Nitin Tandon: AI is no longer waiting for us to define the problem, it's showing up with answers to questions that we haven't even thought about asking yet. This space is moving so fast that it can feel discomforting. But the companies that are going to define the next decade are the ones that are going to lean into that discomfort, that are moving towards the frontier and not away from it.

Today, AI is something that our clients expect, our competitors are organizing around and the markets are actively pricing in. That's what real disruption looks like, a change, a shift in expectations. And when expectations change, standing still is not conservative. It is dangerous. It's risky.

Now, the good news team is that Vanguard did not stand still. We made an early, deliberate commitment to engage responsibly with AI, not because it was in the news, not because we wanted to do it as a side project, but because we believed in its enduring capability to supercharge our mission. And a couple of years later, that commitment is showing real momentum.

First, AI is diffusing across the organization, not as a site project here or an experiment there, but in meaningful ways. It's embedded in the way our teams are supporting clients, supporting advisors, building software, or driving productivity across the organization. Second, we have made some deliberate investments, some pointed bets, not scattered portfolio of investments.

There are four areas where we believe AI will add and drive disproportionate benefits, whether it's client experience, new products and services protecting our clients or through productivity. Those are four areas we're deliberately investing in and you will see during the course of this conference some examples of that.

You'll hear about how we are identifying and preventing fraud earlier in the process, how our teams are prototyping and imagining solutions and vetting their ideas before actually building it. So now we have a better way of designing products, not just delivering them. You will see how agents and Copilots are helping our product teams, our engineers, our service teams shrink down timelines that are typically measured in weeks and months. And the third and perhaps the most important is we are building foundational capabilities to scale AI responsibly. And that team is really important because speed without discipline does not scale and discipline without speed does not compete.

50 years ago, Vanguard was formed in a simple premise that doing right by investors and running a successful company are not at odds with each other. They're the same thing. That belief, that conviction, is truer today than it ever was because Gen. AI, or AI done right is consistent with everything that we believe in. It lowers cost. It scales access. It allows us to provide better advice and better service to everyone who needs it. It allows us to do more for more people with more precision.

Sid Ratna: So let me tell you a little story. I work in the FAS business. The FAS business, the clients are financial advisors, right? They control most of the AUM of the wealth industry in the United States. The number one most valuable meeting that we can ever have with a financial advisor is when they literally say, "Here's my portfolio. Here are all the competitors I use. Here's why. Here's what I'm trying to solve for." We literally have a mind map of what they're thinking, who they're using. From the moment a client tells us, "I want you to help me with my portfolio," it takes us three weeks to get back to them, at which point 30% of the clients are like, "Never mind, the moment has passed." So the micro process, when my team got started, was that we were going to use generative AI to shrink down the portfolio analysis so we could do portfolio analytics in real time. That's going to fix the problem. You know what? We learned the portfolio analytics in this three-week process only took 24 to 48 hours. So yeah, we took something that took two days and shrunk it down to two minutes. But do you know what the real issue was in terms of getting back to our client? Here's what the process looks like: the client says, "Here's my screenshot." The wholesaler or salesperson says, "I got this," and emails the internal partner. The internal partner takes the screenshot and manually retypes every ticker and weight into SharePoint, at which point the internal partner sends another email to the junior portfolio consultant to say, "I've uploaded it." The junior portfolio consultant does the work and emails the senior portfolio consultant to say, "I've done the work." The senior portfolio consultant then emails the wholesaler, who emails the client to get on the call. So this is the problem that AI surfaced, but one we need to solve end to end. That's a very real example of where there's tremendous value, because if we fix the end-to-end process, we can 2X or 3X the output without additional investment.

