MARK BROWN: Financial modelling is about getting yourself to the stage where you can understand the risks associated with your business, understand the capital requirements of your business, and understand the behaviour of your business.
SPEAKER: You're listening to Talking Tech, part of the Rethinking Insurance podcast series from WTW. In Talking Tech, we explore the wide range of technology challenges facing insurers from AI and data science to open source solutions and cybersecurity, with a focus on how we help insurance companies tackle these issues.
CHARLIE SAMOLCZYK: Hello, and welcome to Talking Technology. I'm your host Charlie Samolczyk. And in Talking Technology, we explore the wide range of technology issues facing insurers, from AI and data science through to open source solutions and cybersecurity. And we look at how we are helping our clients to tackle these issues.
Welcome to another edition of our Talking Tech podcast. In this episode, we will be talking about all-things financial modelling. And I'm very pleased to be joined by two of my colleagues. We've got Mark Brown, who is our global proposition leader for life financial modelling; and David Pond who is a director and longtime practitioner in our London office. Welcome, both.
DAVID POND: Thank you. Yes, great to be here.
MARK BROWN: Great. Thanks, Charlie.
CHARLIE SAMOLCZYK: So, today's podcast, as I mentioned, is about financial modelling. I think for our listeners out there, it would be probably a good place to start would be just talking about the state of the market and maybe a bit of scene-setting from your perspective as to what's going on and what are people up against in this space right now.
DAVID POND: Yep, I can kick off on that. I think there's a-- obviously, with the likes of IFRS 17 and other big financial requirement changes over recent years, there's been a lot of very narrow, blinkered focus upon just getting over the line for IFRS 17 and other things around that and not having had time to think about the wider picture and the financial reporting in the wider sense.
I think we're now seeing as people are beginning to draw breath after IFRS 17, and partly for seeing the additional reporting requirements they have on IFRS 17, a lot of people are saying, hang on, what more do we need to be doing? We need to be changing our setup here, finding ways of getting more for less out of the number of runs we need to be doing. And maybe the status quo we have and we've had for a number of years is no longer good. And perhaps, we should be looking around to see what has changed.
In the intervening time, and most of the clients we work with seem not to have done anything in this space for maybe a decade, and so are actually quite behind the curve in terms of what they have and what they could be getting. So there's certainly a lot of interest we're seeing in discussions about what can we do to catch back up in terms of getting more for less out of our systems.
MARK BROWN: Yeah, that's interesting, David. We did a survey about a year ago, asking people where their focus was on improving financial reporting-- interesting results off the back of that. But one thing that came out, possibly off the back of having, as you said, spent 10 years fighting to keep up with regulation, the areas, the clients, and prospects were looking at improving were very much those that they felt they understood the most, the ones they were most comfortable with, and not necessarily where the risks and costs were.
And that contrasts quite heavily because what we've seen from the vendors and the providers is there's a lot of new technology out there in the market that's very much focused on helping people hear, very much focused on making the reporting experience faster, better, and cheaper; but getting people to the stage where they can get their results with fewer errors for less cost and in a shorter time period.
CHARLIE SAMOLCZYK: And I guess, Mark, so lots of potential. And it sounds like, David, there's some, I guess, pent-up-- almost pent-up need for people to innovate in this space. From a life reporting perspective, what does innovation mean?
DAVID POND: Yeah, it's an interesting question in the fact that obviously people always talking about the most latest cutting edge. Anyone you ask innovation, they'll say AI and look to the future.
But actually, from my perspective, I think innovative is things that happened over the last 10 years, rather than things that might happen in the next five. So there are plenty of additive improvement changes that have happened in the systems over the last decade, which most of the big insurers haven't yet had time to incorporate. So I would say to that, yeah, innovative means picking up on the tried and tested solutions to improve things that you don't yet have, rather than some totally untested out-there potential that may seem exciting and more interesting on the face of it.
But when it comes down to it, it is unproven and unknown whether it will help you. And I think we'll see in the next few years a divergence of people-- those who've been almost seduced by this idea of there's some magic bullet coming around the corner, which will solve everything-- spend lots of money to not actually get to working solutions. And on the other side, those who will actually accept the fact that look we're really behind the curve in just the tools that already exist.
Let's just go with tried and tested things, and I think they will see a lot more improvement and catching up to a better state by the things that maybe don't seem so exciting on the surface of it, but are a lot easier to justify when you actually lift the lid.
MARK BROWN: I think you're right there, David. I mean, we all know and we all talk about the technology adoption life cycle-- the fact that at the forefront, there are the innovators and the early adopters looking for the latest technology to give them that absolute leading, bleeding-edge experience.
