Watch Recording
We ran the session twice across different time zones. While the core presentation and demo were consistent, the live Q&A varied between sessions.
Session 1: GMT, CET, EDT time zone friendly
Session 2: PDT, AEDT, NZDT time zone friendly
Learn More
If you’d like to discuss any of the themes in more detail, see a deeper product walkthrough, or discuss your AI journey, please don’t hesitate to get in touch.
Explore our solutions
Transcript from session 1
Welcome
Tom Chard (00:32):
Hello, everyone. Welcome to the very first global webinar hosted by FNZ. I'm thrilled to be hosting you for this today.
I'm Tom Chard, FNZ's Group Head of Growth. And as I mentioned, really pleased to be presenting a walkthrough of the report that we published late last year, which was the AI-Powered Investment Firm report, which we did as part of a collaboration with ThoughtLab. We're going to be walking through some of the data and the findings from that report, and then we're going to jump into a live demo of some of the AI product that we've been developing here at FNZ jointly with clients around the world. So you get a bit of a teaser into some of the products that we're rolling out with a range of different clients.
I'm really pleased to introduce the panel. So joining me today is John Blackman, who is the Head of AI Product with FNZ. He'll be leading us through the demo. And I'm also pleased to be joined by Lou Celi, who is the founder and CEO of ThoughtLab, who has obviously spearheaded a lot of the research in the report that we published. So, without further ado, I'm going to hand over to Lou to take us through the comprehensive study, the playbook, and some of the initial findings.
Global AI Report
Lou Celi (02:03):
Well, thank you so much, Tom, and a pleasure to be here today. Just so we set the scene, this is a study that we did late last year on The AI-Powered Investment Firm. It was a study that covered 500 institutions around the world to understand exactly what they're doing with AI. And you can see on this screen that it was a very comprehensive study. We covered institutions across the industry, so asset management firms, private banks, hedge funds, family offices, broker-dealers, wealth management firms. Eight different types of organizations.
These were organizations of different sizes from one billion in AUM up to over 500 billion, and we covered 16 countries. So it was a very global study just as today's participants are. So we covered parts of Asia, like Japan and China, Australia, Singapore, and Malaysia. The major markets in Europe, such as Germany, UK, France, Spain, Switzerland, Finland, and Sweden, and of course, in North America, Canada, and the United States. Next slide, please.
(03:29)
So one of the biggest findings is that the AI revolution has begun, and it's sweeping through the wealth and asset management industry very fast. 73% of the respondents to the study, and keep in mind, these were senior executives throughout the organizations, feel that AI is critical for the future of their business, and almost two-thirds said that it will help them interact with their clients more effectively.
Another 63% said AI will revolutionize the wealth and asset management sector. They see it as a game-changer, particularly given the latest technologies like GenAI and Agentic AI, which have the power to not just track changes, but actually help to manage responses. And of course, it even is going to be helpful for optimizing portfolios and generating alpha. Some organizations are already doing that. In our study, we talk about Vanguard and how they are improving results and forecast thanks to AI.
So this revolution is sweeping through the whole value chain. So you can see in the front office with customer analysis and AI for conversational support and self-service portals that are now powered by AI used by over 50% of the firms that we surveyed. You also see it in the middle office for regulatory tracking, and tax monitoring, and of course, data security, and privacy, and fraud protection, where it can be very powerful in finding anomalies, and in the back office as well. In fact, folks are using Agentic AI now to not just edit code, but to actually write code and to manage business processes in a more autonomous way. So very exciting times, and it's changing very fast. Let's go to the next slide, please.
(05:35)
But there are some headwinds, and so the headwinds that we see come in three different places, organizational, technological, and regulatory headwinds. So on the organization front, it's hard for these institutions to make these changes because technology and AI is moving so fast, and they tend to have a more conservative culture. So that is one headwind.
Another one is many organizations don't have a clear implementation roadmap or even the skills and talent in place to drive AI transformation. On the technology side, although institutions are improving their data quality, it's still not exactly where it needs to be to support full AI transformation, and some of their systems and legacy systems are complex and not properly integrated. And on the regulatory side, there's still a lack of transparency sometimes in some of the tools they're using which would need to be corrected.
There is still complex regulatory changes. In fact, there's still some lack of clarity on what some of the regulations on AI will be. So those slow firms down, particularly in a highly regulated industry such as the investment industry. So why don't we just stop there a second? And Tom, I'd like to bring you back in. I know you work with many clients at FNZ. Perhaps you can give us your views on the challenges they face and some of the things they're thinking about to overcome them.
