Chris Nobles leads senior industry executives in a discussion of where AI is producing measurable results in wealth management today and how its impact grows when integrated with automation, data intelligence, and modern client experiences.
Good morning everyone. Appreciate all of you taking this time this morning to join us with this panel. Want to thank Steve for the introduction and SIFMA for creating the opportunity for so many people in our industry to come together and talk about the topics that are most important to us. And obviously artificial intelligence is one of those right now. I don't think there's any firm in the wealth management space who is not being directly impacted by AI. AI is reshaping the wealth management industry probably faster than any other technology wave has that we've ever seen. I think it's safe to say that if your firms are still in a point of conversation about AI and you haven't transitioned to practical applications, some might consider you to be behind. So what we're seeing in the industry today is that AI is touching every aspect of the model, the operating model that Bob was just speaking about from operational back office to advisory experience, investor engagement, and even the pesky governance and risk aspects of our businesses.
So we have a great panel today. The panel is representative of a broad view of our industry. I'm excited about the perspective each of them bring. And so I'm going to quickly introduce and then we'll jump right in. To my left, we have Mazi Bahadori from Altruist. He's Altruist COO. To his left, we have Kristie Edeling Day, Executive Vice President and Chief Information Officer of Advisory Experience at LPL Financial. To her left, Gloria Leo, managing director and head of enterprise solutions at DTCC. And finally, Steve Schillingford, the CEO of DeepSea. So thank you panelists. I appreciate you all being here today. And I'd like to start our conversation around operations. And Kristie, I'd like to ask you to get us going. Where do you think AI has the potential to generate and drive the most operational change from how we get work done, how work is organized and ultimately scale?
So let's start with the core of the traditional wealth operating model. For as long as the wealth industry has existed, there's been the three tiers. It's the home office which serves the advisor, which serves the client or the investor. From LPL's perspective, we've always viewed the advisor as the center of our world and our ecosystem. But if I had to guess what many of our advisors felt, I think they probably feel sandwiched between the home office and the investor and the client. And for those of you who work with advisors, they have this term called the swivel chair where they're swiveling between multiple different tools that they're trying to serve their clients. They're trying to call the home office. They're trying to get different sources of information.
I think forward and I say the wealth advisor of the future and I don't think the future is as far as we might have thought once, is probably the most present, the most attentive, the most empathetic version of an advisor that this industry has always wanted and promised. And the enabler of that is going to be AI. It's going to truly make the advisor feel like they are the center of the ecosystem where information comes to them. What they need to serve their clients is available to them where the home office, their workflows, even if they build these great workflows, they don't end at the end of their office. They actually extend into the home office relationship. Today we're not there. There's great point solutions and I think for those in the audience who are starting with point solutions, that's the right place to start.
But fast forward to 2030, 2032 and beyond, you're going to see a shift. And there's this term in the industry right now called Jarvin's Paradox where the better something is and the more widely available it is, the more demand there is for it. And I think you asked about the acceleration of value and the acceleration of wealth. I think as advisors become better and better at serving clients and more and more human and doing the things that AI just can't, the demand for advice is going to increase.
That's a great point that you brought up. A lot of times we think about AI from the perspective of, how do I rethink the work that's in front of me? But to your point, the flow can now connect end-to-end, whether it be from advisor to back office directly or just still within the back office. So Mazi, what do you think about that? Do you think that firms are thinking that strategically at this point about how to change the workflow even across the different personas?
Yeah, I think it certainly varies by firm. I loved a line that Bob used earlier in his talk of just generally...when you have to approach it looking at it at the infrastructure and architecture layer as opposed to just layering on acquaint analogy, but it's almost as if you put a GPS and a touchscreen on top of a horse and buggy or model-T, you're still not going to go that far that fast. You might have a slightly nicer experience with it. You might have some cool features and capabilities, but fundamentally you're still kind of hamstrung by the speed at which you can go do something. And so where firms are really looking at this at an architectural level, you see that tremendous amount of change and impact. Those that are just layering on pieces, whether it was kind of Copilot or whatever it might be to an existing architecture, you're going to continue to run into challenges that are a little bit too tough to overcome.
The only other point is when you think about the capabilities, there are certain things that people are now able to do that they just previously weren't. And so it's not as if there's all these tremendous efficiencies that are being gained. There's actually new unlocks around quality precision that just didn't exist before when you think about the application of the tools.
Oh, that's a great perspective. And Gloria, if I could ask you to kind of tag onto this, you have a perspective of looking across the industry, given what you hear around kind of the target state, where do you think we have the most work to do still?
