How LPL Financial and Anthropic Are Setting a New Standard for Agentic Advice

Peter Nolan [0:01] And we think the ability to lift agentic AI out of the developer and make it more available to knowledge workers is what really changes the world, candidly.

Matt Enyedi [0:16] Welcome back to If You Could with Matt and Taryn. Hello, I'm Matt Enyedi. And

Taryn Huget [0:21] I'm Taryn Huget.

Matt Enyedi [0:23] Well, Taryn, it has been a minute since we recorded our last episode, and I was thinking, I think our listeners need a little bit of reintroduction to their hosts. I'm not talking anything heavy, just an icebreaker to get us back on common ground. So today, my friend, we are going to play the name game.

Taryn Huget [0:38] You know, I do think a reintroduction feels just a little bit unnecessary, but you know what, Matt? I am gonna play along because I do love a good name game. So what are we talking? Celebrity name, superhero name? Where are we going? Both

Matt Enyedi [0:51] Good choices, but no. You know, I was thinking with so much talk of agents these days, I wanna figure out what our secret agent names would be. Ooh,

Taryn Huget [1:00] Secret agent. I like it. So what's the formula for a secret agent? Street you grew up on, name of first pet?

Matt Enyedi [1:05] No, no. The secret agent, it does have a formula. It's this. It's your middle name plus the neighborhood you lived in when you were born.

Taryn Huget [1:12] My middle name is Lynn and I grew up in Scripps Ranch, so I'm going Lynn Scripps.

Matt Enyedi [1:19] Ooh, Lynn Scripps. She seems kind of mysterious. I

Taryn Huget [1:21] Know, she does. What's yours?

Matt Enyedi [1:23] Okay, well, my middle name is Klaus, and I was born in Ocean Beach. So drum roll, I am Klaus Ocean.

Taryn Huget [1:30] Klaus Ocean. I like that one. So if you could, with Klaus and Lynn - Oh

Matt Enyedi [1:35] Yeah, those sound like a couple of nefarious characters. I gotta say I like them already. You

Taryn Huget [1:40] Know, as much fun as the Lynn and Klaus thing sounds, people aren't talking about secret agents these days. They are actually talking about AI agents, Matt. Yeah,

Matt Enyedi [1:50] Yes, yes. I knew that and I knew that's where we were headed, but I wanted to have a little fun along the way. But here's the cool thing, AI agents also have a naming mechanism. So here's what it is. First, it has to be a single word, no last names allowed. And from there, it's either the name of someone who lived in the era of black and white TV.

Taryn Huget [2:10] Right. Like Hazel or Claude?

Matt Enyedi [2:13] Exactly. Or it has to be a word that's seemingly made up, but actually has a deeper meaning behind it.

Taryn Huget [2:20] Nah, we're talking Cyan. We have finally made it to the reveal. Thanks for hanging in with us, folks. Cyan is one of the most transformative AI agents on the scene and the talk of the town in the financial industry.

Matt Enyedi [2:34] And not only does it follow the naming convention, there's so much more to it than that. You see, Cyan is a color. Think a mix between blue and green, and it just so happens it's the mighty blue of LPL's logo combining with the rich history of Commonwealth's green logo. And Cyan is the perfect blend of our partnership, the incredible scale of LPL delivered through the personalized layer of service and care of Commonwealth.

Taryn Huget [2:57] Now that's a partnership, a color, and a name that I can definitely wrap my arms around. We're gonna dive into that with LPL's chief product and technology officer, Greg Gates, the creator of Cyan, along with Peter Nolan, the head of asset and wealth management at Anthropic. And we're gonna talk all things Cyan, AI, and the continued rise of Anthropic.

Matt Enyedi [3:19] I love it. Let's bring him in right now and get started. If you could with Klaus - And Lynn. Starts now. Greg, Peter, welcome to the podcast. How are you today?

Taryn Huget [3:32] Hey, guys. Doing really well.

