Insurance copilots could aid in underwriting, claims
Underwriters often have hundreds of pages of documents to sort through but maybe not for long. Insurance carriers are deploying generative AI in the form of copilots to back-office operations.
Copilots can act as a chatbot, virtual assistant, search engine and productivity tool.
Sandee Suhrada, principal, at Deloitte said this technology can be used across the value chain in insurance with product development, innovation, sales and marketing, underwriting and claims.
“The most traction is in three areas. One is underwriting. Second is claims and third is personalized servicing. A lot more of our clients are building underwriting copilots,” Suhrada said.
Suhrada explained one use case. “One of my clients has actually already built an underwriting copilot for their Spanish specialty insurance business line. And, the reality is, it’s actually a huge lift for the underwriters. So as they actually do their day to day it has helped streamline the underwriting process, because it can help you process vast volumes of data like customer information, medical records, risk factors, and then it provides you the insights into risk assessment, the pricing model, and staying in line with underwriting guidelines.”
The copilots are built on large language models.
Indranil Bandyopadhyay, principal analyst at Forrester, said there are several things happening under the hood.
“When you write something in a language that we understand as human beings, it needs to be processed first, we call it natural language processing,” Bandyopadhyay said. “Then, it needs to be understood and then it needs to fetch the information. It is not just a large language model. It needs to still have an API or query or an old-fashioned ID to go fetch it. And then the biggest advantage of the large language model is that it can respond to you in a human way. It’s a two way traffic.”
There is some heavy lifting done by LLMs, Bandyopadhyay added, but insurers still need an omnichannel strategy and 24/7 operations because a copilot or an LLM will not be able to help in terms of customer experience and better engagement without those.
“Once those things are there, then this can help you. … LLM in this context is the icing on the cake. So if you don’t have a three- or five-course dinner plan, and if you don’t have dessert, and within dessert, if you don’t give the cake icing, it is not going to save the day,” Bandyopadhyay said.
Montoux, an insurtech with AI and actuarial modeling technology, launched Model Copilot, an AI copilot designed specifically for actuaries, in February.
Julian Balasingam, chief growth officer at Montoux, said via email: “Insurers are interested in using Model Copilot, our latest product, to augment the work of actuaries. It assists them in understanding, documenting and maintaining their models. This work is typically time-consuming, manual and is often described to us as tasks that actuaries don’t enjoy doing. Current AI Copilots are good at removing time consuming tasks and enabling insurers to deploy their high value resources to the work that creates strategic value for the business.”
Patrick Davis, SVP Data and Analytics at Majesco, said in early 2023 there wasn’t much of a framework for the idea of what a copilot is, but there was hype around Microsoft’s launch of its product. Majesco also released its own copilot in collaboration with Microsoft.
“We took a cue from Microsoft’s branding and what they were indicating a copilot was meant to do and then we had to figure it out and plumb it all ourselves,” Davis said. “You actually have to curate your data and craft it in a way that a large language model can understand. … We started off with the ability to interrogate complex insurance data.”
In Spring 2024, Majesco plans to introduce a large number of actions, automating parts of a user interface workflow such as copying quotes, canceling policy sets, and creating correspondence, Davis said. The copilot is integrated into the Majesco platform and each screen has an icon and a chat option.
Davis said this technology will most likely never replace underwriters. “There’s a lot of nuances to what an underwriter thinks about and how they price things. We see generative AI and copilots, in general, as being useful throughout the insurance lifecycle but also in claims and underwriting. … I think it can save underwriters and claims agents a lot of time because it can look at a 20-page document provided by an agent and answer questions on it, summarize claims history based on one question or interaction with the copilot.”
Davis added that copilots help with task-based roles.
“If you think of all [the] financial technology institutions in this world, whether it’s a bank or an insurance company, they probably have significant back office operations, like clerical work and sending emails and doing stuff that is kind of low-hanging fruit,” Davis added. “I think as these technologies come into play, it’ll help make the people that [insurers] already have more efficient. The repurposing of some of those back office jobs is probably going to be something that is the first wave that comes from generative AI in insurance.”
Michael Nadel, senior director insurance at Simon-Kucher & Partners, said that copilots and chatbots, for example, aren’t necessarily a new concept.
“Copilots have been around for a while with varying levels of maturity,” Nadel said. “Some of these LLMs are pushing it so much further than where it’s been.”
Nadel added that copilots can act as a knowledge transfer option too.
“To be smart on a policy within a carrier, historically has taken years to build product level specific knowledge,” Nadel said. “Copilots can take that upskilling and onboarding process down to almost no time because you don’t need to spend time researching it, you get all the questions on demand and you’re building that knowledge as you’re doing the job versus training that’s required to onboard as a customer service representative.”