I think one of the things that many people are focused on, we spend a lot of time talking about AI and the great opportunities that firms are seeing in terms of the proof of concepts that they're delivering. One of the key enablers is data and in order for us to truly scale in our organizations is a relentless focus on making sure that our data is discoverable, modeled, cleansed, and can be used in a way that the AI models are trusted, are enabled for not just the one use case that we are leveraging in that particular data set for, but those use cases that we've not yet necessarily considered yet, but the data makes it available for us to really start to think about what is the potential in terms of growing that case out as well. So I think that's a real key for us is that the focus needs to be on the data as a key enabler for AI to scale.
Yeah. It's interesting because data sources have been a topic, a pain point in the industry for a very long time now, but there has never been a better opportunity to solve that problem, as what we can accomplish now with AI. And Steve, thinking about what you see as a provider of AI, what do firms who are going to be successful really look like?
Yeah, I think one of the challenges—and I'm in the technology space—is you want to start with the technology and work out because it's really interesting and fascinating. But I think the firms that really embrace of an outcome-in approach where they look at the orchestration or the workflow that they have in their environments today and they know volumes are never going to go down and you can't scale infinitely with Teams or whatnot. If you start with the outcome, you can actually build a nice blueprint back to the tech. And I think we get too distracted by this model, that model, this tool, that tool. But if you say, "Okay, I need to be here in three years, T+1 now in Europe or other regulatory mandates or just my own business model requires me to improve my SLAs to clients. You want to start there and then orchestrate all the way back to the base technology.
And then the technology actually of unfolds. I believe that models, we talk about them all the time, they're all going to commoditize at some point and they're all going to be equally good or bad at certain jobs. So rather than focus on that or focus on a pilot for a specific agent, which by the way, it's not an agent, right? It's a whole orchestration of arguably the world's most complicated workflows. And I think you can't rely on, well, if I just get my data fixed in this system of record, or if I use this agent to pull data, it's really got to be comprehensive. And I think the other thing that people get a little bit challenged with is, we're never going to be autonomous in a regulated environment. That's my belief. I think it's a lot like FSD, right? We take our hands off the steering wheel, but we still are there as a human-in-the-loop to make sure nothing bad happens.
I think AI powered orchestration is going to definitely unfold in the same way and then has to be paired with good governance. I mean, Bob hit on all the right points. You have to start with governance as a first principle in your deployment. Can I understand what was done? Can I roll it back? Can I repeat it? Do I have operational resilience? Those are all the success criteria that I think the most productive firms are deploying right now.
Yeah, those are all very great points and it got me thinking about when you solve those, how do you measure impact? And so Mazi, I'm curious, where are we seeing measurable impact today and what does that look like from your perspective?
Yeah. I mean, I think first it kind of varies obviously by ops organization. We certainly take the view that we always try to perfectly triangulate three variables of efficiency, cost, and quality. You figure if you over optimize on anyone, you run the risk of potentially detrimenting another one. And so if you're in forever pursuit of optimizing those three variables, you're probably going to run a pretty world-class operations organization. With respect to where we see the AI impact right now, it tends to be mostly on the quality front. When you think about, Kristie, I think mentioned one of my favorite paradoxes, Jovan's paradox of the idea being if you make something more efficient, you'll use less of it. But the reality of what we're observing is we're actually using a tremendous amount more of that resource and that's no different in operations.
I think people are finding that with the deployment of these types of tools and these types of capabilities and architecture, you're able to actually go out and do a lot more things that you just previously weren't able to do. And so we're not observing at all, and this is among the 6,100 advisors that use Altrust, this is among our teams internally, that anybody is reducing staff or finding that they just don't need people to be doing things. Instead, they're finding that they're able to do things that they weren't capable of doing before. They're able to do things better than they were doing previously in terms of quality or cost and now they can redeploy resources to areas of their business that may have been a bit more detrimented. And so some of the points that Steve was talking about of just thinking through governance and the implications of so many of these items, if you're busy racing ahead, deploying the tool, you don't have much time to think about it, but if you get it up and running, then you can actually think about some of these bigger picture items and that's what we've observed over the last 18 months or so.
Yeah. I'm curious from any of other panelists, any other areas of measurable impact that have maybe surprised you as you've started this journey?
