InsurTalk: Building AI enabled follow underwriting with Bernadette Tredger, Apollo

In this episode, Piotr Piekos sits down with Bernadette Treader, Head of Portfolio Management at Apollo’s Smart Follow initiative, to explore how AI and augmented decision-making are reshaping the London market’s approach to underwriting.
Bernadette shares her journey from traditional actuarial work to leading one of the market’s most forward-looking technology initiatives. Together, they discuss the evolution of follow capacity, the balance between automation and human oversight, and how smarter data use can transform underwriting performance.

Piotr Piękoś: Welcome to another episode of IT Insights: InsureTalk by Future Processing. My name is Piotr Piękoś and I’ll be your host today. It’s a wonderful autumn here in London, the year 2025 is maturing gracefully, and so, uh, is the journey towards smarter, AI-enabled underwriting across London markets. Today I have a privilege to speak with Bernadette Tredger, Head of Portfolio Management at Apollo’s Smart Follow initiative. Welcome to the show. Hello, Bernadette. 

Bernadette Tredger: Hello, thank you. 

Piotr Piękoś: So why don’t we start today by giving our viewers a chance to learn more about you? Who are you? What’s your role at Apollo? 

Bernadette Tredger: Sure. So, Bernadette Tredger. I’ve been an actuary in the London market for over 20 years now. Um, and I started my… my journey just very traditionally from the reserving actuary who spends hours and hours preparing data on Excel spreadsheets, then going into pricing, capital modeling, um, managing teams across, um, the globe in reinsurance companies. Uh, and then moving into exposure management, underwriting strategy. And then I came to, uh, Apollo as a portfolio management actuary., 

And in ’23 joined the Smart Follow team where I kind of go full circle now with everything that I’ve learned in the insurance and reinsurance industry, um, and bringing all my actuarial skills together into, um, the portfolio management of the Smart Follow team. So that’s my… my journey to where I am today. Um, before then I had a job as a structural engineer. Uh, I worked in engineering, I worked on building sites, I worked in offices building, um, incredible structures. So creating something, finding solutions, that’s always been something I loved doing, uh, especially when it goes to technical… uh, in the technical world. 

Piotr Piękoś: Wow. So it means that, um, the career for you kind of around the full circle and you have a chance to build something innovative, something that, uh, drives the market, uh, forward. Um, would you mind, uh, telling us a bit more about the Smart Follow itself? What it is and how does it fit into a broader Apollo’s strategy? 

Bernadette Tredger: Yeah, sure. So the Apollo Smart Follow team was set up in ’23 and the idea behind this was to incubate, uh, the Smart Follow idea into Apollo. Uh, initially we started with the idea that we could, um, solve, um, direct placements in an augmented underwriting form. And that was going to the… the London market directly to have box present, to start talking to… to… to the brokers directly and offering our capacity, but through augmented underwriting. 

So a submission would come in, we’d be able to run the, uh, submission through our ingestion mechanism, use AI to augment the data, um, build an entire decision-making framework, and then out pops the answer of… of commitment, having done all the, uh, all the checks in terms of appetite, compliance, etc. So all the package was built. 

We then realized so the shoe leather broking doesn’t fit our model and we are going back into the, uh, spectrum of more like portfolio type solution, uh, underwriting. So if you think the bifurcation of the market where you have your traditional follow… uh, lead… lead underwriters who really do the lead underwriting, um, and then there is the follow market that comes in. Traditional follow market, uh, augmented underwriting, automated placements, portfolio solutions. So there’s a broad spectrum of what follow market is going to be like., 

We found that our space is going to be in what we call smart technology for… with follow only capacity. So that’s where the term Smart Follow is, what… that’s what it means to us. And in particular it’s, um, technology-led underwriting with, um, control by… with underwriting controls. So we are giving up our… we’re giving over the sort of underwriting information gathering to the tools we’ve got, to the latest technology tools, and we’ve got, um, as human loops to, uh, you know, have oversight over the… over the decisions that are being made by the tools and the processes we’ve implemented. 

