In an industry that places enormous value on consistency, predictability and the tried-and-true, the prospect of letting loose fast-evolving and, in some cases, black-box technologies like artificial intelligence (AI) can be daunting.

Alex Taylor, in his unique role as Global Head of Emerging Technology at QBE’s venture investment & development arm QBE Ventures, wants to push insurers beyond their traditional risk tolerances to safely embrace game-changing digital innovations – innovations that will create more viable, better-performing, and more customer-oriented businesses.

A seasoned digital leader and former software developer at the bleeding edge of the innovation curve, Taylor brings a wealth of experience and a unique perspective to the executive table.

In the lead-up to his upcoming keynote at our Future of Insurance 2024 conference next month, the tech veteran walks us through the QBE Ventures’ experimental wheelhouse and latest innovations, his thoughts on how AI is shaping risk assessment and mitigation for insurers, and becoming an AI leader whilst also navigating the complex web of international regulatory standards emerging around the technology.

FST Media: Whilst still valuing consistency and predictability, progressive improvement and innovation have become vitally important capabilities within the insurance sector, and indeed the broader FSI space.

How is QBE Ventures contributing to this ever-evolving innovation landscape? And are there any projects that you’re particularly proud of?

Taylor: It’s something that really cuts straight to the beating heart of the insurance industry. We’re starting to see the relationship between insurance products and the technology that supports them increase in importance, and customers are starting to be curious about what they can do to minimise their risk. The reason I love this is that it stops insurance products from being purely about risk transfer – it becomes a true partnership between insurer and customer and amplifies the opportunities to assist customers should the worst occur.

A lot of work that we’ve been doing recently in geospatial technology is particularly key to this. In property insurance, a lot of the time, people ask us if there’s something that increases the likelihood of loss related to their risk and if there’s anything they can do to address it. Broadly, we call this our ‘resilience thesis’, and resilience to us means giving someone information about their risk to help inform their decision-making. If something is commercially viable, minimises risk, and is actionable, then that’s a very good sign. A lot of people can do things now, particularly in areas like property – both residential and commercial – that minimise that risk.

A recent example was from Hurricane Ian, a huge storm that affected the eastern seaboard of the US in 2022. By using some of our geospatial technology systems that do rooftop analysis and oblique analysis of buildings, we were able to locate buildings that were damaged during the hurricane. We started reaching out to those customers and the response to that was extremely positive.

The brokers love it, the customers love it, and we as a carrier love it because it minimises exposure and risk.

It’s probably one of the most profound changes to the insurance industry that I’ve seen. And it’s increasingly necessary because the landscape of risk is ever-growing.

FST Media: How do you overcome some of the more unique challenges that the insurance industry faces when integrating some of these systems – for instance, the challenges in balancing customers or claimants, employees, and business needs?

Taylor: The insurance industry is highly regulated and the legacy systems that the industry has built can sometimes make it difficult to change existing processes to integrate new and emerging technology. A fantastic thing about recent technological advances is that moment when people realise there’s an incompatibility between a wait-and-see approach and continuing to be at the forefront of an industry or even responding to competitive pressure.

What I love particularly about generative AI, or GenAI, is that it makes insurers look over the fence and see what their competitors are doing and where they are on the adoption curve.

If they’re not prepared to take advantage of that, or their systems and processes aren’t prepared to take advantage of that, then people will soon realise that this is an existential threat.

So, if there’s one positive, other than the technology itself that this has created, it’s that people are maturing their processes to more rapidly adopt emerging technology than we would’ve ever seen a decade ago. A decade ago or even 20 years ago, you might have had more time to take advantage of newer technology; you might have been on a five-year time horizon; you could do a lot of PowerPoint presentations and talk a lot about what you were going to do. Now, you might have three months before an entire product line is taken away from you by a more efficient competitor, because they’re taking advantage of something that dramatically reduces their operating expenses and overheads.

