(Re)in Summary
• Underwriters need to adapt to changing customer expectations by exploring new data models and sources, said
industry practitioners at the Digital Insurance APAC conference in Hong Kong.
• Using new and emerging and technology is key, but must be balanced by privacy and security concerns, with guardrails being noted as key.
• AI should be seen as a tool to enhance risk management, not replace underwriters, but will require a different mentality and comprehensive change management.
• Customer-centric digital transformation is crucial, with a shift towards online transactions and immediate service fulfilment.
Underwriters must begin to leverage innovative tools and data sources as the insurance industry adapts to changing customer expectations, experts said during the Digital Insurance APAC conference in Hong Kong on June 27.
“It’s very easy to look back at risk that you have,” said Saibal Bhattacharya, Global Head of Insurance Products, Underwriting & Reinsurance at Bolttech.
“But as customer experience is driving us to cover peaks that we have not covered sideways, and getting into something unknown without a lot of historical data requires underwriters to ask, ‘What’s the essential product structure like? What’s the customer journey like?”
It’s important to continuously monitor and recalibrate based on emerging data, said Bhattacharya, as underwriters won’t have historical data to look back upon. “Once you have experience, you need to be sitting on top of the data and see how it is performing,” he said.
Saibal Bhattacharya
Global Head of Insurance Products, Underwriting & Reinsurance at bolttechCustomers are also demanding service rather than just paying a premium, he added. For instance, if a customer is experiencing fraudulent transactions on their card, they would expect a specialist who could tell them what to do.
“The key is, how do you differentiate your value proposition? Not just being the dollar value, but also to ensure that, at the moment of truth, you’re giving some actual service,” he said.
A fine balance
The shift comes as insurers explore various data models and sources to enhance their underwriting capabilities, but are facing concerns due to privacy and security risks.
“One side is really moving faster. Commercially, they want to launch different products. And on another side, they have a lot of concerns about their privacy, security and equally sensitive information. But there are solutions out there,” said Dr Chris Ding, Adjunct Professor at Melbourne University.
Different companies are developing and training data models through various sources, highlighting the ongoing debate in the AI industry between large service providers like Microsoft and AWS, who advocate for a cloud-based solution, and proponents of private AI or hybrid approaches that keep sensitive data on-premises.
Firms advocating for a private or hybrid AI solution often tout data security as a key selling point, said Ding. “What they suggest is that their company is financially available to insure this certain data model and data training in their boundaries.”
AI applications in insurance should be categorised by risk, said Ding. “If you look at the complexity of the data, you can probably start to understand why the industry is concerned,” he explained.
Chris Ding
Ddjunct Professor at Melbourne UniversityPrompt injections, for instance, can lead to attacks or misuse that could cause AI models to behave in unintended ways; understanding these risk levels remains crucial for the industry to address concerns about AI adoption effectively.
High-risk areas included applications that might allow for impersonation, direct prompt injection, bias or data loss.
Frameworks need to be built to control different layers of AI implementation, both from a risk mitigation or threat provisioning perspective, so guardrails can be properly formed, Ding added.
“The guardrails support, regardless of what sort of use cases you’re going to have, so you can always use this product to control and detect potential threats.”
Will AI replace underwriters?
As AI and machine learning technologies become more prevalent in the insurance industry, many underwriters are naturally concerned about the future of their roles.
But AI should be seen as a tool to enhance risk management, rather than a replacement for human expertise, said Dr. Porus Peshoton, Executive Vice President of Underwriting, Claims and Reinsurance Admin at Max Life Insurance Company.
“Oftentimes underwriters have this irrational fear that AI is going to take over their jobs,” said Peshoton. “(But it) makes their risk management stronger. Most importantly, it gives the right decision to the customer.”
Dr. Porus Peshoton
EVP of Underwriting, Claims and Reinsurance Admin at Max Life Insurance CompanyTo get underwriters comfortable with using AI and understand that it can help with their work processes, comprehensive change management is needed, said Peshoton. “When we started underwriting, we had very limited tools, and (underwriters are) used to that pattern.”
Crucially, underwriters will always be here to audit, to “make sure the system’s decisions are also accurate,” said Peshoton.
They will also be needed to inform regulators of what advancements might be available.
“Engaging with the regulator, showing him the benefits more from a customer perspective and risk management perspective – I think that is also an important aspect when we are adopting all these technologies and models,” said Peshoton.
Follow the customers
With a shift in digital transformation comes a shift to customer-centricity, as customer behaviour and expectations rapidly change.
Such a shift has come in several forms, including medical underwriting.
Max Life Insurance uses machine learning for diagnostic centers’ fraud detection, which Peshoton said has detected “mirror reports”—fraudulent reports generated without any blood tests done—which account for 2.6% of all reports received.
“While the number may seem small, it’s a huge impact when you price your product right,” he said.
Max Life shares this information with other insurance companies, so diagnostic centres participating in fraud feel the pinch. “The model is very effective at scanning malpractice at a broad industry level,” he said.
Saibal Bhattacharya
Global Head of Insurance Products, Underwriting & Reinsurance at bolttechUltimately, there is a need to reimagine the entire insurance process from a customer-centric perspective. “Customers want real-time processing. They don’t want to fill up lengthy forms,” said Bhattacharya.
Research reports have indicated that more than half of all banking transactions have moved to the internet, and 60% of retail sales going into e-commerce.
“We have medical facilities, primary businesses going into telemedicine, so that is the customer expectation,” he added. “I want my milk by 6pm, I will order it on Amazon and get it right now. (In much the same way,) I want my policy right now, and I get speed.”