Behind that, there is a big pack of companies where there are still a lot of potential improvements out there. That technology or changes to processes can deliver. I mean, if we just look at what's come out in the last few years-- so we've done things like risk agility financial modeler has its new generation, second-generation calculation engine. And we've seen companies picking that up on the liability side, averaging 10 times performance improvement. That's a 90% cost saving.
Clients are looking at equally improving their model development side and picking up things like RAF and team edition. The focus there very much around getting the best collaboration between the team, making sure that everybody that's in the development team, from the most experienced to the most junior, can contribute.
And the recently-launched V test product that does the automatic regression testing so that you don't have errors creeping into your development cycle that go unnoticed when you're on top of the benefits that you receive when you make the take your model development and put it into production. I think the real thing, though, is just being aware that that is a big adoption life cycle. That there are people at the bleeding edge.
But there's a huge amount to be gained for those companies that are still catching up that aren't using a full set of technology, maybe have processes that were implemented five, 10 years ago or end user computing. So there's definitely lessons there. Innovation can be everything, as David talked about, from the bleeding-edge innovators through to the main pack of people who are potentially out there to pick up the technologies of the last few years to improve their business. And then finally, just an awareness that if everybody keeps up-to-date, there's continual benefits just from keeping fresh with what you've got.
DAVID POND: Yeah, I think that's actually to me a really pertinent point because it's maybe the glib suggestion of Moore's law that obviously the computer chips continue to double in power every 12 to 18 months for the last 30, 40 years. Alongside that, I think the software is improving equally. So there should be the speed gains.
But the improvement of usability and all the things we've seen from Microsoft and others, it doubles the same over that kind of period. Too many clients get too close to the coalface. And all they're used to seeing what they've been using for years and years and don't actually realize that the rest of the world has moved on.
So you sort of make do with a certain level of success in your runs, a certain level of security of ease of user interface. And sometimes, it comes as quite as a shock to clients who think that Yeah, Well, suddenly realize they've been using the same package in the same release for the last 10, 15 years.
And then if they go and see what's available elsewhere, they think, actually, we had no idea that things had moved on so much. So I think there's a big challenge not to say bleeding-edge, but just to say, things are continually moving on, even without looking to the totally out-there suggestions.
CHARLIE SAMOLCZYK: Mark, maybe from your perspective, I think it would be also good just to touch on-- we use the term financial modelling. But what does that encompass? And what does that mean from an insurer perspective?
MARK BROWN: Fundamentally, financial modelling is about getting yourself to the stage where you can understand the risks associated with your business, understand the capital requirements of your business, and understand the behavior of your business. We see a lot there with companies that have mature models, that have calculation platforms that have evolved over many years. And they've got large and complex. And potentially, people have lost sight of what's in there.
And I guess this takes us on to the question then of, as David said, do you tinker with what you have, and is that continuing to add value? In my head, tinkering around with what you have is a bit like adding accessories onto your car. Fundamentally, you've got the same car. You've got the same limitations. You've just made it a little bit better by doing. So my brain at this moment is now stuck thinking about the fluffy dice hanging from the wing mirror in those old 1970s and '80 cars.
DAVID POND: It's a fair comment though, isn't it? Because most people wouldn't say, oh, you've got to-- most people aren't driving around in a 10-year-old car. And then if you take a 10, 15-year-old car out, or you take a brand new car out, you think, gosh, the fuel economy is better than I thought, or this is better than I thought. So for the best bill in the world, you're never going to get your 15-year-old car to the level of a brand new one.
MARK BROWN: No, exactly, that's it. And at some point, you do have to think about making a material change to what you've got. The question. I think I'd encourage people to think about is what that material change is. I guess can look at the market and see what people are doing at the moment. I think that gives us some nice perspectives on this and give people a view.
And fundamentally, I think we're seeing people go one of three ways when they make a change and limiting here, of course, to those that are looking to change. One of those is obviously going down the DIY route at building something yourselves.
I can see why that's attractive. It's gives you ultimate control over what you want, but can see a lot of risks in that as well. Quite often, puts key person dependency here and quite often overly focuses on the actual content and avoids a proper assessment of the technology, the distribution, the calculation, the auditability, the governance, the security. There's so much more to having a solution beyond just the model.
The second route is people convinced they need a change, then go out and look at was it the absolute bleeding edge. If you're making a change every 10 years, then people almost feel obliged to leapfrog everybody else and look at those bleeding-edge technologies. There's some great bleeding-edge technology out there, and there's some great innovation that's come up in recent years. I mean, obviously, talk about generative AI is one area of this.