Peter Nolan [3:33] Thank you for having me. Thanks for having me. Yeah, we are

Matt Enyedi [3:36] Stoked to have this conversation. It is one we've been waiting for for quite some time. So we've gone a lot of places, from labor markets to generative, to security, to cyber insurance, to things related to singularity and quantum, and there's so much more to go here. But today we're going to talk about LPL's new agent, which has really changed the landscape of financial services, Cyan. And it's really the brainchild, uh, of Greg, you and your team, and then leveraging the incredible power of Anthropic underneath it. But before we jump in, Peter, I gotta turn to you. I was just reading. Anthropic arguably is the fastest growing technology company ever. That's a, a pretty big headline to have. So how's things going?

Peter Nolan [4:19] I'm hanging on for dear life, Matt. No, things are going fantastically well. Some people say I made a good career decision. I just hit the year mark. It's incredibly exciting. I don't even know particularly where to start, but the pace of innovation is furious and being able to apply this to an industry that I'm passionate about and that I came from is, is like one of the highlights, if not the highlight of my career.

Taryn Huget [4:39] So Peter, I'm gonna stay with you a little bit. You're talking about today, but I wanna talk a little bit about the future. So when you think about where AI is today versus where it's gonna be two, three, maybe even five years from now, what do you think people are underestimating the most? And then also what should our clients and our business be doing today to prepare for that future that may arrive much faster than we are expecting?

Peter Nolan [5:06] Yeah, I think, you know, there are, there are three things to think about. One of them is that the technology is there to do all of the things that your and my imagination immediately race to when you ask that question. The limitation on our ability to deploy the technology to do all those things hasn't kept pace with the technological innovation. And so when I think about the next two years, it's much less to do with model quality and capability and much more to do with things like scaffolding, safety, responsibility and, and regulation. And so my focus anyway in the near term is closing the gap between those two things, which we refer to within Anthropic as the capability frontier and help firms like LPL deploy the technology in a safe way to begin to realize some of the capabilities that the, that the technology already has. So I'm very excited about that. I think the LPL team has a pretty good beat on, on what's coming and, and how we can accomplish that, and we're pumped to be along for the ride.

Matt Enyedi [5:57] Like you said, you're at the front of this unbelievable movement of innovation and where you're spending a lot of your time is on safety and security and making sure that it's applied thoughfully.

Peter Nolan [6:08] I don't think I've spent more time in the SEC and FINRA rules than I have since joining Anthropic, even when I have Fable alongside to read them with me.  But in all honesty, I think like the way that heavily regulated institutions have to think about those things is going to dictate the success of this technology and its ability to make the impact that we at Anthropic hope it makes. Uh, I love to keep Greg on edge. I got to keep him on his toes. He was getting way too comfortable on top of the world at LPL, but very much a partner in figuring this all out and making sure we can do this right.

Greg Gates [6:36] Yeah, and I think the, the partnership is really important here, right? I think one of the things that stood out to us about Anthropic is not only do they see us as a strategic partner, but they also understand the complexity of wealth management. They understand what our responsibilities are, not just to adopt AI, but to help shape how it's applied to financial services in general. Really, advisors need both. They need access to the best AI, and they also need to be able to use it responsibly from a compliance lens. And together, we're building the capabilities for advisors that allow them to do so.

Matt Enyedi [7:08] And I wanna get in from, and, and Peter, I'm gonna turn it to you. So much of what most folks have adopted to date is generative. Like we all use Claude to find answers, to dig deep, to go down rabbit holes and forget about time and responsibilities because we're so deep into philosophical questions. But what we're seeing now is that movement from generative to Agentic. And I'd love for you to just kind of maybe walk our listeners through what that progression looks like. Is that iterative or is that a quantum leap? Is Agentic the future, I guess is the question I've got.

Peter Nolan [7:42] Yeah, I think to take a step back, where Anthropic really caught fire was in the beginning of last year, uh, when we released something called Claude Code. An LPL's pattern of first adopting Claude as a coding assistant, as a coding agent is, is very, is very common in the industry. What Claude Code did differently from the other code tools was instead of just complete a line of code, you could tell at the destination and it would figure out the road to get there. So you could, you could delegate to it. It would take agency on a desktop for a developer and accomplish tasks end to end. Uh, that technology turbocharged the growth of this company. I think a lot of like what has been our success is attributable to Claude Code early days. But something pretty fascinating happened last year and, and midway through the year we saw that non-technical people were using Claude Code in non-technical ways, meaning knowledge workers had picked up this tool that could take agency on the desktop to do things like write documents or to build spreadsheets. And we had a chat interface, we had Claude. And so we launched something called Claude Cowork that basically combined the best of both worlds, and that is user-friendly experience that you could delegate tasks to. So I can now chat and instead of just chatting back to me, Claude not only says it does, I can tell Claude open this Excel document, reformat the file, run the numbers, put it into a Word doc, download it, write an email, attach the doc to an email, and then I click send at the very end. So that's been a really exciting thing to see. And we think the ability to lift Agentic AI out of the developer and make it more available to knowledge workers is what really changes the world, candidly. And so again, this will go back to the scaffolding, go back to transformation, go back to behavioral change, but the capabilities are now there for the knowledge worker to take these things on.