Tom Chard (07:14):
Thanks, Lou. And I think I'll probably touch on both the opportunities and the challenges here. So most firms in wealth management are very, very focused on how do they recruit advisors, bring them into the business, and then how do they enable them to provide a better service to more clients. And big set of challenges that they typically see here are cost to serve, which is a big unlock for recruitment of advisors. And then we're all very, very acutely aware of the advice gap that exists in most of the major financial markets around the world, and obviously, that's getting worse, not getting better.
So I think a lot of firms that are using or are seeing that AI is potentially the saving grace to be able to help facilitate that. Which is why I'd say that the opportunity across the full value chain is really, really important because it's obviously putting technology in the hands of advisors to provide more personalized set of services to more clients than they ever have before, which I think is super important, and that's one of the key opportunities. But also, if you think about how AI can revolutionize the middle and back office, and overall cost to serve, that's another key, key opportunity that executives in wealth management are harnessing.
Just to double down on the points you mentioned around the challenges, I think the pace of change is a really interesting one. You've got here the conservative and potentially slow-moving cultures inside a lot of large-scale financial services firms. Obviously, that's driven by regulatory presence and risk management frameworks. I think there is a dissonance of the pace of change in this technology against that as an obstacle.
It's interesting, obviously. We published this report jointly with you in mid-Q4, I think it was, last year. And by the time now, it's the middle of March, obviously, if we were to run the same set of analysis and process, I think the output would look vastly different, and it would trend even further towards greater levels of adoption and increased ROI, because, obviously, the power of the technology is accelerating at an exponential rate as opposed to... or in respect to many other technologies we've seen before. So that pace of change, I think, is a really critical one, and how that gets landed inside very heavily regulated organizations is a particular challenge.
The other one, I would say, is many financial services firms operate with lots of legacy systems with challenging debt structures or ill-refined data structures. And obviously, having the foundational aspects set up effectively to use data governance and agentic framework are really, really key aspects of being able to build, deploy, and use the technology and product effectively. So I'd say that those are probably the two largest headwinds that we're seeing in our conversations with clients around the world, Lou.
Lou Celi:
Yeah. And I think you're right, Tom. One of the biggest changes since we did this study is what's happening with Agentic AI. It's really taken off, and a lot of the institutions are using Agentic AI to make their workflows more autonomous and using Agentic AI to actually change the way that they even work together with employees. And so there's a lot going on there. And on your point about legacy systems, it's really interesting that what we're seeing now as the state-of-the-art is not just a bolt on AI to a legacy system, but to change that legacy system to a cloud-based system where AI and Agentic is built right into it. And I think that what's happening is... It's moving very fast, but what's happening is that it's companies like yours and others that are building AI right into their system that's helping organizations keep up with it.
Tom Chard:
Yeah, fully agree, and I think that lends itself to... It's a natural evolution from the earlier stages of AI that people were using where the thought process was around, "Help me do this specific part of a process more efficiently," to now effectively need to completely re-engineer processes from the ground up using a different set of technologies, which is, again, a harder thing to do, but obviously, the return on that investment is far greater. But I think that that's the key point, is that there's a process re-engineering aspect now that is also coming into play.
Lou Celi:
Exactly. They're not automating yesterday's work.
Lou Celi:
They're reinventing tomorrow's work. And that's the power of Agentic AI, that you can not just do things more efficiently, but you could do things you never could do before. So I'm excited to see some of the ways that you're doing at FNZ later in this presentation.
(13:07)
Why don't we move on to the next part of the results, which is the lessons from the leaders?
So just to let everyone know, as part of this study, we analyzed where all 500 firms were in their AI journey, and we classified them based on a framework that we developed with FNZ. And so 23% of the firms that we surveyed were what we called starters. They were just planning, starting to run pilots, maybe a few isolated use cases, but still early days for them.
Then, there was this bigger group of organizations we called advancers that were all ready, did some pilots, experimented, were moving firmly into the execution stage, and building an AI foundation, which is important because you're not going to get very far with AI if you don't have your data and your IT platform in place.
And then the third group, and this is a group to watch, were the leaders, 21%. So it's a smaller group, but these are the ones that are first ahead, the ones to watch because they're showing you where the hockey puck is going. And they have already built a strong foundation, and they're scaling AI very quickly across their enterprises and actually, becoming more of a AI-first organization.
And there's a reason that they're racing for this, and that's to gain a competitive edge and to drive performance. So 74% of the leaders are generating moderate or large returns from AI more than other firms in the study. 56% are accelerating time to market from AI, and another 55% are even seeing greater shareholder value from it. So, clearly, AI will drive performance as you can see from those numbers. Let's move to the next slide, please.