What we do see, though, is that those bleeding-edge technologies tend to be a little bit of a one-trick pony. They're very good at doing one thing. But maybe lack, the breadth, and the security that you should be using particularly if you're doing statutory reporting with your software.
And then the third is really looking at the space of the tried and tested-- the companies that are in the market and are used by a lot of the peer groups of the insurers, making sure that you're up to speed on what's available from them and looking at things like automation and speed and control and governance. These don't sound exciting when you talk about them, but they are a huge step forward for a lot of companies.
It gives you all the benefits or the excitement of a big change in what you're doing and how you're working and potentially frees up a lot of people from the tedious work to do more exciting work. But the good thing about looking at these broader providers is you get all those benefits without the risk because they are tried and tested. They are tools that are used by the peer group.
And the benefits that the clients are getting with those is moving themselves a long way forward from where they are. The pack is a long way behind the bleeding edge. And it could perhaps be naive to try and jump to the bleeding edge when you deliver almost all of the benefits for far lower risk just by getting up-to-date and looking at the technologies that are now mainstream.
DAVID POND: Yeah, I would echo all of that and say, I think, yeah, I see those two key themes coming out in the market as well. Up until four or five years back, the received wisdom was always, it's just too much. We've chosen our platform. We'll use that platform forever. It's just too much bother to change.
I think in the last three to five years, a realization is that the world continues to move on. We talk about here where, even compared to 10 years ago, we're now 90% cheaper in terms of compute costs for model run in terms of the amount of hours you need to run for a particular run, and potentially 80% cheaper in the amount of people time to do a evaluation cycle.
So our Unify product where if you switch out-- add in robotics for all the areas where you don't want to spend expensive actuaries. People don't need to be manipulating spreadsheets. They don't need to be doing manual testing of things.
If all of that is replaced by Unify, we see an 80% drop in the cost of a whole valuation cycle relative to where they were at in their old system. So I think that is gradually dawned on the wider market over the last two or three to five years to say actually, maybe now is the time to start thinking because it's a lot more different than we thought. There is still a fear there.
Everybody takes the argument that, yeah, actually, this would save us huge amounts of money, but there is still a fear then to actually make the step to actually change a system just because that's mentally a big thing.
CHARLIE SAMOLCZYK: To me, listening to this-- and this isn't a question, just an-- this feels like a timely topic. It feels like, I guess, in my career, I've seen a lot of the rise of legacy system modernization and big transformation programs in different parts of the industry in the sector. But it feels like-- guys, tell me if I'm wrong. But it feels like this is kind of a hot topic right now in the life, financial, and risk modelling space.
DAVID POND: I totally agree. Yes, I think it's, as we mentioned earlier, kind of this perfect storm of having to do everything they could to get the IFRS over the line, and now actually a chance to finally pause for breath and think about, OK, well, what other problems have we been ignoring? And on the other side, you've got production valuation teams who weren't coping even before IFRS 17.
And now a genuine doubling of their amount of work that they need to do-- those teams, we are seeing the cracks appearing and they're just not coping. So there's no appetite to throw millions more on top of the IFRS budgets just to double the size of the teams perpetually. So think those two elements together are actually saying we need to find a better way, and now is the time to look out there.
CHARLIE SAMOLCZYK: Mark, you talked about kind of theoretically people can kind of tinker-- I think use that term-- tinker with what they have. They can buy bleeding-edge technology, or they can go with the tried and true, which actually gets them a long way down the path. What are we seeing in practice today with the clients that you work with?
MARK BROWN: I think we're seeing a good mix out there. I think we're seeing companies that are going DIY to an extent. But I think we're seeing a lot of clients also potentially having buyer's remorse when they go down the DIY route. Getting the model built, as I said, is the interesting part of the challenge, but there's so much more to it.
I think people are looking at taking on projects where they can get a significant payback. We've had a bumpy few years recently with not only all the regulation that's coming in, but with the inflationary impacts as well. Companies setting budgets and seeing costs go up through the years, and discovering that by the end of the year, the budget they had is barely stretching to meet the BAU, and a lot of pressure being to get those costs back under control now.
So we are talking to companies that are really looking to take on projects that can get a quick and material payback; projects that have got a return of investment of one, two years, maybe. Those are very easy to write the business case for undertaking them and to get the investment.
And of course, once you start freeing up money by getting those safes in place, you can then argue to reinvest the savings in further improvements. So there is this virtuous circle that people are trying to break into where they start getting cost savings, getting more efficient work, using automation tools to run 24/7 instead of 10 times five to drive those savings. And people really looking for something that will make a material difference and looking for confidence from the suppliers and their partners that these projects can come to the end and deliver a positive outcome in the timescales they expect.
