TRENTON MCNEE: My name is Trenton McNee. And I'm the fintech and digital assets industry leader for WTW's UK Financial Institutions team based in London. Today, I'm joined by Assem Marat, who is a senior FI underwriter from Starr Insurance, also based in London.
For those that may not be aware, Starr's London branch started writing FI business in 2019 through both its company and Lloyd's syndicate platforms and works exclusively with the broker intermediary market and serves a wide range of organizations within different industries. Assem, thank you for joining me today.
ASSEM MARAT: Thanks, Trent. Great to be here.
TRENTON MCNEE: My first question today is generative AI can be used to help an FI's operational efficiency, examples being adverse selection, the filtering out of bad actors. However, it could also be a cause for future losses such as AI washing, for example, if companies are exaggerating their claims to artificially pump up the share price or to win new contracts, new talent, or to be acquired outright. Can you please provide your take on this?
ASSEM MARAT: When we look into the financial institution sector and how AI is being used within it, to broadly speaking and generalizing, it is to deliver better efficiencies across the sector by speeding up information processing, and in a more technical and technological level, by generating better developer productivity.
When you look at the sector and the participants in the sector, these are the firms that are primarily focused on providing services to clients. Client service and customer outcome is really at the core of what they do. We don't deal with extraction industry clients, for example, where we're digging up goods from the ground and we're trying to sell them for the highest possible price.
A lot of what happens within the sector is about client service. It's about how they service their clients. It's about how they can process the information to improve that service and/or identify patterns to make the business better, which ultimately delivers a better customer outcome, which translates in a better business performance.
As for the second part of your question concerning AI washing, I would mark it down to just a change competition environment because of AI. As it is with any innovation tool that enters any market segment, that changes the dynamic and how you compete.
For example, maybe a little bit of an outdated example but probably still relevant is when you look into print and publishing, once you incorporate a digitization into that sector and the ability to read magazines and newspapers online, that created a different competition dynamic and a new dimension for firms competing in that sector.
So really, it's the same with AI. And as it is with any new innovation tool that enhances competition, you're going to have examples of companies adapting very well and where it, again, generated a better customer outcome or a better business productivity.
And you will see examples where it hasn't really been adapted in the same way or it's more negative than positive, partially because of the lack of hardware or the lack of understanding around what the tool represents itself.
TRENTON MCNEE: Thanks, Assem. Yeah, I think for the readers, they'll find it's a net positive from what you're trying to say there. This brings me on to my next question. This is being asked regularly in FI client presentations now. What are you most concerned about when addressing a client about their own AI capabilities from a financial risks perspective?
ASSEM MARAT: So financial institutions as an industry is a heavily regulated segment. And it has always been a heavily regulated segment. So when we look into the regulatory angle, we can see that there is already an established framework within it.
So from a corporate governance and a compliance perspective, when we look into the actors within financial institutions, we see a great degree of consistency. There are AI governance committee that assess business cases. And then there are AI risk and control committees that then control how those business cases are being used and implemented throughout the organization.
We also have to remember that for most banks, for example, the usage of models is not a new concept. And when you think about models, you have to think about it in the sense of algorithms that-- any algorithms that run automatically. And banks would have those in the past to assess pricing and capital and risk allocation models.
So again, the concept itself around what the models are and how to use them is not new for financial institutions. What they've done is that they've taken established framework. And they've modernized it. And they've changed it to fit AI and AI regulation.
And actually, regulation in this sense is helpful because we've seen multiple financial institutions working with regulators to assess the direction of that compliance and governance framework and also the degree of changes that they needed to implement. Banks in particular would have worked very closely with regulators around things like entitlement.
What we're really interested in-- and this is not something that we hear a lot when we talk to our clients or we sit on market presentations-- are AI-created exposures. So new exposures that we potentially haven't thought of before or existing risks of different dynamic.
For example, data being the obvious one and the one that we all are worried about. But it's not just about how data is stored and how data is being used and who has access to it, it's also around how does an insurer manage data accuracy once they've completed their own model training?
Models use stochastic and statistical ways to figure it out, the answer to the question that's been given to them. And no matter how good the source data is, there will always be a degree of variability. So it's not just about inaccuracy, it's also about variability and how do you assess the difference in that? And how do you filter that down to what you really, really trying to achieve from a model perspective?
So for some organizations, it would be a question more of the latter around how to filter down variability. And for other organizations, it would be the question of the former. How do they get to a point where the source data is really good and accurate? And for some other firms, it would be both of them.
We also need to remember that within financial institutions, we do have a lot of unstructured data. That unstructured data is things like correspondence and trade instructions, compliance log. And all of that adds complexity into the model training itself and into the human management component of that.
We also think about fraud and how fraud is changing and whether we're going to see increased fraud cases using AI. We've seen some very sophisticated fraud cases in the US and in Australia with the use of AI.