Jennifer Manry: There's a desire sometimes to just bolt AI onto the side. You could have stopped there. You know where I'm going with this. You could have stopped and let the rest of the process stay as it is. And I think if you went out and used Copilot Researcher and asked, "Show me all the places where bolting AI onto the side of things created an incredible amount of value," you'd find research from every major organization—MIT, McKinsey, BCG, you name it. In places where they simply bolted AI on, they did get some value. I'm not saying there's no value in that. But Sid is bringing up a really good point: it surfaced something that said, "Hey, we should actually take the time to reimagine the whole thing." We don't want to just bolt AI onto the piece that captures notes. We have an opportunity to redesign the entire flow. The research that is increasingly becoming available highlights the fact that the real art of this—but also the hard work—is saying, "Okay, we've found a place where AI surfaced an end-to-end issue. Now we should take the time to redesign the whole process to really get the value out of it." The real opportunity comes from reimagining the way the end to end works, not just sliding AI into small, discrete use cases. I'd say the other thing is that when you just bolt it on, it makes it feel optional, like you don't really have to use it. The work that Sid has done in this case, and in several others, has really been about taking the full end-to-end process and looking at how we completely reengineer it to leverage the best of what AI can do while also putting humans in the best places.

Kaitlyn Caughlin: As you're going through that right now, what's hard about it? What's one example of something that's going well, where you're picking up speed, and what's one area where you're thinking, "Oh man, here's a challenge we didn't expect?"

Sid Ratna: It's really two sides of the same coin. One thing that is going really well is that, as the story unfolds, I have a partner in the business who owns the portfolio consulting and analytics offering. She actually comes from another asset manager, so she brings a valuable external perspective. She looked at our tools and said, "This is better than what another major asset manager has." The variance isn't around the model or generative AI hallucinating. The variance comes from the fact that people have different approaches to analyzing a portfolio. The reason I'm saying this has been such a great thing is that the work has actually created more alignment between digital, data, and the human teams.

Nitin Tandon: Real quick, the story from Airbnb—and tell me how much of this is urban legend versus reality. I think in 2009, Airbnb was a struggling startup. Growth had stalled. They weren't getting new bookings. So the founders and management team decided to put their laptops aside and actually go visit hosts and guests to understand what was missing. Initially, they thought it was a product problem, but after those conversations they realized it was a trust problem. People were afraid of strangers staying in their homes. What if they trashed the house? How do I trust where I'm going? Based on that listening tour and those visits, they implemented a number of changes and evolved Airbnb from a marketplace into a trust platform, building in ID verification, property verification, and other features that helped people gain more confidence. Those changes helped restore trust and put the company back on an upward trajectory.

Tara Bunch: Yeah, that's spot on. And it's the journey that really, even to this day, Airbnb continues. You know, it's fascinating. I think the question was like, how have Airbnb and Apple sort of leaned into this whole idea of customer experience being a differentiator? You know, Apple, you know, way before I joined, like 2001, when Steve reentered the company, it was struggling. It's not the iconic brand that it is today. And one of the things he did is he decided, around the time that the iPod was being introduced, and he was really frustrated with the fact that he was selling these, really in this case, the iPod was a very novel product, through all these third-party distributors and they just didn't get it. They didn't know how to sell it. It was just a thing that sat on the shelf. There was no kind of customer education as part of the sale. The service was really wonky. And so, you know, if you had a problem with it, they'd just give you another one. And so he decided that was the time when he decided to stand up all of the Apple retail stores, which was really, I mean, the beginning of a very different Apple over the next several decades. He wanted control of the customer experience from beginning to end. And he wanted it to be absolutely unbelievable. And he didn't put these stores in strip malls, which is where you typically found a lot of the kind of electronics companies. He put them in beautiful high-end malls, and these were some of the most beautiful, you know, shops inside of these malls, all glass and incredible. And people were so riveted by the experience of coming in there and trying out the products. If something went wrong, they could ask a Genius what to do or get it repaired or just get instruction on how to use the product and get more out of it. And that's when Apple really took off, followed by obviously the iPhone and every other product since then. And Steve really got that companies that focus on just the physical product and not the experience of using the product are kind of pushing on the wrong side of the balloon, so to speak. That really, at the end of the day, people will buy a product and they will come back to you again and again when they first trust you, but also when they feel like you really care about them as a customer, that you're investing in them and you want to make sure that they get the most out of that product and you've got their back. And Airbnb, interestingly enough, when I joined Airbnb, it was going through another sort of crazy time. It was COVID. And so you might question my judgment, but after having six kids, you know, it's not really that great. Rational people don't make some of these decisions. But I decided to join the company in the middle of COVID. They were going to take the company public. And actually they were really struggling on a number of fronts. But one of them was that in addition to, you know, obviously we'd lost a lot of business, we could see that as summer was coming and people were getting antsy in their homes and saying, "Hey, I want to get outside of my apartment that I've been trapped in for two months," they began to book Airbnbs like crazy. And so the company just exploded, which is a good thing. But also we were grossly unprepared for that level of growth in the middle of such a chaotic time. And we really doubled down at that point in time on customer experience. We felt like people were afraid of COVID, they were afraid of getting sick, they were afraid of being exposed to hosts or any other people they might come in contact with. And we doubled down on that whole customer experience. We put in a whole cleaning process that we guaranteed every host that was on the platform was being held accountable for. We really focused on our whole homes where people would be isolated, like you said. We really focused on pictures and descriptions that helped people find the right property they would be comfortable with. Customer support was—we went completely virtual as a company and we put our customer support completely virtual, like a lot of companies did. And we really began investing in, at that time, I would say it was more traditional machine learning, but we were really concerned that people fell into two groups that were going on Airbnbs at that time. One group was families that were desperate to go somewhere with their family that was safe. And the other were young people who were not afraid of COVID at all and were using Airbnbs to throw parties. So my first big role, my first big job at Airbnb—and hosts did not want to host if people were going to throw a party—so you had these two groups, right? So we spent a lot of time then building a machine learning algorithm that has now evolved over time to detect when somebody was probably going to throw a party. And you can imagine, as a mother of six, I had some pretty good intuition on what that looked like. Just like, you know, when the 22-year-old books a mansion and says it's a family reunion, you know, I mean, it's like no family puts their 22-year-old in charge or something like that. So it was this whole process of building trust and making that safe for everybody that was involved that was key to us coming out of that. And with the platform that we have today.