DAVID POND: Yeah, I think there's a lot more-- agree with that-- a lot more shrewdness now. Gone are the days of handing over $50 million to one of the big four and then them not deliver very much. I think a lot of companies have been burned one way or another for big projects that didn't deliver.
Certainly, we as a consultancy now, most of what we do has to be on this step approach, as Mark explained, where companies will only part with a small amount of money. And then will prove the first step of the way. Then they'll part with some more, and we can prove the next step.
And so it's the double benefit of you don't have to trust us. We can prove step by step as you see the benefits. But then as Mark says, as you see the benefits, you start to get some of the savings upfront rather than having to wait three years for a finished system. Within the first six months, you're getting benefits, and then you'll get more benefits and more benefits as you go forward.
MARK BROWN: I think so. And just pick a couple of examples on that. If you can significantly improve the efficiency of your actuarial model-- so the calculations genuinely take less time, less compute than they did before-- then you get an IT saving. And if we look at, I mean, numbers taken from Microsoft's total cost of ownership calculator-- but the figures that companies are paying for 1,000 calls on a compute grid.
By the time you factor in everything, not just the hardware, the power, the cooling, the IT support, the licenses, people are spending-- apologies for the American currency but, $1 million a year per 1,000 calls they're running. And there's a big payback on that.
And even if you do it not on the compute, but you look at the time saving, getting two, three, five, seven extra days off of your time to do your year-end close, by automating and by de-risking and therefore fewer errors and faster access to the results and better understanding of how the process is going, again, that saves significantly. It gets more the numbers out to the management faster, but gives you more time for corrections if you need them, more times for additional investigation and a better understanding of the risks, all of which the management of the companies love.
CHARLIE SAMOLCZYK: Mark, talking about the recent developments, I mean, think the other thing-- and I know we talked about it at least internally. There is a difference between, say, the model and the platform. And you're talking about the grid and kind of the could compute. I don't know if either of you want to touch on that because I think-- and David, to your point, I think even if the software has the same name, I think we've made advancements on both sides. And sometimes, it's not as apparent kind of the separation between new modelling approaches or advancements in that space and as well as advancements from an automation perspective or a compete perspective on the platform side.
DAVID POND: Yeah, I could certainly talk on the development cycle side that there have been huge leaps forward. And we kind of used the same argument about self-build. Companies may not do a self-build. But even the proprietary system they're using, they're continually doing a huge amount of build and a huge amount of maintenance.
[INAUDIBLE] as silly as how you read assumptions tables in, we have spent a lot of time of effort developing the tools to actually say, not only, yeah, we want the slickest, fastest-running system, we want the system that has the least effort to upkeep. So anything within RAFM, our risk agility, our main system that is regular and is standard for everybody, we would say you don't need to be coding that.
That is taken away and that's behind the scenes. So all the assumption table reads, all the assumption setting approach is effectively taken out of the code. And you have a significantly lower code base so that, to some extent, you still have to define calculations for how you want particular product features to work.
But everywhere we can, we find ways of reducing that down. And I think that's one area where a lot of companies haven't seen that yet. They're still using an old system that's incredibly labor-intensive to develop, to do the testing of.
And then they still looking at a 50% fail rate on all of the runs they perform over a evaluation cycle due to mistakes and things that have been made where we would say, no, you should believe in better. You should be able to significantly reduce the effort to put into that. And you should be seeing negligible run fails when you actually come to it. Your system should be set up, whether it's using Unify or another automation tool, that says all of the manual things it can do for you. But all of the testing it does automatically as part of the cycle.
So, for instance, finding out that your model is going to fail by doing the runs, the live year-end runs, should never, ever happen. There should be a system whereby by the time you get as far as running, you should know that your data and your assumptions are all ready to go and have been tested out and are pure and clean.
And that just runs because with the ever increasing number of model runs and the ever decreasing working day timetable for both IFRS and Solvency II, you don't have the time and the possibilities that you had in the past of just rerunning. When you find a mistake or a model run fails, you cannot afford to do that anymore, or you miss your reporting deadlines.
So, during that cycle, it has to do all the clever things for you so that nothing goes wrong. And that, I think, is to me is one of the biggest changes we've seen in the last five, 10 years in terms of how is your software moved on? I would say that is a key thing.
That should be the biggest headache of any company. If you can take that away and say, no, no, we're guaranteeing once all the tools are in place, you should not be making these mistakes and not having to do the reruns and the investigation of why the numbers are wrong.