Yeah. So I'll jump in. I have want to build actually on Mazi's point because what I've noticed is something similar where it's the work that people didn't want to do anyway in many cases that is the easiest work to use AI for. As I've worked with business partners and clients to try to take them on the journey, I've compared it to cloud computing, which I think if I think back to the dawn of that, and I'm now old enough to talk in those terms. The promise of AWS was to get rid of the undifferentiated heavy lifting. So managing infrastructure, managing servers, work that many didn't want to do anyway. And to your point, instead of jobs going away, the investment into the industry exploded. And so for us, we're seeing that in things like, and Chris, you asked about unexpected benefits. One is consistency, which is a form of quality.
So, we're using AI right now and annuity supervision review. And when you take something that's inherently requires human judgment to apply rules, the first review of the day versus the 10th review of the day that you start to get tired, you start to see drift—AI doesn't have a bad Tuesday. And so you kind of have the consistency that comes with outcomes that a client knows that they're going to get the same answer every time and people obviously don't want to do that detailed fine grain supervision review anyway. And so you've got the consistency benefits, you've got the human capital benefits and we've reduced NIGOs as well, so we'll take that too.
Steve, I think you were ...
No, I think Kristie on a lot of points, I think there's a qualitative and a quantitative outcome here. And I think on the quantitative side, to your point about early in the day, late in the day, right? I mean, humans are mistake-prone in this kind of work, and so there's an embedded unknown risk associated with that. That always pops up at the wrong time. But I also think in the world of wealth management, and I just happen to come from the position of there's not going to be an agent doing my wealth management as a client. I just don't see that happening. I want to talk to a human being. I think there's a qualitative ability for advisors to deliver on time. So think about the time it takes for them to get something back from the home office or to set up a trade or whatnot.
I think improving that quality and improving that time actually results in more stickiness, more loyalty to the given firm that they're working with. And I think those are really important measures as we think about AUM and we think about how we want to retain these clients and arguably grow them.
It's interesting when you think about the number of client relationships an advisor can manage today and what does the number look like in the future when advisors have access to better tools and faster response time from home office. That's something maybe we can figure out how to measure over the next four or five years.
Gloria, I'm curious as to...DTCC has been on this modernization effort. How has AI impacted that and has there been measurable changes that you've seen in the firms you're working with to kind of bring that modernization together?
Yeah. I think that for many of our organizations, including ours, where we are seeing AI really demonstrate kind of measurable impact is in SDLC and our ability to actually accelerate our own modernization journey and the timeline associated with that. I think what we're looking at as well is the full end-to-end and we talked about it before, right? When you are bolting on some of these capabilities as one-offs, you definitely get impact, but to get true kind of full end-to-end impact, it is looking at your full life cycle. So starting with the business requirements and building out that component pieces and leveraging the AI to really kind of accelerate requirement building all the way through to UAT testing and your regression testing that's needed to ensure that the code that you're developing has no impact to our clients when we release it into market, I think is quite key on that as well. And so we're excited for that because it does help us to deliver on our modernization much faster.
And that's a very powerful story, the shortened timeline to bring product to market and solutions to your operations teams. One thing that's also clear is that there are challenges with AI and I'm curious to hear the panel's thoughts on what are some of the non-technology challenges. Kristie, maybe you can lead us with that. What are you seeing as, and obviously I think governance and risk is one area, but how do we roll out this capability from a non-technology perspective?
I think one of the biggest risks that we don't talk about enough is the trust in the tool, knowing when to trust the tool, knowing when not to trust the tool. We've all heard the term: hallucination. We've heard the term AI sycophancy where it tells us what we want to hear and the importance of expertise in this world actually becomes more and more important, particularly in the financial advice industry. So to make it non-financial for a second, we've probably all had this experience, but I was trying to remember lyrics from a song and I probably should have used Google—I'll put that out there—but I was wanting to discuss the lyrics of the song and ChatGPT invented a whole different version of the lyrics and now I knew the lyrics of the song pretty well, but I started to get excited because I'm, "Oh, maybe this was a version that was never released." And so I started asking it questions and it totally took me down a path and invented a whole story about how these lyrics had been in dark release and here was the story behind it. And it was completely false, but it sensed that I was excited and it was going to go with me on the excitement.