(15:14)
So we found five best practices, five things that leaders do that differentiate them and that allow them to unlock greater value.
(15:29)
The first is they build an AI vision and culture to inspire change. So one of the things they do on the strategy side is one of the first things they do is they align AI strategy with business strategy. They recognize that AI is not a technology strategy, it's actually a business reinvention strategy, and so they tie that together. And they also develop and communicate a top-down AI vision across the organization. But in addition to the strategy, they put the right culture in place because you're not going to get very far if your staff isn't supportive. And so they work closely with their employees to communicate their plans and to provide them with the systems and tools that employees need to start experimenting.
And we're in a good place because ChatGPT has made it very clear to lots of people the value of AI. So these employers are using AI in their own... at home, so they're now experimenting and using it much more in the office. And they also build a culture by building an ecosystem of partners to support AI plans. So, in our study, we talk about HSBC who partnered with the University of Toronto and other startup companies in Toronto to build their Innovation Lab. So that's very important because not all organizations are going to do it alone. So that's the first best practice.
(17:06)
The second one is they make sure they have an AI-ready IT and data platform. So 87% of leaders have made considerable progress on building a modern IT platform. So one that's more integrated, that breaks down the silos, which is very important if you're going to get the full value from AI. And as I said before, they have cloud platforms with built-in AI capabilities, and they even build some of their own add-ons that fit in seamlessly with that.
You want a platform that has the AI built in so you get full orchestration, and that's what the leaders are doing. And then there's three top data strategies that leaders all take. One is they integrate data. So they break down the silos. I know that's hard, but that's very important. They very often put in scalable data lakes, warehouses, or data fabrics to bring it all together. And importantly, and this is critical, they take a security-first attitude. They make sure that their systems are really secure.
(18:20)
If we move to the third best practice, you'll see it's about governance. Now, this is critical, because if you don't have governance in place, you really can't accelerate your AI strategy. You have to be very careful with AI, particularly in a sensitive area like in wealth and asset management. So 81% of the leaders build AI governance frameworks. They have policies and guidelines in place. They even do diligence on the third-party projects and products that they bring in house. That's important because many of their systems are combinations of various systems from vendors.
They do robust AI testing and auditing, which is extremely important too because they have to make sure that what comes out of these AI models is something that's reliable. So those are just some of the things they do, and they do a lot of early dialogue with regulatory bodies. As Tom was saying, it's hard to keep track of regulations. They're changing rapidly. So staying in touch with regulators, really important.
(19:34)
Let's move to the fourth-best practice, please. Yeah, this is very important. They prepare for the future of work. Now, leaders understand that at bottom, it's not a technology change. AI is a people change, and it's going to change how people get things done. So, very often, leaders have a vision of the future. For example, 73% say AI will drive a step change in human productivity. It's a major shift, probably a bigger shift than we've seen for many years.
The other thing is they recognize that AI will create new roles and reinvent current roles. So, based on that, they are taking steps as you can see at the bottom of the screen, and the steps they're taking is they're building internships with the universities and apprenticeship programs. They are institutionalizing strategies to attract, hire, and retain AI talent. And of course, they're providing ongoing training for staff, which is very important. And not just general training, AI training for the actual task that's being managed.
(20:54)
Can we move to the next slide, please, which is the final one, which is Rethinking Business for the Agentic Age? And that's important, because as we said earlier, things are moving so fast. And Agentic AI and the latest AI, like multimodal AI, and even quantum AI, and small language models, and explainable AI, these are types of AI that are being used now by a minority. They'll be growing fast, and they really are going to change the business in a material way.
You see on this slide that GenAI already used by 41% going up to 75%, and these are numbers from last year. I'm sure they're much higher even now. And you can see that Agentic AI and GenAI are being used across the value chain once again. So they're making conversational support much more interactive and personalized, and they're changing up CRM to make it much more active across the whole customer lifecycle.
In the middle office, they're using Agentic AI and GenAI not just to track risk, but actually, to close them in real time. So a lot of cool things happening. Also, in the back office, we mentioned Agentic, the new Claude version. You can write code, write with it, and they're already doing a lot of that as you can see there. So it's a big change, and it's something that institutions have to be prepared for.
(22:45)
So I want to stop there and just see if we can talk a little bit about these five best practices from FNZ's point of view, and how they see these best practices developing within the companies they work with, and how they align with, I guess, the way you do things, John, at FNZ.
John Blackman (23:05):
Thanks, Lou. And hi, everybody. It's really interesting the first point there about creating the AI vision. I think when we started working on this research, I don't think at that point, everyone in the firm has such a ready interest in using tools themselves. And where we find ourselves now is that everyone... it's a change that everyone is expecting.