And again, if we are to assume that AI is a learning tool and it's a tool that is ever evolving and changing and giving you better responses to the same questions asked, so it's only natural to assume that the same thing is going to happen with fraud. The more examples you put about how you're trying to perpetrate fraud, the better the ideal for scenario can become.
We also look into things like energy consumption and dependency on certain hardware but also availability of that hardware, dependency on certain cloud providers, which we know are not many worldwide.
And if we look into that further down the line, we always have the debate around whether we're using proprietary models or we're using open-source models. And that actually in turn will have an impact on the tech sector, which something maybe for a different podcast.
Also, there are other factors that we like to think of, which are the move from how, to what, and why. And that is more in relation to what is the tech departments and tech services within financial institutions which are traditionally your very simple IT support departments. And now coming into front-foot of their organizations because they are helping the business teams and the business units to run better informed decisions.
And again, when you talk about developer productivity, they have to deliver better business outcomes. So they're becoming more and more relevant on the day to day of the business than they used to be.
Again, as a natural from that point around the change decision-making and how informed decisions are being made, we also need to think about, how does accountability going to change from a regulatory perspective?
TRENTON MCNEE: Yeah, thanks, Assem. I could chime in and say this also impacts employees, shareholders, suppliers, executives, and customers. Really good answer. And finally, where are the greatest insurance opportunities for the general insurance industry?
ASSEM MARAT: Already talked about efficiencies at the very beginning. And of course, insurance is no exception for that efficiencies that will be the obvious one. And I think when we looked into how and where we can deliver efficiencies, we have an enormous amount of data within the industry which, unfortunately, is not used to its full potential.
So if we're able to deliver a more efficient way of processing that data, there's a number of benefits that we can achieve. The first one would be around actuarial models and how it's not just about giving you better pricing points but also giving you a better risk assessment tool with the ability to identify patterns, also making certain systemic risks a little bit more predictable, which is future learning for us in the FIBI side.
Potentially, we can even use it to predict market cycles, which can then feed into your business strategy. It can also be used to create better capital allocation tools and better capital usage tools, which, again, will be hugely important to your business performance as an insurance carrier. And then
We can look into the development of new products. Like if you analyze your loss data and you have the ability to analyze large amounts of loss data and you can identify gaps in coverage, you might be able to create new and more innovative products in that space.
We have also spoken about underwriting efficiencies in the market where, again, submissions, informations can be processed quicker. And the AI model can help you identify red flags straight away.
We do have a couple of examples of automated underwriting platforms. So again, AI can be used to deliver some further sophistication to those. It can also lead to a more sophisticated way of underwriting because you can also tie a lot of externally available data or a lot of publicly available data. And the obvious one would be everything that is out there around US litigation.
TRENTON MCNEE: Thanks, Assem. And just to summarize today, what are your three key points you would like everyone to take away from today's podcast.
ASSEM MARAT: Firstly, AI is now broadly used within a variety of industries, not just financial institutions and tech. And the aim seems to be to extract a net positive business impact.
Secondly, we might see some further regulatory challenges around that the more the usage of AI is being adopted. We're also probably going to see some business infrastructure and operational challenges, going back to the debate around proprietary and open source models, for example.
Thirdly, AI is most definitely, again, going to get bigger and has a potential of significant impact on general economic components like production, consumption, and exchange activities on a larger scale.
TRENTON MCNEE: Thanks, Assem, for your time today and for those who have listened to this podcast. Look out for our next podcast episode in the All Eyes on FIs series.
SPEAKER: Thank you for joining this WTW podcast featuring the latest thinking and perspectives on people, capital, climate, and risk in the financial services industry. For more information, visit wtwco.com.
Willis Towers Watson offers insurance-related services through its appropriately licensed and authorized companies in each country in which Willis Towers Watson operates. For further authorization and regulatory details about our Willis Towers Watson legal entities operating in your country, please refer to our Willis Towers Watson website. It is a regulatory requirement for us to consider our local licensing requirements.
The information given in this podcast is believed to be accurate at the date of publication. This information may have subsequently changed or have been superseded and should not be relied upon to be accurate or suitable after this date.
This podcast offers a general overview of its subject matter. It does not necessarily address every aspect of its subject or every product available in the market. And we disclaimer all liability to the fullest extent permitted by law. It is not intended to be and should not be used to replace specific advice relating to individual situations. And we do not offer and this should not be seen as legal, accounting, or tax advice.
If you intend to take any action or make any decision on the basis of the content of this podcast, you should first seek specific advice from an appropriate professional. Some of the information in this podcast may be compiled from third-party sources we consider to be reliable. However, we do not guarantee and are not responsible for the accuracy of such. The views expressed are not necessarily those of Willis Towers Watson. Copyright, Willis Towers Watson. All rights reserved.
