this a new risk or the same risk just coming in a different way in terms of advice? That's a thought-provoking statement there of what we're seeing. We always have to do regulations around delivering advice in a heavily regulated industry, but I may throw it out to the group here what they think on that statement.
Erik Allison (34:27):
Well, I think that's an interesting topic. I think it's the same risk just delivered in a different way. Humans can give wrong advice. You can give the wrong trust structure that actually doesn't accomplish the tax minimization strategy that a client might have while also accomplishing charitable giving or philanthropy needs or transitioning wealth from one generation to another. A human can just as easily give the wrong advice, and then come tax time, the CPA goes, "Who told you to do this? This is not the tax move that you needed to do with your trust structure," or things like, "Hey, you shouldn't have gifted this property to your kids before your parent died or something because now they don't get a step-up in basis and they owe a lot more money." So things like that, that inexperienced advisors or advisors who are delivering advice without oversight of managing partners or directors who are running a team of agents, the AI can do the same thing. So same risks, same problem, just delivered in a different way. So it's like now there are two, right? You have to make sure you're checking on the AI advice, and then the advisor has to consume that information and then make sure that they're understanding it right and utilizing it and what does it mean, and then delivering the advice again. So
Geoff Moore (35:42):
I think there is some data retention risk. I think we've, in some areas, we have to keep a lot of records for compliance purposes, but there's been some other areas where we've specifically deleted certain records, and AI does better when it has more data and more records. So I think that's been at least one of the things we've struggled with is what is that right level of data retention to keep to make the AI work really well, especially in areas where we necessarily haven't historically kept those types of records.
Rob Pettman (36:14):
I think there are two sides to the coin, and I'm going to be the optimist here. But look, I think that there are real risks that need to be managed, and to advance the business, you have to wrap your arms around and manage those risks if you want to just be better and create better wealth outcomes for end investors and really harness the scale problem we have in wealth management today.
Brian Wallheimer (36:34):
Right.
Rob Pettman (36:35):
On the flip side of the coin, it would be a shame not to recognize the risk management potential AI actually has and the value it can bring from that perspective. So for instance, being able to make sure that the right documentation is in place for particular advisory accounts, articulating the investment strategy, trading rationale, and things of that nature, being able to get your arms around a rep's book of business in a more organized fashion.
Geoff Moore (37:06):
Even if you do have a human reviewing, having AI be that extra set of review that wasn't possible before.
Erik Allison (37:12):
Or
Geoff Moore (37:12):
be able to cover more cases than you could because now it's AI and not just a human.
Erik Allison (37:17):
I like how you flip that, Rob. You go from, "There's more risk," to, "Hey, it also can take risk off the table," and it is totally true. You can look at it both ways, but there have been so many instances where I've been able to be more accurate and quicker and come up with better solutions because initially I missed something, and the AI goes, "Hey, you didn't consider this solution." I went, "Oh, man." So that's a cool way to look at it, too.
Stephen de Man (37:42):
On that topic of usage, I'm curious, Jeff, you might have insight on this, having a usage policy either for internal use or for advisor external use. When we collect the data on this, we see of the firms that are using AI, only 40% have some type of written verbiage or usage policy, so I'm curious, how do you all approach it? Yeah.
Geoff Moore (38:05):
I mean, we just recently updated our AI policy and put a little bit more teeth in it. I think one of the things that we had not maybe defined as well in the past is what we do with, I think everybody gets PII, right? Everybody gets PII and PHI. But proprietary data is something like we've recently defined, and being really specific about certain systems. There's a lot of information in your firms that you might just think is not a big deal. It's not PII, it's not PHI. You're putting it into these systems. Depending on what system it is, did you just leak your advisor list or your client list or something like unique code or some unique process or operational thing that you have that's helpful to your firm that you didn't even think of it being as maybe sensitive data, but it actually was? And depending on what AI system you're putting it in, you might need some more caution around that.
David Breakstone (38:56):
That's guardrails, and that was one of our first endeavors and how we structured this: the guardrails, but also the zero copy, right? You're not sending this PII information out to the public, where that could be tagged. Just to close, I like what Rob said on the compliance piece. If you think about fraud, fraudsters are going to be using agents and have started to use agents. So there's a whole opportunity within our back offices to leverage AI and agents for fraud detection and other regulatory compliance issues that could potentially come up, complaints, or any other things that they could do, and it's such an accelerated pace. So that's another avenue where you're going to start seeing, while this conversation is talking a lot about agents supporting the advisors, this will be agents supporting mid- and back-office compliance teams, legal teams, RFP management, everything that's going to just transform our overall industry, which is very exciting.