Now, if you apply that to the financial advice industry and you've got a financial advisor who kind of has a feeling about a financial plan and he's become, or she's become emotionally committed to this financial plan asks AI as a second opinion, AI will pick up on that. And if that advisor doesn't have the depth of expertise or doesn't really understand that the tool is probabilistic, not deterministic, you don't just have a silly story, you have a real regulatory problem. And so making sure that users understand that for the foreseeable future, it's not just from a relationship standpoint. I think FINRA would agree that the financial advisor is the financial advisor and they hold the license. Understanding that AI as a tool is a strong and powerful second opinion, but like a physician who owns the relationship with the patient, they make the decision at the end of the day and trying to bring people on that journey to understand the limits and where they have to have the expertise to be able to say, "Ooh, I don't think these lyrics are actually the right lyrics," I think is a human journey that we have to take people on as we build tools.
That was so spot on. I think being hyper aware of the biases, as Kristie pointed out, a colleague sent me its really great meme the other day that said, "The dumbest person you know is being told you're absolutely right by ChatGPT right now." There's some challenges you got to navigate. I think one of the other things that we certainly observe is this race to action. Everybody feels like you got to go do something if you're not talking about AI in your board meetings, if you're not talking about AI in your management deck or your quarterly business reviews or your monthly business reviews or whatever it might be, you're behind the curve and your board, investors, your management's not going to be happy with you. I think whenever you have that insane race-to-action, you tend to get poor decision making, not surprisingly.
That then contrasts with, you don't want to handcuff or constrain your teams unnecessarily. So we, for example, observe our marketing team is very bullish on rolling out agents. Our ops team is bullish. Our engineering has some ideas, CS has some ideas. We could attempt to centralize things and have everybody kind of go through a formal intake process where you have one firm-wide approach on this, but then you're just layering in a lot more bureaucracy of something that's designed to kill bureaucracy and get you moving a whole lot faster and that creates a challenge. There's no necessarily right or wrong way. Every organization's going to take a different risk-based approach with this, but those biases versus that race to-action and risk tolerance of a company, really, really tough to navigate and not one that the tech itself can quite solve.
There's an embedded concept here that I think is really important and that is like all the consumer driven models that we use are all trained on consumer data, right? They're all trained on Reddit, Wikipedia stuff out there. And I think to your point in the regulated universe that we all live in, you really have to have a governed AI model, which means proprietary data is really important, super valuable, and you want to make sure that you have the proper guardrails so that you don't get the lyrics of a made up song and you don't get the other things that cause really bad outcomes. I think that's another success criteria when you deploy this for use cases like we're talking about.
And Gloria from DTCC's perspective, how do you manage that at scale across so many firms that you're interacting with?
Yeah. So I often get asked the question about responsibly enabling AI within our organization and in the industry as well. I actually think the first pillar of that and component of the way that I think about it is people first, right? And so we have responsibility of upskilling the people within our organization to really understand AI and how they can leverage it for automating the work that they're responsible for. That certainly is a key, but actually a big part of ensuring that we are democratizing AI within our organization is ensuring that they understand they remain accountable. And actually Ken mentioned that word quite a few times this morning as well. Steve just said it. AI is an accelerator. It is not an autonomous decision maker. And I think that that's really key for us in understanding how we employ AI within our organization and with the industry as well, right?
Because people do really need to understand we don't want the made up songs. We recognize that there is a very significant impact of that going downstream into the broader industry in any of those potential use cases that are being used by our firm and by our clients as well. The other thing, and I think it's something that will resonate really keenly with this particular audience is that when we think about new processes that we introduce into our organizations, the control framework and the layering on of dual controls, the way that we think about what needs to be true as we introduce new processes is no different when we introduce a process that has AI that's associated with it. And so in some cases, we have seen instances where that control is a human. In some cases that we've seen that control is another AI model that really kind of oversees and confirms that actually the results that you were looking for are the ones that actually match the potential outcome that you were looking for as well.
And so there's an advancement in a way of thinking about how we implement controls around this as well. I think in the early stages and certainly within our own organizations, within our own organization, I would say we are conservative in nature and ensuring that actually there's in many instances, additional oversight until we can build that trust layer and that assurance that the models are delivering what we expect them to.
Just one caveat, I guess not to totally qualify my remarks or speak for the group, but there's an organization called METR, that...it's a nonprofit funded by the kind of large model providers with the goal of coming up with a relatively standardized approach of measuring the efficacy or the productivity of various models. And given that this is also new, there's no universally agreed on unit of measurement to determine, well, how good is it one model versus another? And so that's what this organization attempts to solve and is doing so with decent traction. They publish a chart regularly every time there's a new model update and it's been out for I think four or five years now and it's exponential. I mean, the way that they're measuring it is to say the average amount of time that an average engineer will take to perform a certain task, how long does it take that engineer versus the model?
