Ferma members told AI will bring risk and reward for corporate insurance buyers

The growing use of artificial intelligence (AI) by insurers promises big benefits for corporate insurance buyers, but also comes with risks, Ferma members were told. AI has the potential to cut costs, speed up processes, standardise submissions and generate risk insights, according to the speakers. But the technology could lead to less risk differentiation by insurers, while the technology is not yet suitable for predicting risks, they warned. The application of AI is particularly relevant for risks with large volumes of homogenous risk, such as personal lines, explained Salman Siddiqui, senior director and insurance practice lead at Moody’s, during a Ferma webinar. “Actual underwriting does not even happen now [in personal lines]. Retail insurance is all about pricing using a model and having some risk appetites,” he said. But in corporate insurance, where risks are not homogenous and data is not standardised, specialist risk still needs to be actively underwritten, said Siddiqui. “As a result, there are not as many applications for generative AI models as you have for retail insurance… That does not mean that corporate and specialty insurers will not use AI. They will,” he said. According to Siddiqui, insurers will use AI to interrogate and summarise information, for example to speed up the submission process. It can also be used to automate and accelerate underwriting and claims processes, such as running analytics or assessing claims, and even predict potential loss behaviour within a submission. At the higher end of capabilities, AI could be used to automate underwriting or advise underwriters on the best risks to underwrite. As a result, insurer response times should improve, according to Siddiqui. “Insurance submissions should start to become more standardised because insurers and brokers should become clear about what data they need and what is superfluous, which will help simplify requests over time. As a result, hopefully there should be faster responses to your needs,” he said. More than a third of an underwriter’s time is spent gathering data, but with generative AI, insurers can assemble information more quickly and gain insights, freeing up time for decision making, said Siddiqui. “Hopefully in the future you should get a faster response to your needs. It is inevitable and will happen over time,” he said. AI could also help insurers offer new products and services to help corporates manage their risks, including challenging areas like climate change, according to Lloyd’s head of emerging risks Marco Lo Giudice. Cyber is a good example of where AI and machine learning is being used to marry insurers own experience with third-party data. This helps identify threats that pose an existential risk, and also provides risk management services that help mitigate such risks, explained Lo Giudice. He believes there is potential for this approach to be applied to other lines of business. “This is a good example of what we see in the market, and we see potential for that in other lines of business. One aspect I hear about a lot is climate. It would be nice if in the future we underwrite property in a similar way. Instead of moving away from areas where climate risk might be heightened, providing a service to clients to increase resilience,” he said. “Some of that is in its infancy and it is not quite in the market yet, but I see the seeds for it,” Lo Giudice said. However, implementing AI use cases in corporate insurance is not without risks and challenges, according to Siddiqui. For example, AI could result in less differentiation between risks, if corporate insurance moves towards a more homogenised approach that lumps insureds into pools of risk that are priced accordingly, he warned. The Lloyd’s market has already seen a move to algorithm-based underwriting, where AI is used to automatically select and underwrite risks with limited oversight by underwriters. “The use of data is great because it is efficient. But it does create pooling of behaviour that says these risks look the same so we will price them the same,” said Siddiqui. “Does that mean all your risk mitigation work goes out of the window? I hope not. It will still be taken into account. But it may mean that in the future there might be – and this is speculation – less differentiation,” Siddiqui said. With increased underwriting efficiency, buyers might also expect lower pricing as savings are passed on. However, this will probably not happen, according to Siddiqui. Insurers will need to recoup their investment in AI, while pricing is driven by hard and soft markets, he said. Lo Giudice warned that AI will also generate hidden or “silent” risks that will only emerge and evolve as implementation reaches a critical mass. “This risk will happen and will appear in all classes of business. I call this silent AI. It’s there, it’s going to happen. But we don’t know where it will happen and how. So we need to raise awareness within the insurance industry, and also with our clients. If the insurance industry gets caught by surprise, and is unprepared, ultimately the people who suffer are our clients,” said Lo Giudice. There are also limitations on the capabilities of AI when analysing risk, according to the webinar participants. “You can’t use this technology as a crystal ball yet,” said Lo Giudice. “We still look at the world around us and apply a human element to it for horizon scanning. But where the technology is very powerful is to provide some insights into counter factual [what has not happened]. You look at what almost happened and use this technology to give an idea about what might happen. Counter factual is the perfect application for these tools at the moment,” he said. Barbara Mayer, risk management expert at SAP cautioned about using AI – which is trained on the past – for emerging risks. There is a high risk of low-quality data in large language AI models, she noted. “Don’t use normal LLM without special architecture for any risk management purposes,” she warned. Siddiqui also warned about the limitations of AI. “This technology can be used to predict and look at long-term risk. But for horizon scanning and looking at emerging risk, it still needs a lot of human touch. I don’t think yet that ChatGPT has evolved to a level where it can do that. What it is good at is looking at data, not predicting but forecasting a few scenarios,” he said. The adoption of AI technology by the insurance market will not remove the need for human interaction when it comes to complex corporate risk and specialty risks, said Lo Giudice. “Our role as professionals is to remain human. Let the machines do what the machines are good at, and enable the humans to do what humans do best. This means the transaction of insurance should move into a world where it is streamlined and more efficient, and it is transparent,” said Lo Giudice. “You will still need humans to be humans. No client will ever be happy to talk to a machine when they speak to their insurer,” he added.

Ferma members told AI will bring risk and reward for corporate insurance buyers

Submissions should become more standardised and deliver faster response

Want to read this article?

Read Next

AXA P&C increases premiums by 7% as rates rise at AXA XL

Hong Kong and Shanghai to cooperate on captives and ILS

Swiss broker Kessler names head of special risks

Supply chain cyber risk concentrated in 15 top vendors with ‘below average’ cybersecurity

US insurers expected to increase casualty pricing as social inflation rises

AXA P&C increases premiums by 7% as rates rise at AXA XL

Hong Kong and Shanghai to cooperate on captives and ILS

Swiss broker Kessler names head of special risks

Supply chain cyber risk concentrated in 15 top vendors with ‘below average’ cybersecurity

US insurers expected to increase casualty pricing as social inflation rises

Want to read this article?

Read Next

AXA P&C increases premiums by 7% as rates rise at AXA XL

Hong Kong and Shanghai to cooperate on captives and ILS

Swiss broker Kessler names head of special risks

Supply chain cyber risk concentrated in 15 top vendors with ‘below average’ cybersecurity

US insurers expected to increase casualty pricing as social inflation rises

Related Articles