So, for instance, when we announced the AI Champions program within the firm, we had hundreds of volunteers. So it's a change that everyone is anticipating that we will bring in, and the way that we've done it is we've... I mean, aside from the points that you made about generating code and actually using AI in our delivery process, we're also using AI in all of our product development, we're using it in our operational efficiency programs, and of course, everyone is using it in their day-to-day corporate lives.
And in order to do that, and I think this is one of the key pillars there, is that you have to have a robust governance process in order to introduce this into a large, regulated entity. So that combination of having a governance process that everyone understands, layering that on top of a data model where we've clearly permissioned and introduced data into the whole organization, these are the bedrocks of how you can roll AI out through the organization.
Now, quite early on in this process, so over six months ago, maybe a year ago now, we developed a very strong partnership with Microsoft, which has helped us a lot in that governance and data model rollout. They had a lot of experience already. We've worked closely with them to really embed some of our core functionality. That enables us then to address your next point about how we rethink business processes, because we're not using AI just to fix broken pieces of the operation, we're using it to actually rethink the jobs that everyone does.
And if we look at that through the role of a wealth manager, through the role of an advisor, you can see that you can create for them a whole agentic framework that they can use to run their daily processes. So we'll do that for advisors, we'll do that for ourselves within our operations, we'll do that in our corporate teams. So maybe, Lou, maybe we should think about moving on a little bit and showing a little bit of the software that we've developed.
Live AI Demo
Tom Chard (25:52):
Why don't we jump into the demo, John? So we've obviously spoken quite a lot about the products that we're developing, particularly in the context of making advisors more efficient, offering more personalized service at scale. Let's jump into a couple of examples.
John Blackman (26:09):
So thanks, Tom. So what you can see here, this is our standard FNZ platform. It's a microfund and built up in many apps. And the very first thing that you'll see here is that we've got the advisor's homepage, and we used to get super excited about our in-depth library of widgets and how we were expanding them. But AI has blown all this away, right? So we can deliver to the advisor's homepage the advisor's dashboard. We can deliver a configurable experience based upon prompts which they have invented, if you like, themselves.
So this panel at the top, this AI Insights panel, this is... Individual advisors can put their own prompts into this framework. So maybe if I show you a little bit how that works. So as I go into our AI agent, let me share a few things with you. We've got some standard prompts. Now, these are prompts which FNZ has run thousands of time through our own Evaluations Framework, and these are prompts which the advisors can just use themselves. They can run these whenever they like, but they can tailor them, and they can take it a bit further forward, and they can have their own prompts, and they can save prompts.
Now, what I was just showing you before is that some of their prompts, if they want to, they can flag those to be shown on the dashboard when they sign on. So any of this prompt library, they can display on their home dashboard, making it immediately configurable.
Now, what you'll see also in this section, these are their skills, if you like, right? So I've got a client review one here. So this is a detailed prompt that I've created as an advisor to describe a particular task that I want to do. So, in this case, it's a client review. And I've set up a library of these skills. These are things that I would use on a regular basis.
I've got one here, this quick retirement calculation, and perhaps I'm particularly proud of this one. So if I want to, I can share this with other advisors within my network. So it means that all the effort and time that I spend working through these, I can then share with other people so that we can actually have a common set of... These are almost like markdowns, right? You can have a common set of skills, which we can provide back to the model.
But then, the next piece of it, and this goes back to your point, Lou, about reinventing how we work, is we can start to create an agentic framework for the advisor based upon those skills. So I can create a schedule prompt. So, in this example, I'm going to go back to my prompt library. I've got a simple one here that just looks for clients that are overweight in cash.
And I can tell it to run this at the beginning of every week and to create me a list of those clients. So let me just check that that's all going to run correctly. And then maybe what I'll do is I'll create workflow from that. So that's going to run once a week. It's going to find all my clients who are overweight in cash, and it's going to create me a workflow task. So, Tom, I think you can see that already, for an advisor, I can start to build out a framework to make their lives more efficient.
Tom Chard (29:30):
Yeah. Absolutely, John. Quite impressive. How do you see the evolution of the interaction between advisor and platform, and this hybrid experience now between connecting via agents versus connecting with traditional platform experiences?
John Blackman:
So the next example I'm just going to show you is one where we have the... As an advisor, I'm not only using my save prompts, but I'm interacting with the platform on a round and round basis. So if we start off, let's just prepare a client snapshot for Thomas Wood.
So what we've built at FNZ is an agentic backend for the advisor that brings together various different guardrails, different regulatory structures, ensures that the questions that I'm asking as an advisor are not outside the boundaries of perhaps, "Should I be giving advice, not giving advice? What should I be doing?" And we've provided connectors to our agent back into our platforms, back into research, and so on and so forth.