And when they first started measuring it, it was something like six seconds and today it's up to, I think, 11 and a half hours was the last one. So that rate of growth, just truly exponential. There's no guarantee that'll continue. Obviously anything could change in the future, but every indication that we have shows that it will continue at that rate. And so all of this could look very, very different in six months if we were doing this panel in this time next year, the answers could be completely different. And so one caveat that it's just a very hard tech to keep pace with right now.
That's a very true statement. That's a great segue into...I was curious to hear from panelists on a year ago or maybe 18 months ago, AI was the experiment, but what is the experiment now? Where are firms experimenting now that AI is becoming a part of our core technology and infrastructure, what's the art of the imaginable, right? What can we do with it? Kristie, maybe you want to kick us off?
Yeah. I mean, so I would say firms should be moving beyond the experimentation phase at this point. I think we said it at the beginning and really starting to think about, "Hey, how do I take what I have learned from the experiments that I have been doing and start to think about how they string together at scale?" And by that I mean, and that doesn't need to be overwhelming, it can be what are the building blocks. So for those of you who have done POCs with point solutions, how do you start to think about how multiple point solutions work together to create a bigger solution and how do you actually, for those of you, most people working and starting to think about agents, how do you allow those agents to begin to work together and to start to produce outcomes and act as opposed to just reason. I couldn't really echo enough some of the comments that were made about the governance need and the guardrails and making sure that as firms start to experiment with how do you actually start putting some of these units together, that things like, I'll use a technical term for a second, observability, which is a favorite of mine right now.
But if you've got multiple agents that are working together that have little discreet responsibilities at each step in the process, what happens if one of those does the wrong thing somewhere along the way? Do your systems or your systems and your oversight systems smart enough if you get an outcome on the other side to be able to say, "Actually this outcome isn't valid because this agent went off the deep end for a while." Those are the sorts of things that I would say firms should be hopefully starting to move beyond, but if not, for sure, experimenting with now to make sure that as the systems that we build with AI start to get more and more smart, that we are getting more and more smart about how we monitor what they are doing to make sure that the answers and the actions that they take on the other side are actually something that's justifiable and that we're comfortable with.
I just want to add to that because I think it's great. In addition to observability, when you start you do an automation or orchestration for 100, 100,000 times, you not only need to have the transparency, but if you've ever used ChatGPT or whatever and asked it about yourself, just tell me about me, you'll be shocked. I'm personally shocked at what it knows about me a little bit, but in our business it can tell you a lot about the business. It can tell you a lot about the hotspots, where you're spending. Maybe you're spending a lot of time with a counterparty that isn't really your most profitable or your best trading partner. Those are the kinds of things that these orchestrations that are powered by AI, you're throwing off all kinds of telemetry, all kinds of insights that then compound your advantage from the original deployment.
So not only are you making sure that the agents are behaving, so to speak, but you're also getting essentially competitive intel about your own business. Those are things that you can use to go exploit opportunities that may be hidden in the past.
I'm curious too, what is a firm who's successful in some of these experimentations becoming a part of their production workflows versus firms that are really struggling to take that last step? What's the difference? What do you see, Steve, that really holds firms back?
Yeah, that's a dangerous question. So I think everybody thinks they can build anything all the time. And I think that's just a natural bias. We're all talking about AI and we're using it for a lot of different concepts, but in our business, in regulatory environments, you have to start with first principles. You have to start with governance. You have to start with resilience. You have to start with an operating model that includes what if AWS goes down or what if my cloud goes down? How am I going to roll back? And every time I meet with a compliance or external regulator, am I going to be able to show not just what the model did but how the whole process was actually transacted through that AI powered orchestration or whatever you want to call it.
I think you've got to really think about building that first because sometimes we get high centered on this model, that model, and there's always a battle at the top, but that's literally the last thing you should worry about. You should really worry about where's your data coming from? How's the model going to be trained? Not just accurately, but precision wise, is my process going? Do I have proper human-in-the-loop? And oh, by the way, do I have proper controls around human-in-the-loop? Because not everybody should be in the maker checker world, not everybody should be a maker, not everybody should be a checker. And then I think just that fundamental, do you have ease of transparency for your own assurance, but also for all the other assurances that we have to provide?