Brian Wallheimer (40:01):
Great. I want to go to one other quick question here. Steven, in earlier comments, you mentioned a study divided between operation and growth, and you gave the operations side of this. Can you talk about the growth side of it?
Stephen de Man (40:14):
Yeah. The number one challenge for firms, both our high-performing firms and everyone else, is capacity constraints. So that dovetails right in line with what we're addressing here. But advisors are stretched thin. Capacity constraints for their teams, growing their teams. I mean, we were talking in the pre-green room just around skill development, too. The number of conversations I have with advisors that say, "Oh, we've got great farmers, but we need more hunters." Developing that skill is something that people don't want to pick up the phone. Younger folks don't want to even call a prospect. So I think a question for us and our discussion is, how can agents either help train, develop—David mentioned using that within Salesforce there—but I think that's something that we see come up quite a bit on the growth side, those capacity constraints.
Rob Pettman (41:12):
If I could add on the growth side, I think growth is a harder area for adoption for a lot of enterprises because it takes a big leap of faith and investment in risk because you're essentially making this growth bet that growth is going to happen alongside of it. Some firms, depending upon where they are, don't have a risk appetite for those sorts of things, but some do. But there are applications that exist out there today. For example, we have one focused on growth for wealth firms that essentially puts its arms around a significant amount of data and helps advisors understand out of their client base where they have the greatest potential to expand their relationship and gain a higher share of wallet. It's again, learning on a predictive nature and helping them understand with a level of explainability as to why they should call that client and have a different level of conversation. We've got a client in the market right now who's generated over $2 billion in net new assets using something like that. So it is possible, but people are seeking commercial proof points first, I think, a lot of times before they make that leap.
Geoff Moore (42:24):
Rob, is that client you referenced, though? I feel like there are some of these digital tools, but I feel like they had to do something alongside with your technology. I mean, is that a fair statement? It is not just the technology to have some of this growth. Sure.
Rob Pettman (42:37):
No, you also have to surface it to the advisor inside of the workstation in the right way so they're getting the signals, and we code it just on sort of zero to a hundred, and a hundred being the highest level, and then they see the client list, and they get to understand why as a result. So there's interconnectivity between the two, and there's some strategy as to how you deliver those signals for people to take action upon them, and then there's obviously sales management that goes above it.
Brian Wallheimer (43:05):
Great. We've gotten into obstacles, and we've talked about the potential. Let's have a little bit of fun with some predictions real quick. This is the hard part, right? Because you can probably be fairly accurate with what we might expect with Agentic AI in the next few months or so, but so much can happen in a year or five years that it's hard to pin that down. But let's be bold. So who wants to jump out there? What are we looking at in terms of Agentic AI, its role in wealth management, possible outcomes, maybe in a year or, if you want to be so bold, more than that, maybe up to five years?
Geoff Moore (43:46):
I'll go. This is what's in my brain, maybe a three-year timeframe now. I was the kid that grew up on Star Trek, so everything's possible. So I'm just going to—I'll start there that it's probably not realistic, but that's okay. I kind of go back to this thing I was mentioning with Walmart. I think we're going to have an AI workstation. I think advisors are going to have some, and whether it's in the CRM or it's in the note taker, it's somewhere there's going to be this centralized tool that advisors can just talk to. They don't have to be programmers. They can just talk to it. We're going to use these tools like Eric mentioned, this MCP server or APIs, and they're going to connect. I mean, I've already done this with some CRM systems. Literally, you can just tell it, "Here's my list, update these records," or, "I just had this," and it'll talk to multiple systems for you.
(44:35):
You won't even have to fully explain it, and it will just happen. I don't know who exactly is going to be the tool set that David's point earlier, who's going to be that central point first. I don't know. I think that's up for grabs, but I think whoever can make it easy to talk to as many other integration points as possible and then just make it easy for advisors to say, click, click, click, "I want this," and just either speak or type into it, "This is what I want to happen," and it just does it, and I need to be able to schedule.
David Breakstone (45:04):
Yeah. Let me plus one to that one, Jeff. I think you heard from Marc Benioff, "We're the last generation of managers that are only going to be managing human employees," and we're going to have a high-grade workstation. I firmly believe what you're saying, AI will be the new UI, but structured data will still play a very key role, plus.