So you can see that I can very easily get a nicely-formatted client snapshot. I can download this straight into a PDF if I want, so I could maybe do a few more terms, and change the format, and just download it into a fairly standard PDF format. But I might also want to dive a little bit deeper. So I had an example that I prepared here earlier where I was saying, "Okay. Well, what about if I actually want a detailed forecast for this client? He's coming to the office. He wants to talk about the future."
And so without having to download all the data into Excel, without having to do a lot of work, if you like, I can just get it to produce me a forecast. It shows me the numbers, it shows me the calculations, and it produces me a graph that I could then cut and paste into a PowerPoint, or I could take this whole section, unload it into a PowerPoint.
And we know from our experience with advisors, because we're putting this software live with one of our clients in Australia right at the moment, we're going through Advisor Labs. We know that this process that I just showed you can take three or four, sometimes five hours per client for an advisor to bring together all that data. So that interaction is already going much, much deeper than perhaps it used to, and it's saving advisors an enormous amount of time.
Tom Chard (32:10):
Yeah. I think that's phenomenal, John. And in some of the analysis we've obviously done, we're seeing advisor time saved in simple tasks, like meeting preparation and simulation analysis on certain activities in their portfolio, multiple, multiple hours, which we're obviously seeing is a key unlock for advisors being able to provide deeper services to their existing clients and also scaling their practice to support more.
John Blackman:
So the other thing, Tom, I think, which is super important, is that we are enabling advisors to manage more end investors, right? So this is serving two purposes, really. One, it's making advisors more efficient, enabling them to be more successful, if you like. But it also means that there are end investors for whom advice would otherwise not be available that are actually getting access to some kind of service from the advisors.
So let me just show you another example. If you remember, we ran a report from our agentic model looking for excess cash. And in this particular example, this is a UK example. And what we're suggesting for these smaller cash balances is that they should transfer them into an ISA. And normally, this would take quite a lot of time for the advisor to go through the process. But what we've done is we've created an Agentic model which allows us to automate this process.
So let me just show you how that works. So, in our example, the AI agent, if you like, is called Kai. And in this particular example, we've noticed that Walter has a cash balance, an uninvested cash balance in his general account, which we think he should put into a tax-wrapped account.
Kai (34:00):
Hi, Walter. It's Kai from FNZ. I've noticed you have a healthy cash balance with us. Have you ever thought about making some of it work a bit harder through a stocks and shares ISA? I can walk you through how it works. This would be information only, not advice.
John Blackman:
That sounds really interesting. Could you tell me a bit more about it, please?
Kai:
Absolutely, Walter. I'm glad you're interested. A stocks and shares ISA lets you put your money into investments like a ready-made portfolio, and any growth is tax-free. Now, the one we're talking about is the Ready-Made Portfolio 3, Moving On Up. It's a balanced mix of shares for growth and bonds for stability, designed for at least a five-year horizon. It's moderate risk, so the value can go up and down, but the aim is steady long-term growth. Remember, this is just information to help you understand how it works, not a personal recommendation. Does that make sense so far?
John Blackman:
I'm interested, but can you maybe give me some more details so that I can come to a conclusion? Have you got a fact sheet or some details about the fund?
Kai:
I've opened the Fund Fact Sheet and Key Investor Information document for you. Take your time to have a look through the details, and let me know when you're ready to continue.
John Blackman:
I've read all the documents. I think it looks like a great idea. What should I do next?
Kai:
Great. Now that you've reviewed the documents, the next step is to check your eligibility. I just need to ask, have you paid into another stocks and shares ISA this tax year that-
John Blackman:
No, I haven't.
Kai:
Got it. In that case, you're eligible to open this ISA with FNZ as long as your total contributions across all ISAs this tax year stay within the 20,000 pound limit. Now, let's move on to the investment amount. Would you like to make a one-off payment, set up monthly contributions, or both?
John Blackman:
I'd like to do a one-off lump sum of 5,000 pounds, please.
Kai:
Great choice. So you'd like to invest a one-off lump sum of 5,000 pounds. Just to confirm, that's 5,000 pounds as a single payment, right?
John Blackman:
Actually, let's do 2,000 pounds, and can I use Apple Pay for that?
Kai:
I've opened the Apple Pay interface for you to complete the 2,000 pound payment. Just follow the prompts on your device and let me know once it's confirmed.
John Blackman:
It's all done. I've confirmed that.