Nitin Tandon: Fascinating. I'd love a copy of that machine learning model you've seen now at least for four years in the time that you've served on our board. You know, one aspect of financial services I'm sure through your personal finances you've seen, and other aspects as well, how would you compare and contrast the financial services client experience compared to some of the high-tech experiences that you've seen at Apple and Airbnb?

Tara Bunch: Yeah. And I, you know, obviously I'm not an expert in all of the financial services areas, but I'm ramping up.

Nitin Tandon: Trademark qualities, like I've learnt a lot from Tara over the last four years. You know, humility and insights are two things that distinguish her. So don't believe her when she says she's not an expert. She's a deep expert.

Tara Bunch: But here's what I think. Not just financial services, but I would say many industries sometimes are too transactional in the way they treat customers. They treat that interaction like a transaction, and it's not. It's a relationship. And I will say, I think Vanguard is one of the companies that figured that out early on. The trust that Vanguard has is because people believe in the brand, they believe in the relationship that they've built with the company over time, and they know they can trust this company with their money. And I think that is an incredible strength to be built on. But I would say that in general, in the industry, treating customers as a transaction is a weakness. And I think in the age of AI, it could become a death knell because AI will do transactions better than any industry, company, or brand can ever do it. But they cannot build a relationship. Only a company can do that, one that really cares about their clients and is willing to invest in that relationship. And so I could not be more excited, honestly, about the position that Vanguard is in because I think you already have a relationship with your clients, and now it's how do you build on that and strengthen it and make it a differentiator that is impenetrable by any other company, but also cannot be disrupted by any other technology. In fact, it can actually help you build on that.

Jes Koepfler, Ph.D.: I mean, look at her. She's got my great hair. She has my smile. For some reason she's missing an earring. I don't know why that is. But listen, if we do this digital twins thing right, she's going to have my insights. And our objective here is to make sure that she doesn't forget that I'm the human on the other end.

