ROUNDTABLE
Generative AI: running before we can walk?
Love it or loathe it, Gen-AI is here and the worst thing re/insurers can do is ignore it. Experts at a special Intelligent Insurer roundtable in Monte Carlo debated how to make the most of this technology.
IN ATTENDANCE:
1. Isabelle Clausner vice president, client executive EMEA, Xceedance
2. Michael Cook claims advisory and London Market consulting lead, PwC
3. Sandra DeSilva chief executive officer, Nova
4. Richard Hartley co-founder and chief executive officer, Cytora
5. Richard Lawson chief executive officer, Polo Commercial Insurance Services
6. Hartmut Mai group president, Cyberwrite
7. Brian McGregor head of property, Africa Specialty Risks
8. Frank Perkins founder and chief executive officer, Inari
9. Bill Pieroni managing director, Acord
10. Ben Rose co-founder and president, Supercede
11. Jose Seara founder and chief executive officer, DeNexus
12. Rosina Smith chief product officer, McKenzie Intelligence Services
13. Carlos Wong-Fupuy senior director, Global Reinsurance, AM Best
14. Chair: Bill Zuill editor, Bermuda:Re+ILS
Following the launch of ChatGPT and the buzz around generative artificial intelligence (Gen-AI), talk of machines replacing humans, and bringing huge jobs losses, was rife, but it is not necessarily true right now. As an industry already au fait with AI, machine learning and other forms of data automation, re/insurance has taken a keen interest in this latest development. But how far will the industry embrace this innovation, what are the real opportunities and benefits and what are the biggest risks?
With so many unknowns in play, Intelligent Insurer invited industry experts to a special roundtable discussion titled “Harnessing Gen-AI for re/insurance success”, held at the Monte Carlo RVS, to talk about the biggest opportunities and threats, how far Gen-AI has already penetrated the industry and how much further this developing technology could go.
Frank Perkins said there were plenty of “evangelists and experts in AI” when ChatGPT launched, while others felt it conjured up sentient robots and job losses, “which we are far away from”.
“But if someone’s job is copying and pasting from a PDF, should they ever have been doing that job? Let’s have people doing people jobs and robots doing robot jobs.”
He said that re/insurance still has a problem ingesting data, but the next evolution of AI could really help with this.
Rosina Smith emphasised that “AI is capable of seeing relationships between non-linear variables, which could transform underwriting”.
She said: “We’re keen to focus on value creation and we’re seeing the fruit of these labours in customer satisfaction.
“We’re not unleashing the full power of these technologies because we’re in the mindset of incremental improvement.”
Michael Cook
“With the ‘AI winter’ that lasted from the 1970s to the early 2000, when there was little interest or developmental strides, you could draw a parallel with innovation in insurance, or the lack of it. Now this latest wave of interest can be harnessed for a fruitful ‘AI summer’,” she added.
For Brian McGregor, Gen-AI presents an opportunity to supplement an organisation’s underwriting and help customers with chatbots. “You could train a ‘large language model’ (LLM) on insurance information and use it like an underwriting mentor,” he said. This would help Africa Specialty Risks to increase insurance penetration levels across Africa, a continent where the protection gap is wide.
Richard Hartley said that the next level of AI is already here and pervasive across the value chain. “Gen-AI will create a step-change in functionality. You will see the interaction in the value chain becoming more efficient between brokers and insurers,” he said.
Caution urged
Many at the roundtable urged caution when considering the possibilities for the industry. Ben Rose said his firm had experimented with Gen-AI to see where it could make something better.
“The issue with using AI LLM to create something is that it doesn’t deploy logic or calculations. It guesses the most likely next word, but in terms of logic it doesn’t understand and has made things up. For those of us who care about accuracy, reliability and data, Gen-AI is a distraction more than anything else,” he said.
Bill Pieroni suggested that it could help make operations “a little better, a little cheaper” with opportunities for new value systems that could “challenge the current operating models”. However, he said, in terms of transforming the industry “it will be another decade or so at least”.
The Gen-AI development was important, according to Michael Cook, because re/insurance has had issues with data for years. “Gen-AI can help address those data issues and I think it has potential to address the protection gap,” he said.