Peter Nolan [9:14] We've

Taryn Huget [9:14] Talked about the partnership with Anthropic. We've talked about how Anthropic is powering many of the Agentic experiences we're seeing across the industry, including with some of our competitors, which we know Hazel put a stake in the ground as being sort of an early mover in the wealth management space. And I'd really say a lot has changed since then. So Greg, as you watched that first wave of AI agents enter the market, where did you see that they got it wrong and maybe where they fell a little short, but also maybe where they got it right and how did that shape your team's vision for Cyan?

Greg Gates [9:49] Yeah, I think there are a couple of things here. First, I think it's really important for us to care for the cybersecurity governance, data sharing controls that are designed for a highly regulated industry. So what that means is advisors need to be able to safely trust that they can take advantage of AI within a framework that's built to address the important risks and compliance and security considerations that, that they might have. I would, you know, maybe propose that advisors often thought that those things were cared for and sometimes they were and sometimes they weren't and it was hard for them to know. What we're focused on first and foremost is when people use Cyan, we want them to have confidence that those, I'll call it table stakes, are actually being cared for and cared for in a responsible manner. And then we saw some areas that weren't being focused on as much, how to help an advisor grow their practice, the, the data insi- insights around their book and moving beyond efficiency to how to become a real business driver for them. This is where we could use the power of our proprietary data that we know about the advisor and their book and match it with the LLM to really deliver some on-demand growth coaching and highlight opportunities for advisors to think, you know, within specific households, how to recommend practical next steps for helping advisors run their practices more like the top performing firms. So I think there's a little bit of all the above in there, like how do you care for the basics? How do you deliver solutions where there's definitely demonstrated value being delivered into the market? And then where do we see some opportunities that maybe others haven't seen and we can leverage the best of our company and our data set to, to be advantageous for advisors?

Matt Enyedi [11:33] Exactly. And Greg, I think that's the key point is we have to give advantages to our clients. And for our listeners, if we take a step back on Cyan, Greg, when you guys built it, you started in three key areas, account maintenance, so how do advisors more efficiently manage their accounts? And then you really added on the, and doubled down on financial planning and growth. How do advisors and institutions advance their business? And those last two areas are where we believe Cyan is delivering something truly unique to advance our profession.

Taryn Huget [12:01] And to be fair, there are other agents out there solving for financial planning as you would expect, but what I'm really excited about is how we're helping our advisors grow. We're building on the work we did last year with the LPL Advisor Growth Index, think a generative output, and we're taking that and turning it into identic outcomes. Which

Matt Enyedi [12:22] Is a nice preview to an upcoming episode where we're gonna dive all the way into Cyan growth with Kraleigh Woodford. But I wanna move forward in building on this whole what's coming next idea. And Peter, I know you're chomping at the bit to tell us all about the latest announcement from Anthropic.

Peter Nolan [12:38] When this podcast goes live, we will have just announced Claude for Financial Advisors. We're very excited about that. What that looks like is bringing this specifically to the financial advisor. And what that looks like is having Claude Cowork be connected to the rest of the advisor tech stack with many of the usual suspects that are part of that tech stack and being able to orchestrate across all those components using Claude to truly get end-to-end automation, which I think is what AI was always promising us. So we're, we're very excited about that.

Matt Enyedi [13:01] So Greg, maybe you could tell our listeners what that actually means for them because in the spirit of scaffolding and protection, it doesn't necessarily mean it's unleashed completely to all of our clients.