Peter Borysov, Ph.D.: How do we know that something that we've built is working as expected? Well, we can run an experiment, we can collect client feedback from surveys, or we can extract insights from phone conversations. We can do other types of user research. So we thought about this problem. How can we scale it and how can we identify friction points as early as possible? Well, luckily there was this thing called AI that was super happy to talk to us about anything and answer any question. So we started talking and asking questions, and we built an AI system that simulates how real clients would interact with and experience various things at Vanguard, maybe marketing content or web journeys and so on. So what is it? What is a digital twin? A digital twin is a safe way to predict client preferences and behaviors and predict them early. How does it work? Let's dig into some code. What you see on the screen is the set of functions that define Digital Twin and how—I'm just kidding. I'm kidding. We just had to put the slide up to qualify for the IT conference. But this is how it actually works. It has three components. The first component is context. Who are we trying to simulate? Is it a first-time investor or someone heading into retirement? What are their goals? What are they trying to do? With digital twins, a few parameter tweaks will allow us to simulate virtually any client, even those that we realistically would never be able to reach. Billionaires, for example. They are typically too busy to give us an hour of their time to answer our questions. But digital twins will happily answer thousands of questions. They will never get tired, they will never complain, and they will never say, "Can we take this offline?" Second is the experience itself. What are we trying to get feedback on? Is it a web journey? Is it marketing content or is it a new product? Digital twins can walk through any client interaction and tell us what's working, what's confusing, or what might cause a client to pick up a phone and call Vanguard. Last are the insights. With the right context and experience, we can uncover things that are potentially problematic before they go live. You can think of it as test driving in reality, but without production rollbacks or angry emails. So if you're a skeptic like I am, you're probably thinking, are digital twins actually reliable? Can we trust them? I mean look at these guys. They are powered by LLMs, which are non-deterministic, so they will happily hallucinate any response. How do we know that simulated responses and behaviors are actually consistent with our clients? Over the last year, we went on the journey to answer that exact question. We built client personas grounded in real data. We back-tested those personas against real client feedback and conversations. We worked side by side with multiple teams, including analytics, research, and behavioral science, to close the gap between digital twins and our clients. Our goal was not to create a one-to-one match between the client and the digital twin. Our goal was to create directional confidence and deliver that confidence early and fast. Twins don't replace talking with our clients. Rather, they give us an early insight into what's working and what's not. They allow us to test hypotheses and ideas very quickly so that we can eliminate bad ideas very early in the process and double down on the really good ideas. And in those experiences where we do need direct client feedback, like in-market A/B tests, twins can provide feedback before experimentation starts so that we can provide the best content to our clients.

Vanguard crew member: Who would you invite to a dinner party? What would you serve and who would do the cooking?

Salim Ramji: OK, I'm going to go. I just finished a biography of Ben Franklin. Somebody gave it to me two years ago when I moved to Philadelphia. I figured I should get it done before the 250th anniversary, so I would invite Ben Franklin. The reason I would invite Ben Franklin is that even if you took off his resume, co-founder of the United States of America, he was a fascinating, fascinating individual. And I think he had a wide variety of interests in science and technology and civics. And he was, amongst any of the founders of the country, the most widely travelled, the most worldly. So I think it would be fascinating to talk to him about all the technologies that we have today and get his view on it. I would definitely cook because, as much as I could kind of figure out in the biography, in the 1700s in the United States it was mostly boiled meat, boiled vegetables, like the food did not look good. I would either make him—I’m pretty good at making Indian food. And so within my family, that's my thing. If that's what we're eating, then I will cook it. I think he would find that fascinating too. He lived in France. He lived in a few different places. I don't think he ever had any exposure to Indian food. And so I'd be curious as to his view.

Evan Swartley: OK, Ben Franklin and Indian food sounds fascinating.

Salim Ramji: The other thing that I found out when I read this biography was that he was kind of the nation's first financial advisor. He wrote this book in the 1760s called The Way to Wealth, and what he basically talked about was that wealth and financial independence was in the grasp of every American. You had to work hard, you had to save, you had to kind of be diligent. And it was a lot of the same principles that, if you read any of Bogle's books, it has the same degree of influence. And even if you look at the foreword of Enough, he cites Ben Franklin as being kind of the originator of a lot of these ideas that we've come to know as kind of Vanguard and Bogle ideas. He traced it back all the way to Ben Franklin. So he was also good. He was unlicensed, but he was a good financial advisor.