On the topic of whether Gen-AI is just a “useful tool or a transformative technology”, Hartmut Mai said the technology is still in its relative infancy and that we’re just starting to see and get a feel for it. “It’s difficult to predict today where we will be in 10 years from now, as with blockchain.
“Blockchain was actively used and drove innovation,” he added, highlighting a live use-case where blockchain was used to make transactions happen over eight countries, issue policies, make the payments, and refer matters to the local tax authority.
“The problem for the industry is that when something new comes along people don’t give it a chance, they kill it. And that is not helping innovation in the industry.
“The insurance gap is an area where it could help, and where re/insurance needs to stay relevant. If they are not embracing innovation they will lose their relevance.”
“It can be used to refine and rebuild—that is where it can make a difference.”
Sandra DeSilva
Get the tool right
Describing Gen-AI as a tool, Perkins said that when something new comes out there’s a tendency “to want to use it to do everything, even calculate the treaty rate. That’s silly—it’s not the right tool”.
He added: “With this tech we’re slapping all of this over existing operational processes. What we will see is how these things fuse and change. It goes back to being the right tool—Gen-AI has a space but it’s not the right tool for everything.”
Richard Lawson said it can be used to reduce expense ratios and save time. “This time saved can be used to write more risks. Underwriters can view more risks and assess more risks. It’s all about enabling humans to do more and do it faster.”
Sandra DeSilva said she has seen clients optimising selection and creating optimal portfolios, and risk mitigation. “For example it can be used to refine and rebuild—that is where it can make a difference.”
Cook suggested that it has the potential to do much more. “Since the 1970s we’ve laid it on top of something that already exists. We have blockchain and AI and Gen-AI, and we’re thinking about them with an incremental improvement mindset.
“But actually we could just tear it up and do things completely differently. We’ve only seen small examples of that.”
He argued that the tech can help humans as he cited how Gen-AI has taken away admin and can even produce decent reports. “We’re not unleashing the full power of these technologies because we’re in the mindset of incremental improvement.”
DeSilva said that her firm works with startups which need to get licences and regulatory approvals. “You need to show things where you are minimising the risk of uncertainty. Once you have that you can certainly accelerate.”
But she said that when AI does the underwriting, companies will need to insure their AI. “So for now there needs to be a more of a balance,” she countered, adding: “We will get there but at the moment you can’t even insure AI, so it is a progression thing.”
What does the industry want?
Carlos Wong-Fupuy said that to get the best out of this technology the industry needs to think about the current situation and what it, as an industry, wants.
“Gen-AI has a number of applications on the admin side. We need to remember we are in a regulated industry and we need to get the right balance.”
Commenting on regulation of AI, he referenced a comparison that people make with the aviation industry when they talk about the impact of restrictions.
“They say there are too many restrictions, and with fewer of them humans would have gone to Mars already. But we can’t go to Mars with all the crashes that would have happened in between,” he said, with implied warnings for unregulated AI development.
Isabelle Clausner added that in this initial “Gen-AI hype phase”, people can see the potential but they don’t really understand the value it brings.
“Tests where underwriting had taken hours, took seconds with Gen-AI.”
Isabelle Clausner
“We’ve just got over the internet of things and blockchain excitement, and the wider industry was only just implementing them when Gen-AI came out, and those techs are still quite immature. One of the threats or risks is that we are trying to run before we can walk.”
She said that Xceedance is trying to lead this transformation and identify use-cases across the process. “We’ve been running proofs of concept and the results are stunning. Tests where underwriting had taken hours, took seconds with Gen-AI. The accuracy is there, so the process is more efficient.”
Smith warned there is a danger that the industry looks at Gen-AI only as a tool. “This toddler that is Gen-AI is going to turn into something smart and powerful. We can’t be passive about it. As we move out of this winter of AI and to get to that summer point, we need to nurture it until it’s ready for us.
“We have to drive AI and show where it can be useful, we have to take ownership of it rather than have it forced on us.”
Jose Seara added that this technology will likely be able to do things that would not be possible otherwise. But, he said: “The challenge is how you validate the outcome of that. How do we make it credible?
“It represents a powerful opportunity to add visibility to risks that are dynamic and fast-moving, as well as remove the noise from risks where there is a lot of data.”
Perkins smiled: “We will know it has been successful if it provides value without anybody even talking about AI.”
Main image: Shutterstock / thinkhubstudio