Geoff Moore (45:30):
You can't stress that enough. A lot of people keep forgetting that. They act like structured data is going to go away. It's still really important. The AI, it needs to park the data someplace, and it needs to pull it in a meaningful way, not just a document.
David Breakstone (45:44):
So AI will be, I think will take over as the new UI, and we're all talking about Google search and being replaced now, and that's starting to take hold, not in the last quarterly results, but it is coming. I think the AI command center, you're seeing firms roll out this command center to manage all of your different agents and viewing of productivity. That's not three years. Now, whether or not it gets adopted and used is going to be, but I think the promise of AI is think about the time that you had the best assistant working for you, best teammate working for you, and now you could hire the best teammate that you've ever had on your team to manage all of these tasks that you didn't want to do so that you could focus on working with clients. I don't think it's by any means going to replace any human. I think it's just going to make that team interaction more important and that financial advisor personal engagement more important, and it's going to free up their time to focus on more strategic initiatives. So I'm very optimistic about leveraging AI in the wealth practice.
Erik Allison (46:56):
Yeah, Jeff, you're right, from a perspective of the ease of use. This is a magic to just going, "Hey, would you go ahead and find me this and then put it over here, and then by the way, remind me about this? And then on my way over here, you said this to me. Now do a research report on that and then get over that, and then when I get done working out, remind me about that call I had." And then just talk. David hit on that. That exists today. I've built stuff like that for my own personal use that just natural language talking back and forth. But to put that into a workstation or put that into an environment where from an enterprise level, you imagine being an advisor and trying to pick a new firm that you're looking to go to. Maybe you're going from a wirehouse to an RIA firm, and you're trying to interview firms and find out who's got the best tech stack, where am I going to save time. One firm over the other goes, "Hey, so the way our system works, you just come in, you just talk. You just say what you want to have happen. I want all these records to go over here. I want to do research on the top 10 viral things that wealthy ultra-high-net-worth individuals are looking at today. Write me a script and an image that I can post on LinkedIn, and then actually just post that every day at eight o'clock in the morning." The advisor is going to go, "Oh yeah, I want to pick that firm." That's a pretty sweet setup, right? So yeah, that's coming. I like that.
Geoff Moore (48:19):
The other thing
Erik Allison (48:19):
though, the talking is what's really nice because I don't know what else, what next level you get to that besides just thinking it, right? And that starts getting computer chips in your brains, but talking is the easiest thing besides just thinking it. So I think there's going to be a lot of just talk and it gets done type of solutions, which is cool.
Geoff Moore (48:40):
The other thing we keep talking about is all this efficiency serving more clients. I was talking to one of our advisors a couple of weeks ago, and we were talking about AI note-taking, and he was telling me how good it was, and actually, a short video we were going to share with some other advisors. After the video, he shared something with me that just, man, it just hit me like a ton of bricks. He's a younger dad; he's got three kids under five, and he's like, "You know why I really like this AI note taker? After my meetings all day, at night, I was spending time typing it up and doing my to-do list, and now that I have the AI note taker, I spend that time with my kids." So I don't know how that translates to the future. So are we going to get better work-life balance maybe because
David Breakstone (49:22):
of some of those? Yes, time saved for an advisor is either going to be driving new AUM or just improving your golf handicap. But that's a nice story. Spending more time with your children, I think it's also a huge advantage. It's going to change the world as well.
Rob Pettman (49:38):
I do think a single advisor office with no support staff in the future is going to be a possibility, albeit I don't think that ends up being the model. I think that people in the office will be changing their roles around client service and other sort of more human-valued interactions than the sort of brute force clunky operational nature that they might have today.
Erik Allison (50:09):
What are they talking about? The first one-person billion-dollar company, right? You're talking about the first one advisor managing a billion-dollar AUM office.
Rob Pettman (50:21):
But either way, the net outcome is better wealth outcomes for more investors. That's what that is.
David Breakstone (50:27):
How about that advice gap, too? Yeah.
Rob Pettman (50:29):
Yeah.
Brian Wallheimer (50:30):
I mean, we're always talking about advisor shortages, too, right? I mean, if we wind up in some of these worlds that you're talking about, I can't imagine that it's going to take up all of that slack, but then it makes it a little bit easier to deal with having more clients, deals with that. But it also maybe opens up the possibility of serving clients who maybe aren't traditional wealth management clients, right?