Greg Gates [13:11] Yeah, not, not at first, but I'll back up a second. Getting to Cyan, what we believe the opportunity was, was to create an AI experience that really connected those pieces together, reduce the swivel chair experience that advisors have, remove friction and help people move from insight into action. So that's why we built Cyan, and it is directly embedded within the advisor workflows, not just around the model itself. When you bring in Claude for financial advisors, I look at it as more tools at our disposal to build those types of products going forward.

Matt Enyedi [13:44] Right. And to build on that, Greg, we're bringing the best of this innovation to advisors in ways that are practical, yes, but also secure and compliant. And so when it comes to Claude and many of the Anthropic MCP integrations, we're exploring ways Cyan will leverage those with all the additional controls needed to safeguard sensitive data, because for financial advisors and their clients, we know they prioritize security above all else.

Taryn Huget [14:07] Absolutely. And that said, we're thrilled about all these advancements that our partners like Anthropic are making, but critically, Greg, how you and your team are using our unique data sets and integrated capabilities to optimize them and accelerate our AI roadmap. But now I'm gonna pivot us just a little bit. You've mentioned risk, security, we've been hearing the words trust. In our industry specifically, trust is everything. So as we start to introduce more advanced AI capabilities, especially the agentic AI, we have to be equally focused on privacy and security. I know so many of our clients, it's a thing that's top of mind for them of, "Oh, I wanna go there, but I don't know if my privacy and security, both my business and my clients, is protected." So when you think about AI in financial services today, what's the biggest privacy or security risk to our advisors that they should be paying attention to, and what safeguards do we have in place to help protect them and their clients?

Peter Nolan [15:09] I think privacy and security are like two of the things that we think about most, because again, model capabilities are there. What's holding us back at the moment is our ability to accommodate the existing regulations and then to make sure that everything's done in a safe and responsible way. With regard to privacy, like everybody should know who is getting your data and what they're doing with it. It's kind of as simple as that. Anthropic is explicitly not training on your data. LPL is an enterprise customer, and that should be table stakes for an LLM. In terms of supervision and accuracy, we've recently launched our compliance API, which Greg and his team is very familiar with, but that was largely in response to feedback that we were getting from heav- heavily regulated customers that they needed pieces of data from our technology in order to roll it out in, in a regulated setting. So those are two things to be sure that are, are top of mind for advisors considering adopting this technology, because I think you're right. Like the last thing you wanna do is, is create some sort of data leak, uh, or make clients feel like they can't trust the technology at the end of the day. They're with advisors, and I think advisors have staying power in the space of financial advice strictly because, like, they're so well-trusted, so we would never wanna jeopardize that.

Greg Gates [16:13] Yeah, no, it's well said. And then what we add to that from a, just a responsibility perspective is we had a couple of security layers. As Peter mentioned, making sure that everybody knows where your data's going and, and not training on it. We also encrypt data so that it is not really usable from others. We hold the encryption keys to that information. We also are working to make sure that access to these tools is done so through secure browsers and have things like masking of PII data in them. So we add additional security controls in there as well, just to make sure everything here is very, very safe.

Taryn Huget [17:05] I wanna go back to trust when it comes to Cyan, but, but with a little twist. When AI agents confidently generate information that's inaccurate or just simply not true, what protections are in place to minimize that risk? I mean, how should hallucinations be managed and what responsibility really sits with the human?

Greg Gates [17:27] I think trust goes beyond just the security concerns. I also am making sure that the advisor trusts the responses that they're getting, that they understand what the solution is doing and what it's not doing. So for us, advisor trust is a non-negotiable, and all of this goes into just a general thing where we do not believe that AI replaces human or advisor judgment. We view these tools as augmenting humans and human judgment, and an AI agent might surface, you know, which clients warrant attention, it might draft suggested outreach, it might take work off the hands of somebody, but the advisor remains the decision maker. The human review checkpoint is at the point of action. The advisor always is evaluating the recommendation. The advisor is always applying context that the model might not have. And then ultimately, the human does decide whether, when, and how to engage. And in a primarily relationship-driven business such as ours, that human interaction, that human accountability, ultimately that has to stay with the advisor so that they trust the solution and also so that their investors trust them.