Kai:
Perfect. Your 2,000 pound payment has gone through. Now, let me quickly recap where we are. You're opening a stocks and shares ISA with the Ready-Made Portfolio 3, Moving On Up, and you've invested a one-off amount of 2,000 pounds. Remember, this is information only, not advice, and your capital is at risk as the value can go up and down. Next, we'll go through the final terms and declarations. Are you ready to review the key terms and conditions now, or would you like a quick summary?
John Blackman:
Just give me a quick summary. I think I'm happy.
Kai:
Of course. Here's a quick summary. By proceeding, you're agreeing to the FNZ terms and conditions, the Key Features document, the costs and charges, and the ISA tax year declaration for 2025-2026. You're confirming you're over 18, a UK resident, and you won't exceed the 20,000 pounds total ISA limit across all providers. And remember, the decision to invest is your own, and no advice has been given. Does that all sound good to you?
John Blackman:
We can stop it there, Tom, but I think you can see how that kind of technology could enable an advisor to just increase the amount of business that they can get through on a regular basis. And all but that last human in the loop piece can be completely pre-scheduled and is kicked off by events that are happening within the client's portfolios.
Tom Chard:
Yeah, completely agree. I mean, it's absolutely amazing, John. And as we think about some of the changes that are coming through in a variety of markets, like targeted support or other variations on guided advice, you can clearly see how extrapolating that interaction model that you just demonstrated for us could be used as the key delivery mechanism for a range of different workflows, or solutions, or engagement with clients without requiring huge amounts of advisor time.
Live Q&A
(38:40)
All right. Well, thank you for that demo, John. That was fantastic. Every time I see that, it puts a big smile on my face. I think it's just awesome. So why don't we move into a few minutes of Q&A. I can see we've got a handful of questions that have come through. I'm going to kick us off with a pre-submitted question that we had sent through prior to the start of the webinar. So, John, maybe I'll start with you, and then, Lou, if you can give some industry context on this. What is the change that people have observed in productivity levels, and how has this been converted?
John Blackman:
I mean, so from my own personal experience, I mean, really, this was the Advisor Labs that we were running when we first rolled out the software, and it just saves so much time. Although our screens that we're delivering to the advisors have all the information that an advisor would need to prepare for a client report or whatever it might happen to be, being able to super quickly collate it into a predefined report ready for a client meeting just saves so much time.
And the really exciting one that's coming next as we've just plugged this in with our partnering with Microsoft is we've plugged our agent into either Excel or into PowerPoint. So you can have a conversation within the framework of PowerPoint saying, "Okay. So I want to prepare an annual review for this client. I want this heading. I want this summary. I want this content." And as you're talking, it just brings it all in. It just saves hours, and hours, and hours every week. And we've built into the platform a tally, so that every time we run it, we say, "Well, this has just saved you three hours. This has just saved you four hours." So we get that feedback loop.
Lou Celi:
Thank you. So just to add a few things to that, first of all, what we have found is that currently, a rep or advisor is saving up to a day, a week in their time within those organizations that were seen as leaders. Okay? So the leaders that are pretty far advanced are already unlocking a day's worth of productivity.
Now, what we found that people are doing with that time is they're using that time to do what they do best, be humans, and to make a human connection, and to be thinking more about how they're going to manage this data and this account for the attitudes that their client have. So it allows them to do more higher value work.
The other thing I would say is that in our study, we found that 60% of leaders are already unlocking productivity, and there's another 26% that are going to be unlocking even more productivity going forward. And the one thing to keep in mind that I think is important is that productivity doesn't change overnight. There were some studies from MIT and others that said, "People using AI, but they're not seeing ROI or productivity. Why not?" Because it takes time.
It takes time to get that productivity, and you have to be thinking... Because productivity comes not just from cost savings, it comes from when your revenue actually increases or when you have more per revenue, per head. So it takes thinking through the roles of people and what they would do, and lifting the revenue when you get the real productivity gains.
Tom Chard:
Yeah. Awesome. Thanks, Lou. And maybe just doubling down on that point, there's a question here from Semesh. So, Lou, of the 21% leaders and 56% advances, what percentage of those are achieving their estimated ROI? And again, going back to that point, it may be lower than anticipated right now, but they're playing the long game because they know that they've got to build the utility and infrastructure before the ROI is realized. It'd be great to hear your perspective on that as well.
Lou Celi:
Yeah. No. First of all, I agree with that point. Our study found that on average, it takes 22 months to get to payback, and payback is just breaking even, and then the ROI comes after that. So it doesn't happen overnight. So that's why MIT didn't find the ROI so high a year ago, but that's changing.