Erik Allison (50:58):
Yeah. Either moving upstream or downstream, and in your case, what you just mentioned, moving downstream: folks that maybe it takes too much time to advise someone on a $50,000 portfolio or a $100,000 portfolio, but they still have planning needs, they still have advice needs, they still have tough decisions and complex matters that they need simplicity on, and they're looking for clarity on. I think that's one of the big parts. You've got different types of areas in the space. Some focus on investment management, some do holistic planning. I think where AI and the Agentic side, where the future is really going, is more people are going to be getting better holistic advice because now it's more accessible. It can be replicated quicker, easier, or it could be packaged in a way that more people can access it. So I think part of the future is people getting better outcomes and making smarter decisions, which is cool. That's what we're all here for.
Stephen de Man (51:54):
I'll build on the talking to the system. My wife gives me a hard time. I love talking to ChatGPT. I have it give me pep talks all the time. I'm a big user. My wife calls it my girlfriend. She'll say, "Oh, ask your girlfriend. Ask your girlfriend," all the time. But my prediction for our industry, I won't be "boldly go where no man has gone before" like Geoff there, but I really like that future. But I think from a hiring perspective, as advisors start to look at the skill sets that they're looking for, I think Rob mentioned broadening those skills, but I think looking at what's your comfort with these types of tools, with either generative AI or Agentic AI, you don't need to have gone to prompt engineering school, but how comfortable or open are you to learning about those? I think hiring for these types of skills will definitely—we're already starting to see a little bit more of—but I think your mainstream advisor will start. If you think about the tech adoption bell curve, as we start to get more into the mainstream, those folks will start to be hiring for some of those skills.
Brian Wallheimer (53:03):
Great. We have just a minute, maybe a minute and a half. The last thing is, what is no one thinking or talking about that should be on everyone's radar? Just quick. If anyone has
Geoff Moore (53:14):
MCP server, MCP server, MCP server. You're going to hear it more and more. Not very many people have rolled it out yet, but it's coming. If you've never looked at Model Context Protocol (MCP) server, big deal.
Brian Wallheimer (53:25):
Tell us what it is real quick, because someone asked that earlier, and I didn't quite get there.
Geoff Moore (53:29):
Basically, if you had to connect to an API endpoint, it's kind of time-consuming, somewhat complicated. You got to tell the AI how to do it. It's not fun. I've done it. An MCP server is literally, you just drop in a link that says, "Go here," and the other side has basically redone their API to make more sense to an AI engine. So it just makes integrations go very, very quick.
Erik Allison (53:54):
And you could have a hundred, 200 different tools inside an MCP server. So for example, with my personal agent, I've got MCP servers that have 143 different tools. So when I'm talking to it, that's how you have that fluid motion of, "Hey, do this and then remind me that, and put this on my calendar, and then send a text to so-and-so, and then do research on this." That's how you can do those types of things because the Agentic part of it is it's going, "Okay, he's asking me this and this and this, and how do I access these tools?" The way it's happening is through MCP servers. So think of the power of that. No longer do you have to go one task at a time. You can have multiple tasks running in parallel, and that's why I just bring it up: it's powerful.
Geoff Moore (54:39):
If they're familiar with something like Zapier or any of these automation platforms, it's like on steroids. You don't even have to click through to create your automation. You just tell your chat and hook it up, and it just does.
Stephen de Man (54:50):
It's a beautiful story of collaboration. Just like when you think of HTTP, this universal language, that's where this comes from. It was coming together saying, "Hey, let's create a common way that we can all engage these tools." So it's actually a nice story. This is making a lot of people's lives a lot easier.
Erik Allison (55:08):
That was really nice how you put that, too. In the beginning, it was like a poem.
Brian Wallheimer (55:12):
We're out of time. I want to thank everyone here. What a great conversation. We probably could have kept this up for a lot longer, but it's a great point. A lot of great points, a lot of things for people to follow up on as they go. So thank you all for your time, for your expertise. We're going to take about a 10-minute break for those of you in the audience, and we're going to come back with Samir Munchie. He's head of Behavioral Science and Simulation at EY. I'll be chatting with him for about 45 minutes, so don't miss it. We'll be back in 10 minutes. Thank you so much.
Erik Allison (55:41):
Great. Thank you. Thanks. Thanks, everybody.
Opening Remarks & Personalization is Everything: Wealth Agentic AI Use Cases to Deepen Client Engagement
July 28, 2025 5:14 PM
55:50