Insurers’ AI transformation is hitting structural limits, as scaling AI systems collides with technical debt and unclear accountability, insurance executives said on the opening day of the InsureTech Connect Asia (ITC Asia) conference in Singapore on Wednesday (1 July).
“The state of our industry is not as healthy as we think it is,” said Will Ross, founder and CEO of AI-native commercial insurance platform Federato, during his keynote address. “This industry has survived many trends through many times; the fact that it is resilient does not mean that it is efficient.”
Insurance professionals are highly exposed to automation, but labour costs only represent 8% of carriers’ expenses, Ross said. “We have to actually be looking more holistically, at where the efficiency really comes from, because in our industry the inefficiency is not human, it is not process, it has to do with the basic idea of allocation between capital and risk.”
The path to AI transformation is likely not to be “bolted on”, Ross added. “We are going to continue to confront the basic challenges that prevent us from moving forward — incentive misalignment, technical debt, and dated workflows, functional fragmentation,” Ross said.
Understanding how AI works and knowing the basics is not enough, said Rajnish Pal, SVP and head of transformation at MSIG. The general insurer is taking a three-pronged approach to reduce operational cost and drive efficiency, target core profitability, and make its products more accessible to customers.
“Pretty much every single role — whether it’s a risk manager versus an actuary or a compliance person or ops person — kind of knows where to focus on leveraging the basics,” Pal said during a panel discussion on the measurable business impacts of AI. “Now, does it really move the needle commercially? It does not.”
The right angle
Speaking at a panel on operations transformation, Sourabh Chitrachar, chief technology officer at MS First Capital Insurance, said that insurers and insurtechs should avoid looking at only one process or problem, rather than “looking at boiling the ocean”.
“What we should look at is not introducing tech for the sake of introducing tech, but first, looking at what is the problem statement,” Chitrachar said. “What are we trying to solve? We talk about AI being a magical solution for everything, but sometimes that’s not the case.”
AI has been an overused buzzword in the industry, admits Maya Lee, Chief Operating Officer for Asia Pacific at Sompo. “Yes, AI is a really big topic, but I think it really is difficult to do in the right way and also find the right use case,” Lee said.
Insurers have been focused on using AI to better automate or digitise processes, but AI can also be used to make insurers more proactive in serving policyholders and improve customer experience. “By using AI, insurers can predict the next best action from the customer,” Lee said at the same panel. “Instead of reacting to customer requests, proactively serving the customer by using AI — that’s maybe a use case I will think about scaling.”
In a separate panel discussing insurance personalisation, Sachin Dutta, chief operating officer and director of Technology at Canara HSBC Life Insurance, said that AI solutions should be seen from the customer’s point of view.
“(It’s) very important to reach out to the customers, to connect with the customers, to sit with the customers to understand their perspective,” Dutta said. “Customers are not wanting AI at the end of the day; the insurers are wanting AI to make it better for the customers.”
Technical debt
74% of carriers globally are still running on systems that can’t quickly respond to demand, said Federato’s Ross.
“We are an industry of 20 to 40-year-old systems, and with new technology change, while it’s a pleasant idea to believe that all the advantages of AI can accrue without actually addressing some of these underpinnings, it is a misleading one,” Ross said.
It’s a reason why AI can’t fully solve portfolio management, said MSIG’s Pal. “Long-term better risk selection, portfolio management… are hard to do in today’s world, because of a lack of information, disjointed information,” Pal said.
“A lot of stuff sits in PDFs, complex Excels, and not on our system. Systems could be 15 years old and legacy, it’s hard to extract, so that’s where the value would get unlocked.”
In order for AI to better resolve distribution and capacity issues, insurers have to put AI to work with filtering potential business, Pal said, and that’s where insurers can unlock value.
Structural transformation
Beyond legacy systems, however, insurers are also facing issues with adapting their organisational structures, procurement processes and culture to AI.
From a compliance perspective, insurers have not moved to change their organisational charts in response to how fast AI is moving, said Partha Rao, co-founder and CEO of Pints.ai.
“I think the challenges really are in terms of who’s accountable for the decision that is taken by AI, and that decision accountability has not caught up with the pace of AI as an industry itself,” Rao said. “Who’s responsible for the AI output? Is it the underwriter, the CTO, is it the compliance head? And this round robin, it delays to a point where the technology landscape completely shifts.”
Procurement cycles have also not caught up, Rao argued. “It still takes 6, 12, 18 months for a new vendor to be onboarded for a new use case, (and) everything changes in that particular period,” he said. “Imagine where AI was 12 months ago. You wouldn’t want to be evaluating a proposal which started 12 months ago to be onboarded today.”
In moving through sophisticated systems that have underlying dependencies, inefficiencies will compound, said Federato’s Ross.
“The reality is that we cannot continue to pretend that the underlying systems that we have from a capital allocation perspective, from a human perspective, and from a process perspective, can remain completely unchanged while making these changes,” Ross said. “You can’t draw a new future on a broken canvas.”