Peter Nolan [18:42] I think it starts in the model provider. Anthropic's models are explicitly trained to be helpful, honest, and harmless. Being honest involves providing accurate information that is interpretable and often has prominence, meaning Claude will, will often cite its sources and, and provide you links back to those sources, trying to make, like, its thinking as transparent as possible. In fact, when Claude is reasoning, you can see the steps that Claude is taking along the way to get to the answers that it gets to. So we are always trying to think of ways to make it more transparent as to, like, where the outputs are coming from and why Claude has landed on them. Increasingly, in some of our more recent models, like Claude will tell you when it's not sure, Claude will kind of give you its degree of confidence in the answer that it's giving to you. And so I think a major problem in hallucinations will, you know, at least begin to diminish to some degree just because of model training and model quality. That said, I think it's still critically important that advisors are reviewing everything that comes back from an LLM. Like it's, it's made very clear in the regulation that it's incumbent upon the broker dealer or investment advisor to make sure that anything that goes to an end client has been vetted, just like, you know, they would be responsible for doing so if it came from a deterministic piece of software. So I, I don't think that responsibility will ever go away, but I think a lot of the technical capabilities and, and what we're launching in the models and the harnesses around them should hopefully help minimize this problem moving forward.

Greg Gates [19:57] I very much agree. I think there are a couple of things, uh, to just add to here. One, we will strive to separate deterministic calculations from those that are subject to LLM reasoning. If the LLM helps interpret inputs, explain outputs, et cetera, but like a v - a really authoritative compliance decision is going to be required for that specific use case, we will probably continue to use technology that is very deterministic and think of a, a rule engine that, like, it will enforce a certain action here or there. I think the hallucinations for those who have been using AI, they're getting less, right? That actually makes the risk, in my opinion, that you might accidentally approve something that was a bad result, actually go up, because you're more likely to see a higher percentage of the time, a really good result, and you might lower your guard, as it were, a- against finding a bad one. That's a, a risk that I worry about a lot. I think it's important to have a mental model for this. Think of agents and the support you're getting from these AI tools as almost like an infinite number of really, really smart junior employees. With that mental model, you can think these are highly capable, highly intelligent, high work ethic, they're gonna work around the clock, and they're gonna be giving you really, really great work. Also, you need to provide some oversight. You need to look at the results. You need to make sure that you ultimately understand that you're responsible for them because you are in charge of those workers, right? And, and I think it's an important mental model to have.

Matt Enyedi [21:35] Look, the technology is ready. The security is challenging. The change management is complex, and the iterations will continue to evolve and maybe even accelerate, and you've gotta dip your toe in the water now or you're never gonna learn how to swim, and the current's only gonna get faster as we go. And then there's these questions about how much do I use it and, and how much can I trust it? And I wanna just kind of go to the other side. And we're now a couple years from here. People are using it all day, every day. They're leveraging it everywhere they go, yes, for generative, but certainly for Agentic, and it is doing work for them, and they are increasingly comfortable, and they have put their business in cruise control. But sooner or later, they start to fall asleep, and they go from cruise control to no control. And they actually start to lose the very essence of their business, and maybe even now I'm getting existential, the very essence of their humanity, because they're no longer asking the tough questions. They're no longer reading between the lines. They're no longer watching their client's eyes to see if they match their client's words. And suddenly, we've lost what makes it special. We've lost the nuance. We've lost the friction that causes us to dig deeper, and perhaps we've lost the differentiation that makes humans necessary and actually really critical in this business. Peter, I know you swim around in this all the time for both business perspective, but also your own personal health and your own personal life and how you're absorbing it and leveraging and seeing all the benefit, but also as a true consumer, potential harm too.

Peter Nolan [23:07] I think I'll start from an advisor's perspective, and then I'll double click into my own. I think, hopefully, all the things you just described, eye contact, like, tonality, these are all things that happen in face-to-face meetings. And what we hope to do with AI is allow advisors to have more of them, because we think that's where differentiation is generated in this industry. So hopefully, anything that's differentiating an advisor is, is being sort of emphasized and, and amplified by this technology. They spend less time in the back office preparing for meetings. They spend more time in the meetings with clients, helping them through life's problems. We'll see if that actually pans out. We hope so, but that's, that's certainly our aspiration for what the technology does.

Matt Enyedi [23:44] It's pretty cool to think, Peter, how digital for so long made the house call no longer necessary. And I think what you're saying is digital is now making the house call more possible and more necessary than ever.

