What would you do if you had PhD-level capabilities in every function of your insurance business?
That’s the question Marc Low, director of innovation and emerging technology at KPMG in Canada, asked attendees during a keynote session at KPMG’s recent 2024 Insurance Conference in Toronto.
“What could you achieve as an organization?” he asks. “It’s superpowers for every one of your employees.”
In the 18 months since ChatGPT 3.5 arrived on the scene, the generative artificial intelligence (GenAI) platform wrote the LSAT (Law School Admission Test) entrance exam and originally scored in the 65th percentile, Low reports. ChatGPT 4 scores in the 99th percentile — “better than every human who wrote that same test.”
The latest ChatGPT model, o1, is now doing PhD-level reasoning “better than any human in that domain,” Low says. “The thing to understand is if it’s doing PhD-level reasoning in one domain, it’s going to do PhD-level reasoning in every domain.
“Meanwhile, the cost to use these tools declined more than 90% — the most deflationary technology we’ve ever invented.”
GenAI is a top priority for 81% of insurance CEOs in Canada, according to a recent KPMG study. The survey also finds 90% of insurance company CEOs in Canada feel GenAI is going to be either ‘very important’ or ‘extremely important’ to their competitive advantage moving forward.
Related: Meet your new underwriting assistant: GenAI
One dollar spent results in at least $3 of return, with some organizations seeing as much as eight or 10 times return on that money, Low adds.
“And what we’re already seeing is that it’s impacting every part of the value chain inside insurance — front office, middle office, back office,” Low says. “Whatever part of the business you occupy, generative AI is a lever for you.”
As an example, Low points to one financial services company that uses GenAI to help answer inbound customer queries. “They’ve reduced the volume that goes to the actual agents by 55%; no drop in quality.”
Other examples from two life and P&C insurers include one that uses it for claims processing to allow agents to do more with higher accuracy. The other uses GenAI to reimagine back-office IT and push Google tools into every part of their business.
Looking ahead, certain factors will improve the speed of GenAI.
Currently, the context window — essentially the amount of information pushed into the tool to then perform analyses — is relatively limited and the software forgets part of the interaction, similar to how the human brain works.
“Well, the context window is going to radically expand, meaning you can upload, for example, every piece of customer interaction you’ve ever had, every piece of insurance legislation…and then use tools to query it and create insights,” Low says. “If the tools have infinite memory, that means they become very, very useful for the folks inside your organization, because they contextualize everything they’ve done before when they ask a new question.”
Another factor is the concept of agents: actions that can be taken autonomously or semi-autonomously on your behalf. For example, a client comes to your website and generates a lead, the lead gets updated and scored in the CRM, and your pipeline and financial information gets updated in real-time, all driven in the background by GenAI.
These AI models are improving exponentially, with Low predicting a 3,000-times to 100,000-times improvement in capabilities by 2029.
“What I get very excited about is the idea that it’s going to radically accelerate the work that every legacy organization does…”
Feature image by iStock.com/wenjin chen