Frankly, having worked in a lot of industries in my career, this is the time when an industry is the most exciting because people have to make quick decisions.

They also have to pay attention and work closely with regulators. They have to put due process and safety into place, and when you’re incentivised to move more quickly, that’s where the action happens.

FST Media: Could this kind of fast-paced innovation and transformation in the industry lead to a more collaborative, or alternatively more combative, environment?

Taylor: Yes, to a certain extent there will be more collaboration.

Insurance, by its nature, is a competitive landscape, but ultimately the customer will win because you’re seeing this competition to reduce expenses and better understand risk. That’s obviously something important to regulators, too. One way or another, the customer wins because products become more efficient, more effectively targeted, potentially cheaper, and where a risk looks good, there’s a lot of very good opportunity. We’re starting to see the emergence of entirely new product lines that were infeasible previously because the time and expense of operating them would’ve meant that it wasn’t commercially viable.

Now that we have machine intelligence that we can treat a bit like a partner to help augment our decisions and processes, we’re starting to see a lot of these ideas that have been on the shelf for a long time get dusted off and say, ‘Hey, you know, this is something that we can actually do!’.

FST Media: With this flood of new technologies coming into the industry – some, admittedly, more useful than others – and limited budgets to explore them all, how can an insurer effectively choose the right tool that best fulfils their or their customers’ needs?

Taylor: The nature of the technology we’re seeing emerge at the moment is looking at solutions for legacy industry problems from multiple fronts. For example, document ingestion and OCR (optical character recognition), which are now moving from their infancy into maturity, means that we can obtain important decision-making data in a way that doesn’t cost as much or require as many human hours as it once would have, freeing our people up for more higher-value tasks.

There are further interesting examples that I’ve seen recently in other industries that offer potential for insurance. If it can be cheaper and more effective for the customer, and support our employees by reducing the time burden on them, then that’s definitely a technology worth exploring.

FST Media: When considering a new technology, what are some of the best ways for an insurer to balance the benefits and limitations that a particular solution may offer?

Taylor: My job is to stay at the forefront in this space – and probably for the first time in my career, I feel as if things are moving faster than even someone who spends their day focusing on it can understand the adoption curve and the progress curve of. For example, the other day I had a demonstration from a company that had created a multimodal generative AI model which can take in pictures, watch videos, and use tools. In this case, they gave it a claims information packet including some CCTV footage, a police report, and an audio call of the customer reporting the claim, with the expectation that it could read the documents, watch the video, and understand the claims guidelines. And it figured out how to do that end-to-end all by itself to essentially process the claim! That’s the rate at which we’re evolving.

The key thing we need to be very aware of, however, is that we’ve got an emerging regulatory landscape alongside this. We must adhere to the standards, protect from biases, ensure safeguards in relation to ethics and morals, and protect customer and third-party data to make sure we’re operating with appropriate oversight across multiple global regulatory landscapes, as QBE operates in dozens of countries.

FST Media: You might be familiar with some experiments recently where researchers were testing whether AI could pass the LSATs (Law School Admission Tests). There’s still, I feel, a lot of room for improvement here.

Taylor: It’s interesting looking at the progress in human benchmarks, or ‘evals’ as they’re called in the generative AI space. People get lost in the minutia of particular things to the point where they can’t see the rate of evolution. But even with the LSATs or the medical boards or the AP calculus college entrance exam, now that multimodal models exist that can look at and interpret diagrams, we see these tremendous advances. That is a higher-order task, and it’s this kind of advance that lets us understand the vector at which this technology will intersect with human skills and abilities.

FST Media: On the point of emerging regulatory standards in AI, how can insurers effectively implement new technologies like AI whilst still adhering to compliance requirements, especially if, like QBE, they operate internationally?

Taylor: That’s probably the most fundamental question in this entire space. And, conveniently, a common theme that we’re seeing across industry groups and regulatory bodies globally is around oversight.