Peter Nolan [23:57] That's absolutely right. And the old house call used to be a lot different than the new one. The new one's a higher level conversation. It's about the things that really matter to people. It's about things that can't, for whatever reason, couldn't or shouldn't be conveyed via writing. Now advisors are able to go deeper with their clients. They're able to make more meaningful relationships, and they're able to affect more change.

Taryn Huget [24:15] In the spirit of If You Could, which is where we like to end every episode, if you could describe what happens next, what does the future look like from your perspective? Where is AI taking our industry and our businesses and our clients, you know, and how quickly do we get there?

Peter Nolan [24:31] I think, you know, in many ways, it's gonna benefit people that are curious. Like, if you sit down and you're talking to an LLM and you actually have some time, you can ask it any question in the world, and it's becoming smarter and smarter about everything. In fact, like, if you ask, you know, Fable some very existential questions about very complex and nuanced subject matters, like, it can go very deep and, and tell you all about them. You can ask follow-up questions and you can really learn a lot in a very short amount of time. And so on the one hand, that makes you think, like, education is gonna be, like, unlocked to many more people who haven't historically been able to afford it or otherwise get exposure to somebody that knows all those things. On the other hand, it, it almost encourages you to get lazy. Like, I read a lot less, like, now, which is probably a, an issue, and hopefully I fix that at the tail end of this year. But if I can get to the root of whatever I'm trying to get after, like, there's no necessarily reason to, to take a roundabout way of getting there by reading an article or a book when I could just, like, get a very clean, clear answer that has all the relevant context of my life taken into account on the way. So, totally TBD on whether that lands in a great spot or, or a not so good spot, but I tend to focus on the bright side of things, and I think if we, if we focus on the core fundamentals of, of our personal lives and try to keep everything in perspective, uh, I think it should land well.

Matt Enyedi [25:41] You know, Peter, it's funny. I'd listened to Marc Andreessen say something similar to the first side of that, which is in the history of mankind, all of the great scholars were all taught by private tutors, and that was only reserved for the aristocracy. And now, every human who has broadband or Skylink has the ability to have their own private tutor and learn just about anything and everything they want. And on the flip side, we could just become complete lazy bums and never read a thing. But Greg, I wanna talk about how it goes to business. When you think about where we go from here and how we help advisors and institutions grow their businesses, serve their clients, what gets you pumped?

Greg Gates [26:18] I get pumped by a lot of things. I don't think that this is necessarily, like, in the future. I think it's kinda happening right now and accelerating. Whatever an advisor thinks is their value definition, the answer right now is they can do more of it, right? They can serve more investors. They can go deeper with more investors. They can offer more capabilities to investors, whatever they wanna define as their differentiating value. The cool thing is, you can do more of it now, and you can do more of it partly because you are spending less time as a human working on things that ate up a lot of your time, and you can also do more as a human because you're powered by some of these great tools that augment your knowledge and skillset and your ability to get some things done. The trade also is an interesting one, and Peter said it actually, in business and in your personal life, there's not as much downtime anymore. Those mindless tasks that used to eat up a lot of your time that now you have digital assistants doing on your behalf, that was time where your brain recovered a little bit. And so I do think we're going to have to be intentional. You mentioned me reading books. I intentionally, because that's how I recover, read some books, watch a baseball game, go spend time with my family, right, and make sure that I'm having the human interactions, because it's really easy to get excited about all the value you can deliver and to lose your attention span. I think it's important for our own humanity. It's also important, if you come back to this industry, the differentiation, the real competitive advantage will come from the combination of systems that surround an advisor, that bring you connected data, deep business expertise, strong governance, and also the human element of the confidence and trust that you have between yourself and one of your investors if you're an advisor.

Matt Enyedi [28:09] What I know is this, is in a world that is increasingly digital, partnerships like the one that you two have and the partnerships like the one our two firms have are really powerful, and I think it's gonna generate and drive our clients forward, and I am so excited for what's next. Peter, I know you never have any time, so thank you for spending some of it on us. And Greg, don't worry, we're safe, we're secure, you can go to sleep. We're gonna be fine.  Sounds good. Thank you both.

Greg Gates [28:31] Thanks for having us. My pleasure.

Matt Enyedi [28:33] Thanks, guys.