Piotr Piękoś: Where do you think… because now since the introduction of generative AI techniques there are a couple of syndicates who are well embedded into that type of, uh, approaching underwriting. Can you, um, maybe indicate the points where Smart Follow initiative differentiates?, 

Bernadette Tredger: Yeah, absolutely. So it’s very important for us to to understand what can AI do, what can the technology do, and what are the humans just better at. And we’ll always have to have a level of oversight. So we can use AI to do most of the underwriting. They can do much more data gathering and… and find, uh, connections that, uh, we wouldn’t be able to see in the, uh, traditional underwriting. We learn it from experience whereas there it is all brought together. But we have to have the, um, the oversight. 

So if you compare us to more the ones that are fully automated embedded, what is important to us is, um, the… the connection with the producers of… of business. So it might be a lead… a lead underwriter, a lead of a very interesting book where we can work with them to embed our technology into their underwriting process. So they can grow and defend the book without having to worry about building all the technology because that is not what they are supposed to be doing. It’s not supposed to be their… what they’re there for. We can take that off their plate and they focus on the lead underwriting and make the best decisions and… and go out and find the business., 

So it’s working with… with these, uh, lead underwriting teams and also working with the brokers to find solutions that are effective for the market in terms of, uh, placements that don’t come to London easily. Let’s go and buy it, uh, find a technology solution where we have trusted leaders that write… underwrite the business, price the business, uh, and we provide the technology angle to sort of syndicate the business on the back of it. 

Piotr Piękoś: Okay. As we spoke offline, uh, you mentioned the collaboration with other follower markets. Uh, um, can you elaborate on that? 

Bernadette Tredger: Yeah, so the collaborate is really important to us. Actually one thing that’s always been close to the heart of the people in the Smart Follow team is to give back to the market. So just taking a step back, technology is not, uh, is not cheap in that sense. Um, technology itself is expensive. The people to operate technology are scarce and expensive. Like data scientists are very new in insurance. Um, and in that sense we’re looking for scalability., 

Now what happens is everybody has always tried to solve the same problem in the London market. A piece of business comes in and everyone is cleansing the data and everyone is doing the modeling etc. We want to take that off and say let’s just go for it. We clean the data, the lead benefits, they get the data and the follow market gets the data, gets the insight. Um, and that way we are becoming more efficient, we’re bringing efficiency back to the market. So it’s really taking that step and say it’s not propriety to us, we want to share back. That will hopefully bring the whole market along to work more efficiently and more cost effective. 

Piotr Piękoś: Well that’s something that is, uh, really impressive, uh, and certainly not a common theme, uh, that we see in, uh, in the market. Uh, but that… that’s the conceptual part. You are now, um, quite embedded into execution. So how does the Smart Follow initiative with, uh, the paradigm of having the market [at] the center, uh, performs today? What are the metrics against you benchmark the way of working?, 

Bernadette Tredger: Yeah, so that’s a good question. So first of all the Smart Follow team is class agnostic, so we could write across all classes of business. Um, and we’re still very small but we’ve made some very important wins that bring us on the journey. So, uh, we’ve started with a… we had a successful collaboration product with, um, project with another, uh, consortium where we are starting to, uh, ingest data for them and bring it back into the, uh, post-bind data processing. 

Uh, we have, um, another initiative which is really, really groundbreaking is we’re writing with an MGU in the US. We’re providing them with a… a quote referral platform. So the MGU is sourcing the… the… the business for us, can underwrite under the rules that are coded in the platform and the referrals will get automatically back to the lead underwriter. So the… there’s still the underwriting control by the lead and we are backing the lead, um, with the technology and the capacity and building… building the platform for them., 

And that initiative has now led to a… a very big, um, initiative with a US wholesale broker, uh… US wholesale broker where we are providing a… a fully actively portfolio managed quota share facility into the US market. So that is… that has started. We’re starting small, we’re doing proof of concept pieces of work and we are also looking at what initiative is scalable for the next product. So we call each class of business is a product for us. So we start with let’s say a cargo product, what are the other cargo opportunities we can… we have… we… we find and then we can increase the… the footprint of our cargo product working with MGUs, working with consortia, working with MGAs, uh, and growing that way. 

So yeah, um, that is our growth plan and we are seeing that success. At the end of the day we want to bring better underwriting control to the underwriting teams and with our, uh, technology and the insights allow the teams to defend the positions in strained markets and grow… grow their share where the opportunities are there. So essentially we want to get paid through loss ratio performance because of, um, improved data visibility, improved exposure management, and the data being fed back and improved portfolio management statistics back to the lead. That… that should pay back itself., 

Piotr Piękoś: Well these are bold, uh, statements. I think that, uh, um, it’s great to see this type of approach, uh, particularly its success in… well landing new business. Uh, essentially apart from the marine broadly understood as, um, subset of… of classes, if you were to predict, uh, what would be a classes of business that would be most susceptible to this type of innovation that Smart Follow is bringing to the market, what would there be? 