Just to kind of build on it, and Steve had said something at the start of the panel I really, really liked and would super emphasize with that question. Starting with the customer experience first, like what is the actual outcome that you want and then working backwards to the technology. 99 times out of 100 will always lead to a better implementation than you have a piece of tech and you've got a solution hunting for a problem almost that seldom ever works anywhere. And so I think focusing on, what is it that you're actually trying to solve? Is it NPS? Is it CSAT? Is it SLA adherence? Is it response times, resolution times? All of those types of things, I think setting that north star first and then determining if AI is even relevant or not is probably paramount.
Chris, I wanted to just go back to your question in like a lesson learned. Building the capability without adoption does not deliver value to the organization. So many organizations have had very successful proof of concepts and have deployed them to their organizations, but the deployment doesn't necessarily equal the adoption in the organization to really kind of embed it into how teams are operating on a day-to-day basis. And so that's a critical component of making sure that that adoption is there...going back to the process and really understanding what is preventing people from really adopting this and what they do every single day and ensuring that it is getting attraction that it needs within the organization so that you can start to build the end-to-end components. Because if you have a component that is not being adopted by a part of your organization, it's not going to work.
Very, very good point. We've got about five minutes left. I thought we could wrap up with a lightning round. I'm curious to hear from our panelists. AI has created a lot of opportunity. There are a lot of firms that are at different places in their journey. For the leaders in the audience here, what's one thing they should be doing or walk away thinking about doing based on your expertise and vision into the current state and future of AI and our industry? I'll start with you, Mazi.
Yeah. I mean, I'd probably keep it very simple and just say, just learn, encourage everybody on your team to learn, everybody in your organization and not just learn about what AI is and how you implement, but really deeply understanding what are the potential, what are the use cases? How do you actually build agents? How do you deploy them? It's super easy. Anthropic does this phenomenal online course. Anybody can take it completely free. Obviously they're incentivized to get you to sign up with them, but it's still incredibly impactful. I almost think of, I don't know, back when was it, 30, 40 years ago, rolling out the Microsoft Office product suite and eventually within organizations, everybody kind of identified those who were masterful. If you ever needed anything in Microsoft Excel, you had that one person you could go to that knew how to manipulate any sheet and create any formula.
That's almost irrelevant today with so many of these models, you just don't really need that skill anymore. But the ability to build an agent that's bespoke to a certain use case I think is the new version of that and will continue to be for the foreseeable future. So really encouraging everybody on their teams to go out there and just learn how to use these tools.
So mine is stop piloting and start deciding. You can pilot your way into oblivion, honestly, especially in as fast as the technology is moving and it can be overwhelming to make a decision. It can be overwhelming the number of opportunities, but this is a technology where you get better only by doing, and you get better by doing it end-to-end. So I couldn't agree more with the first principles comment because you think about, hey, the flashy, shiny part of the technology, the truth is there's so much underneath. It's all the same stuff. It still matters and it's easy to get distracted by the pilot and just the superficial technology unless you say, "Hey, I'm going to go deep on this one and actually get the thing working end-to-end." Don't be overwhelmed because especially if you pick the right problem and the right problem space, there are adjacent problems around that that you can start to solve and automate as well and pretty soon you actually have a full solution, but it starts with the decision to actually take something end-to-end to production.
I'm just going to pile onto that. Really, it is the focus on the enablers. What are those enablers that are key for developing and accelerating the work in this space? And so, I mentioned it earlier, the focus on data needs to be a critical component of that and that really kind of accelerates the ability for organizations to take it forward.
I mean, what I want to say is, ignore the nerds because in this room there's an amazing amount of subject matter expertise. And so, to extend what you guys have said, I think every firm should have an operating model. You should think about what my three-year operating model's going to be, where are my pain points, what are the outcomes that I'm either charged with providing and the tech is changing so fast that by the time you finish that, it won't matter what you thought you were going to start with. And that's why these pilots, you can be in pilot hell if you're not careful. But I think if you think about it from the, where do I want to end, start with the end in mind, I guess, and then back into how you plug in these tools and how you validate all the checks that we've talked about, I think is the best possible strategy because operations people understand that...they understand that world and I don't think the tech should be the first place you start. I think that's the last place you start.
Very interesting. I think one of the things that I see is, firms cannot quickly just roll out dozens and dozens of bolt-on AI solutions. You have to reimagine your operating model, the operating system that Bob was talking about because there's really a more efficient new way to work to drive scale.
I want to thank the panelists, thank all of you for attending the session today. Very exciting times and can't wait to see what AI looks like a year from now. I feel very strongly it will not look anything like it does today, so looking forward to being here next year. Thank you.
Source: SIFMA Ops 2026