AI: Finding the balance between risk and reward

Generative Artificial Intelligence (AI) will create big opportunities for companies but also raises the prospect of aggravating some of the risks they face today, participants in the French chapter of our Risk Frontiers Europe 2024 survey say.

They are particularly worried about the prospect of facing new risks that are still not well understood, and consequently not covered, by the insurance market.

“AI will be useful for the organisation of logistics, but it may also be at the core of its disruption,” said Frédéric de Serpos, a member of Amrae’s board. “We can imagine that we will see increasingly automated supply chains in the future thanks to artificial intelligence. If they collapse or are attacked by cyber terrorists, a whole industry can be affected. Which raises questions about the insurability of that risk.”

One example of an emerging risk is the introduction of autonomous vehicles that have the potential to greatly increase the efficiency of supply chains.

“Autonomous vehicles are the link that will lead us to the automation of logistics. Will we be able to insure the whole of this chain, knowing that there will be no loss history data about autonomous vehicles and insurers will have to deploy AI to price those risks?” De Serpos said. “I see insurers working on it, that they are stepping on the gas, but I feel that innovation still moves faster.”

The tantalising blend between opportunity and risk from generative AI enhances the role that risk managers must play as their companies go through the inevitable process of adopting the new technology.

“Generative AI is an interesting topic because some are seeing it as the innovation that will transform all activities, and some are seeing it as the scariest thing that can happen to the world. Our role as risk managers is to find the right balance,” said Amrae president Oliver Wild. “If used efficiently, generative AI will simplify many human tasks, especially white-collar tasks. But I do not believe that AI can work only on its own. It can produce absurd results, and so it requires human activity and human intervention.”

“I do not see risk managers as obstacles, but more as enablers,” he continued. “If we know our risks, or are aware of the risk environment, we will be able to make the best of opportunities, without jeopardising the whole company. So the idea is not to stop the business and say that we do not want to use a technology at all. It is more about developing, by test and trial, the best strategies.”

The introduction of generative AI in companies’ daily activities goes well beyond simply using Chat GPT. Companies must first decide how and when the new technology can be useful for their organisation.

“For the moment, AI is not relevant for our activities,” said François Malan, a member of Amrae’s board and head of risk management at Eiffage, which works in sectors including construction and energy. “We are working on it to imagine what AI could do for us. It is clearly an opportunity to support our operations, but it should also be managed because it could be a threat.”

He cites themes such as ethics, the GDPR and cybersecurity among the main points of concern linked to generative AI.

Insurer SMACL Assurances, which specialises in local government risk in France, has already started to implement the new technology, according to Marie-Elise Lorin, the firm’s risk manager.

“We see AI as an opportunity to fight fraud and to optimise our contact with local governments,” she said. “We have a tool that synthesises the insurance documents presented by local governments, which sometimes reach up to 400 pages, and also to moderate the parts of our website that are open to comments by the public. AI can be also employed to target customers in the underwriting process and to manage claims.”

At the same time, SMACL is aware of the risks entailed and is already on the lookout for ways to mitigate them, even though they may loom far in the future.

“We have invested in start-ups that work with AI, and we have considered the risk of being cannibalised by new companies that come out of that area,” Lorin said. “But we have concluded that it is too early for that, as there are still barriers that prevent an AI company from becoming an insurer.”

Wild points out that many companies are developing their own chatbots and ring fencing them to make sure that the information will not leak out of the organisation. This is the kind of careful approach that is required by a technology that is in a constant state of flux and about which companies are still learning.

“AI is going to happen, but we need to know more about it. I think a lot of people do not actually know what it means, what it can do and how far it can go,” he said. “Even some of the people developing generative AI are actually scared of how fast it is moving. So we need to be careful.”