Bernadette Tredger: Okay, so this asking me to have a crystal ball, but we… we all know that there are certain classes which require less specialism, some require a lot. So initi… what comes to mind obviously is the property market is almost like a trade that could be packaged up into a traded commod… commodity. We end up with like clearing houses type solutions where risks are being put onto platforms and the… the markets go and put their shares on and take them back. So these sort of commodity classes, um, would be the first ones where we think we’ll see a lot of uptick of that., 

We are also, uh, doing it in casualty. We’re seeing, uh, successes in casualty. Um, and even within the specialty markets there are pockets where we can accelerate with, uh, our insight. Not just on like bread and butter vanilla risks but even complex risks benefit from, uh, data augmentation and make the underwriting decision process more transparent, uh, more easy to follow, more consistent. 

Piotr Piękoś: You’re past your minimum viable product, u, phase that’s for sure. But with scale there come challenges. Um, would you mind sharing with, uh, our viewers some of the key obstacles that, uh, you are facing when scaling up the Smart Follow, uh, initiative?, 

Bernadette Tredger: Yeah, there certainly two challenges which are very obvious to us. One is data is still very scarce, especially if you want to start a new initiative in the Lloyd’s market. How do you prove you’re going to perform? Um, so we’re using a lot of benchmarking, uh, in order to justify, okay, this project has this range of, uh, expected performance. And that’s where the portfolio manager comes in. And this is the framework under which we ensure that we… this… this… this project can succeed. Uh, and where do we have, you know, our, when you say pillars of safety to ensure that, you know, this is not going in… into a loss-making portfolio etc. So there’s the data is a challenge. 

The second one is people who really understand how to use the technology, uh, and, uh, just getting high quality, um, people through the door. I think there is a lot of technology firms out there that are able to help, like with yourself Future Processing, and they have the… the talent and is matching those talents with the problems and finding it… uh, finding them the most suitable candidates for the problems you’re trying to solve. So this is sort of on… on that front., 

Another one which I haven’t mentioned I think it’s just the… the… the pace of change, uh, of the market and how the, uh, Smart Follow is perceived, but I think that’s just a timing issue we will overcome. 

Piotr Piękoś: When talking about the data availability, obviously it’s a well-known challenge at least in the… in the London market, uh, perhaps even a structural, uh, problem. Um, what would be the solution if you were to decide, um, when it comes to the data availability challenge? 

Bernadette Tredger: Yeah, so you just look at what is out there. You have to be open-minded, um, and see, uh, what… what benchmarks you can get before you start the full… fully augmented underwriting journey. There’s always something. It’s so surprising what data sources there are out there, how much structured data there is out there for the augmentation piece. If you want to demonstrate performance this is where the actuary comes in where we can have, um, ways of how we could model a potential performance. Is it, you know, uh, creating new ILF curves, creating frequency severity models, where do you get the data from?, 

We’re benefiting from being a syndicate of… of multi classes to have some insight of how classes would perform and then the actuary is there to sort of tailor, uh, the existing data to what a potential new solution would look like. So this is sort of the portfolio management angle of it. The augmented underwriting is… it’s amazing of what’s out there for us to, you know, join the data up and bring it together. Um, especially with the ingestion of… of claims as they are at this point in time to get more… more of the claim feeling where the claims comes from and then going into the claims books and say, okay, this means we might have to look and find more data that, you know, around a certain area that attracts more planes etc. 

Piotr Piękoś: [Right, so] you mentioned that part of the game is really a proper modeling of the exposure before you engage with a particular book. So how does the AI augmented portfolio analysis differ from the traditional modeling approach, uh, from an actuary perspective?, 

Bernadette Tredger: Yes, so where we come in… we… we found the easiest way for us to come is at the post-bind stage. So it’s that building that portfolio management cycle where we start setting rules at the start saying we’ve got our risk appetite, our maximum lines, our portfolio optimization set. Then we bind risk, we ingest the risk into our systems as quickly as we can post-bind, building the overview of what the portfolio looks like, what it is susceptible to, and then feeding it back into the, uh, live underwriting decision. So is the post-bind analysis actually where we’ve got a bit more… more time to think and monitor, okay, how we are going, how we expecting to perform, and where we can respond to deviations of expected performance. 