The other thing I would say is that not all use cases are the same. Some are going to deliver higher ROI, and that's important for you to keep in mind. So, for example, in our study, the leaders were getting the highest results, high or very high ROI from things like writing and editing code. I'm just reading off my list here. Client reporting and financial statements was an area of high ROI. Of course, they're getting high ROI from robo-advisors in the front office, and their conversational support is working very well to achieve their ROI. Portfolio management. In the middle office, predictive and scenario analysis is showing very high results for them and as is data security, and risk and fraud protection. So these are some of the early use cases leaders are using to really drive ROI.
John Blackman:
But, Lou, we talked about this before, right? In this particular technology, I think what makes it different is the pace of change. So even since the research then that you did, whatever it is, four or five months ago, I think I'm really hearing in some of those areas that you've picked out. They're really resonating with me that two or three months ago, I would've said, "Oh, that's not working particularly well." And now, it's just absolutely flying, particularly with code writing, right? In the last two or three months, we've seen the capability of the code writing has just gone through the roof. So I would say that the ROI will accelerate.
John Blackman:
I just don't think in the use cases that I'm seeing, it's going to take us 22 months.
Lou Celi:
Yeah, it may be that it will take a shorter amount of time because these new technologies like Agentic AI are so powerful in what they do, and so you're right. But back when we did the study, that was what you had to keep in mind, that it wasn't going to come overnight. Well, nothing does, right? You invest, it takes time. Even if it's not 22, maybe it's compressed. What would you say now, John? When would you get the-
John Blackman:
Honestly, particularly on our operational efficiency, I mean, it would be definitely sub 12 months. I mean, our targets are way less than that.
Lou Celi:
Okay.
Tom Chard:
I think that the converging phenomena here is that the partners that we're using to help support a lot of this, so Microsoft, Anthropic, whomever else, right, that the pace at which they are delivering new models, and capabilities, and features, each new version is delivered faster than the previous one to the earlier one, and the capability is far, like it's a big leap. So you have this genuine exponential trend in the technology which has not been seen before. So I do think that that will shift left the ability for those to realize ROI. But again, it will be the leaders who have already set up the right framework for their organizations to consume this appropriately that will be the beneficiaries.
Lou Celi:
Yeah, I agree with that. And I think when we were doing this study, a lot of people were still experimenting and doing some of their own AI build-outs. But I think what's changing is that the ecosystem like you and others are coming with out-of-the-box solutions that can be put into place very quickly.
Lou Celi:
Like Microsoft, like you said.
John Blackman:
Can I pick up, Tom? There's a question from Mo, hi, Mo, that says-
John Blackman:
How do you balance governance and controls without impeding development and trial by learning to see what's actually possible? And this one is super close to my heart because we really want to innovate as fast as we possibly can. And so within our governance process, we've expressly created the space to allow that to happen.
So, for instance, if I want to build some agents that I want to test in my own personal workspace, that's one approach, that's one framework that I have to go through a process for. And then if we want to roll it out more generally, it's a completely different process, but I allow people to ideate within their own world using various different models, different agent builders, Copilot Studio, and so forth. So on a fairly small blast radius, we're allowing people to do really quite a lot, but then have a separate governance process to embed that into FNZ's core functional approach. So having that two-speed process baked into the regulatory governance framework is really important, because otherwise, the whole thing comes to a halt.
Lou Celi:
Yeah. And what we found is a lot of sandboxing, like you said, and also, it's very hard to understand some of the regulations because they haven't been released yet. So what a number of institutions that I'm working with are doing is they're thinking through what would be future best practice. So transparency and results clearly going to be an issue for the regulators. Obviously, privacy and security. So they're building that in.
Tom Chard:
Yeah. And John, so just moving on from that, there's a couple of questions here, one from Christie, around, "Any chance of hallucinations? How can you make sure there are no errors in the overviews and analytics?" And then there's another question on a similar topic that's just coming from Johan, "In an agentic workflow where AI performs automated compliance reviews and client briefs, what specific guardrails are implemented to detect and mitigate the risk of hallucinations before the output reaches the advisor or client?" So maybe if we can talk a bit about the evaluation framework and how we have our inbound and outbound controls inside our agentic framework, I think that would be helpful to speak a little bit about.
John Blackman:
I mean, also, maybe just going back to Mo's points as well a little bit. So when we first invented the concepts of the software that I've just shown you today, we were effectively taking the data and putting it into the knowledge context of whatever LLM and just asking the questions. And we were getting good results, but they weren't repeatable. And to your point, Tom, there was some hallucination going on.
So then, with our partnership with Microsoft, we've built this advisor agentic backend. And so now what's happening is this is all going through a pretty sophisticated agentic structure where we have a regulatory agent, we have guardrails, we have anti-hallucination. I don't know exactly how many agents are in our backend, but it's a lot, and it's complicated, but the output is therefore pretty deterministic.