Fundamentally, a human or a company can be prosecuted for lack of adherence to a standard. A machine cannot make a decision that would have criminal culpability or even civil culpability as long as it’s a machine-assisted process. And that’s the key here: that it’s not an automated process. A machine-assisted process has oversight by an appropriately qualified human and, ultimately, it’s the human that’s required to be the arbiter of what’s correct and what isn’t. Yes, the promise of this technology is that we could do more and it’ll cost less. We can be faster; we can be better.

But all the best use cases we’ve seen in recent generations of AI have been directing a human’s attention to what they need to pay attention to that would’ve otherwise taken substantial periods of time.

One of my favourite examples that we are starting to talk about right now is something that we did in our North America business between September and December last year. We looked at cyber insurance submissions, some of which span 400 pages. It takes about a day to read through one of these and can be costly. From September to December, we implemented a product that employed a technology called Retrieval Augmented Generation. Essentially, it can take all of this documentation, ingest it into the system, and then ask the language model to pick out the most relevant pages to answer particular questions.

We had 196 questions that we asked as part of our cyber insurance underwriting, which is very time-consuming. We went from the initial review period, being four to five hours, to about 15 minutes. It’s that kind of transformative capability that is the promise in this space. But, again, it’s the human that’s the fundamental arbiter of truth. And as long as that’s the case, and I believe it will be for the next few years at least, then a lot of this experimentation has guardrails. Now the guardrail is that we are pushing due process and human time and attention to a problem that’s been accelerated, but not replaced by AI.

FST Media: Do you have any success stories where QBE Ventures has implemented new technologies and successfully navigated regulatory obstacles?

Taylor: What I find is most effective is when a challenge is brought to us that we can solve by being more adventurous with the technology that we’re starting to engage with. One of my favourite examples of this is with a company that we invested in a couple of years ago called TensorFlight, an amazing organisation founded in Poland. They do machine vision analytics of buildings, so they can look at a picture of a building from the side and from above, and can tell you how much it would cost to rebuild based on what it’s made out of using its model.

One of the most profound things that we discovered early on was its ability to detect flammable cladding. We gave it some examples of identifying buildings with flammable cladding – and in some cases, people had physically been to inspect the building and missed the presence of this material – and the model was able to pick up on it. We found 55 new buildings that had this material in the greater London area that we insured, and many of those have now been remediated. This is the kind of example I like to highlight in terms of how innovation can be applied in a truly meaningful way because, ultimately, insurance is about protection and protecting lives has to be the most important thing. Lives and livelihoods – that’s what insurance is.

FST Media: We’re excited to have you as one of our featured keynotes at FST’s Future of Insurance 2024 event next month. What do you hope to hear about and share with your peers from across the industry?

Taylor: I speak to a lot of people in insurance globally every day and the message that I’ve been sending recently is that change is possible. QBE is a 138-year-old organisation. I wouldn’t have believed if you’d asked me 12 months ago that we would have generative AI use cases in production today. Yet here we are. It’s because our leaders across the organisation believe this is something we need to do. We need to have a position on this and understand what the timeline is.

When you have a combination of vision and willingness to engage, hurdles start to disappear, roadblocks start to disappear, and doors open.

I’m starting to see that right now, and I’m keen to hear other people’s experiences in this space, their experience in the last 12 months particularly, and where they see that going.

There’s a lot of fear of AI, particularly among people who don’t truly understand what’s taking place. But the message that I’m sending is that with appropriate controls, with appropriate care and due diligence, it can be a force for good and everyone can benefit regardless of where you sit in the equation.

Alex Taylor will be a featured keynote presenter at our upcoming Future of Insurance, 2024 event, exploring how rapid advancements in artificial intelligence and machine learning technologies are reshaping the insurance sector.

Event registration is now full. However, extra places are being made available for FSI industry members early on our waitlist. Don’t miss out!