Piotr Piękoś: So what about the volatility of the, uh, Smart Follow enabled portfolios? 

Bernadette Tredger: Yeah, so I think for us it’s important to small steps. Mm-hmm. Don’t take oversized positions, the usual thing. Make sure exposure management is the most critical stuff… uh, thing to us and that’s again where we think we can add more value than, uh, what it… the traditional angle does. Um, and just limiting your maximum exposures, limiting your RDS, growing that with your income and finding a lot of business that is the vanilla bread of… bread and butter business to… to cushion the volatility, potential volatility of the book. But we believe because we can surface exposure, we can surf exposure clashes that haven’t been, uh, seen before perhaps, we can walk in with our eyes open, uh, rather than, oh gosh, we didn’t expect there to be this kind of accumulation risk., 

Piotr Piękoś: Okay. So certainly you have, um, visibility into the portfolio dynamics at much more granular scale. So where does the human factor come in? Where is the underwriter? 

Bernadette Tredger: Yes, the underwriter has to have the oversight. The underwriters are becoming more and more portfolio managers with the insight that they have available around them and they can then decide, okay, we overweight in this… the example of blue ships and… and red ships. Are we writing too many blue ships? I think we should go in red ships and, uh, the rates in the red ships are improving so let’s just play a bit more here, keep our balance. So they will have a lot more overview of a portfolio play. Um, that’s what the, um, follow underwriter does and we write it… we… we played that back to our leads and we play that back to, uh, to the brokers etc. So we… our risk appetite is very, very well communicated and understood., 

Piotr Piękoś: Okay. Speaking of communication in terms of practical tools, do you provide a form of, uh, intelligence real time to underwriters and brokers alike? How does it look like? 

Bernadette Tredger: Absolutely. So, uh, I mentioned before a successful product, uh… project we’ve already, um, created, uh, with… with another consortium. Uh, so we can play back these insights of the portfolio directly to the lead as and when the we ingest the, uh, the data. So it’s… it’s live dashboarding, live making availability of data where, you know, they can draw the conclusions they want but the data is there., 

Piotr Piękoś: My last question, um, would be about more than just Smart Follow but, uh, thinking more broadly about the future of AI enabled syndicate. Can we try to embed into a small thought experiment and then… and try to imagine how, um, this could look like in the future where on this granular level those decisions are actually AI augmented and underwriter is more of a portfolio manager? What would the syndicate look like? 

Bernadette Tredger: Okay, that’s very interesting. So first of all there will still be people there. It won’t just be computers running by themselves. Uh, so we will have I guess experts in particular classes or multi-classes, um, who oversee… uh, what they… somebody needs to tell the AI what to do and what to change. So there’s obviously that oversight and we need the skilled… skilled underwriters to do that. Um, and act as well. If you think about an actuary, their job is to make sense of the future. They take decisions, uh, data that is there now and project it in the future. AI can’t do that., 

So we will have people who are more projecting into the future and making sure we haven’t got a black box because no one is going to invest into a blackbox syndicate. I think people will be unnerved by that idea. So it’s people who bring the transparency on the table, how the syndicate is going to perform. I hope… I hope it’s going to perform well with the right guard rails in place. Uh, whether it’s going to happen sooner or later it will. Regulation needs to catch up in terms of what… what governance they expecting to be in place. 

Like I think even about my profession, the actuarial profession has always been very strict on documentation of every decision, of every material decision. So, you know, as AI is being implemented in, you know, just generating documentation the thoughts have to move on. It’s like, okay, where do we need the human to have the oversight, at what level, and what needs to be, uh, sort of signed off by, um, a qualified act… um, yeah. So it will be a syndicate as it is now but people just focusing more on the output of the data rather than the individual, um, risk., 

Piotr Piękoś: Bernadette, thank you so much, uh, for your time, uh, um, great insights, um, and, um, a wonderful place here at the Apollo. And, um, to you dear viewers, thank you so much for your attention, uh, through this episode of IT Insights: InsureTalk by Future Processing. Until next time.