MARK BROWN: Good thoughts there, David. I like what you were talking about on that. And just tying it back to what Charlie was saying, I think in terms of doing the calculations, there's two halves to it. There's all of that technology framework that ensures that the model works seamlessly and slickly when it comes to being used in production.
Certainly, I'd be genuinely surprised if any insurer wanted to be or should be worrying about how to build and deploy all of that. This very much feels like the piece that needs to be got out of the business and passed on to a vendor or a supplier. And we've seen the move from desktop to enterprise technology to cloud. And I think it'll continue down that route with the SaaS offerings.
As companies look to push more of the raw technology side of this-- the security, the distribution, the frameworks, the integration-- push that out as an area that really isn't the distinctive competence of that insurer. Then you've got the other half, the model that's doing the calculations. And I think there's a healthy discussion to have with every insurer here as to whether the model they have today is the best model for them going forward, or whether they should be looking at something simpler or more generic, or potentially even the other way-- more customized.
But that understanding of the business and taking the distinction as to how accurately you want to reflect your own products and your own capital needs feels like the decisions that insurers should be engaged with, making and getting their models to reflect that. But all of the other pieces are about making sure those numbers can be produced on time, reliably really is something insurers shouldn't be anywhere near. You need to get out of the end user computing world into a world of automation, governance, and even software as a service.
CHARLIE SAMOLCZYK: One of the recent podcasts we recorded-- granted, it had to do with personal lines. But we were talking about machine learning and AI. And since we're on the topic of the underpinning technology is, what's AI's and ML's role in the life and reporting space? Does it feature?
MARK BROWN: It does feature. I can see it coming, and I think it has a big role to play. I think the key difference between the life and the PNC space at the moment-- the AI capabilities, particularly generative AI and the LLMs-- they seem to be more of an accelerator rather than a solution in the life space. So we haven't quite got the same machine learning in the financial reporting space.
Clearly, it has a role to play in setting assumptions and deriving assumptions from experience data. That's much the same as it is on PNC space. But we're seeing people talking about generative AI at the moment and how it can help people.
It will be an accelerator. Generative AI will come along. It will help people build their models efficiently by assisting them with the coding, alternatively giving them prompts as to what. It could do it will help people understand their models by allowing natural language questioning. Definitely one to keep a watching eye on at the moment.
It'll take a little time to mature. It has great potential at the moment. We're not there yet. But once we can get to the stage where we get more confidence in the reliability of the answers from generative AI, then I can see it's stepping out from being an accelerator, where it maybe saves 10%, 20%, 30% of a human's time in doing their job to actually being able to take over completely some of the more basic roles, and the modelers becoming much more of a reviewer and less of a builder.
DAVID POND: Yeah, I think that's right. I see at the moment rather like the robot butler approach, where the promise is great. But actually, the reality, we aren't quite there yet. So we, like everybody, are investigating this. We are investing in looking at this.
But in terms of what we cannot-- what we and others can actually offer today, the robot butler doesn't work yet. But think of it more like as the junior student role where they didn't want to be doing it anyway.
But the basic coding and the first draft of the report or the first draft of the documentation to go with the model, the computer model-- those sorts of things I think we are nearly there on stuff that's worth using and saves enough time. But people have to realize that it isn't 100% efficient. It's not guaranteed.
But so long as you're saying, well, if I asked one of my junior students to write this report, I wouldn't expect it to be perfect either. I would still review it and check that it was OK. So effectively, I think we're at that level already that you can actually ask for these things, and on the caveat that you're going to have to still do the same amount of work that you did as a senior. But it takes away the need for the junior to do it.
And as, Mark, I think there's a great promise for the future, I think eventually they will get there, but there are a number strong legal requirements, both from the actuarial requirements modelling and other areas, whereby you're not allowed to rely on it, and quite rightly so. You have to be able to prove the code that and the modelling that you performed.
So for some years to come, it' going to have to have human input. And I think it will need to be proved to be a lot better yet before anyone legislatively and regulatory is going to start saying you can rely on something that's been purely created by AI.
CHARLIE SAMOLCZYK: Fascinating. Why don't we leave it there. I thank you both so much for participating. It's been a really interesting conversation. And hopefully, we'll have you back on another podcast soon.
MARK BROWN: We'd love to talk to you again on this, Charlie. Thanks.
DAVID POND: Yep, thanks.
SPEAKER: Thank you for joining us for this WTW podcast featuring the latest perspectives on the intersection of people, capital, and risk. For more information, visit the Insights section of wtwco.com. This podcast is for general discussion and/or information only; is not intended to be relied upon. And action based on or in connection with anything contained herein should not be taken without first obtaining specific advice from a suitably qualified professional.