And then what we did was we actually used the same process to build an evaluation engine so we can run these same things through thousands of times and produce a properly documented, evaluated analysis of what we did, what the... Because the thing is the answers aren't binary. It's not right or wrong here. And so you have to have a grading of whether you think it's an acceptable answer. And this comes back to your point about best practices, I think, Lou, is this is all part of the learning. Particularly within a regulated entity, there has to be a framework between the users and the models in order to protect against exactly Johan's point and Christie's point.
Lou Celi:
Yeah. And, John, I think what you're saying about having multiple agents is an important one because one agent doesn't do it all. You can now have teams of agents. So one agent can just be focusing very heavily on compliance or on risk, and so they can be doing different tasks and working together just as people do.
John Blackman:
Exactly. And what we've discovered, this is even more recently, is that really think about it, think about the skills rather than the agents themselves, right, because the frameworks that we're now being delivered spin up the agents themselves depending on what needs to happen. It's all about segregating the core skills required to ensure you get the right output.
Tom Chard:
And maybe just time for one or two quick more questions. So another one here from Alan, which I think we've partially covered, but there's a specific bit towards the end of the question I think would be helpful to answer. "Can the advisor that we've just..." or, "Can the advisor set a risk profile and boundaries for the agent to recommend solutions? And does the AI advisor learn from the decisions taken by the advisor in how they interact with it?"
John Blackman:
So, to the second point, we haven't done that yet. I can imagine how we might do that. But to the first point, configuring the boundaries of the model effectively of that Agentic framework to giving it risk parameters in particular and guardrails within which it should make suggestions, give advice, even do summaries, that's all configurable in the backend by the advisors themselves. So it's not something that we have to code up separately. These are all part of the skills that I was describing. The analyzing the human in the loop process and working out whether a particular advisor, whether they've accepted the answer or not, we haven't built that in yet, but you can easily imagine that would come next.
Lou Celi:
And people are doing that. And that's the cool thing about Agentic AI is that it does learn, and it can learn from behaviors and change its own course of action. So, yeah. For sure. I'm sure you're going to be doing that soon, John.
John Blackman:
What we have found though, to be fair with our primary big client we're working with down in Australia, was that they would actually prefer to put some boundaries around that that don't get altered through the behaviors of the advisors. And this was a point you brought up right at the beginning, Lou. The rationale and the explainability of all this is really important. In the examples that I demoed, I switched off the detailed explanations and rationales of what it's doing because it gets a little bit tiresome, but it's really important that it's there. So, in all the logs, you can go back through and see how it got to the conclusion that it got to.
Lou Celi:
Yeah. I guess the lesson there is right now, people want to keep a human in the loop. That's not changing yet. Although, I guess over time, that could change. At least Agentic AI will have the power to do it.
John Blackman:
Well, there'll be some use cases perhaps where it does, right? Absolutely. It's certainly got the power to do it.
Tom Chard:
Last question, and then we'll wrap up. "With the pace of change and rapid adoption of AI tools and capabilities, what is this secret source or competitive mode that firms can deploy? Essentially, is adoption of AI going to become table stakes, or can it create sustainable competitive advantage? Keen to hear both of your perspective on that."
Lou Celi:
There's the first-mover. I am seeing right now that those who are moving faster are gaining competitive advantage. They're gaining even shareholder value. I think I honestly believe that, and the study is showing this, that AI is a true game-changer for the industry. It will disrupt the industry. The industry three, five years from now will be very different from the one that we've been experiencing. And those organizations that think through their business and say, "If I had started as an AI business, how would I organize this?" and then think how they would put people in to make it even more effective. The ones that actually start thinking AI-first are the ones that are going to get the highest lift in performance and are going to change their market share and competitive position. So I think the answer is yes, it can, but it takes thinking not just about using AI for efficiency, but actually for transformation.
John Blackman:
I completely agree, and I take you back to your best practices. Data and governance will enable firms to have a long-term sustainable edge over firms that don't embrace that.
Tom Chard:
Awesome. Well, thank you both. Why don't we wrap up there? Just wanted to thank you all for joining today. Hopefully, you enjoyed the session as much as we did. Massive thank you to John and Lou. That was a lot of fun. And to all the guests today, we'll distribute an email with the recording and presentation shortly. And if you've got any additional follow-up questions or you want to get in touch to see more of the product that we've demonstrated today, then please feel free to reach out. Really look forward to speaking to a lot of you on your AI journey. So we'll leave it there, and thank you very much. Cheers.