“We spend all this time talking about AI and innovation for the underwriter, but, as far as I can tell, the only thing we’ve done is continue down the path of the dumbification of underwriting.”
– Roger Paré, Retired CUO, Hanover Insurance, Industry Advisor
I spend most of my waking hours thinking about the intersection of AI and P&C underwriting. A rough cocktail napkin calculation in a Delta Sky Club suggests I’m just eclipsing my 10,000 hour mark. I still have more questions than answers, but what I can tell you is that these simple words the dumbification of underwriting have given me new clarity.
Federato will soon be announcing a suite of new capabilities available on our RiskOps Underwriting Platform. And, as always, our team has been putting in the effort to make sure we are giving insurers and their underwriters capabilities they can trust. In these times, I often sit down with people like Roger, quoted above, when we’ve already done plenty of user research, but things are still a little foggy and I’m struggling to see the matrix.
And so, a few weeks back, I was explaining a new feature to Roger – its basic premise is using AI to generate relevant questions based on hundreds of pages of underwriting guidelines in the context of an account. We’d heard great things about this feature from underwriters and teams:
“This is perfect! Asking questions efficiently is the core of underwriting.”
“As an underwriter, there are a million things you could ask about a risk. This helps you to discover what you should ask.”
“We have to choose our own direction and stop using shared risk scores as a crutch.”
The enthusiastic response from actual users boils down to a common theme: Finally, a solution provider was using these technologies to do something that wasn’t focused on limiting an underwriters’ analysis or telling them how they should think about a risk. We weren’t dumbing it down.
There’s a common, but perhaps tired, adage in underwriting technology that value lies at the intersection of art and science. It has the right spirit. If you’ve spent enough time in the space, you’re unlikely to believe that a purely algorithmic approach can outperform the instincts of a seasoned underwriter. But sometimes it takes an insurance veteran like Roger Paré to remind us that, despite our good intentions, we really have been complicit in that dumbification.
Of course, AI can automate manual tasks (those countless time-consuming tasks underwriters would rather not be doing in their day to day) and streamline the underwriting workflow, but as an industry, we’ve been focusing a huge percentage of our time on automating away expert human analysis or attempting to conform an underwriter’s thinking to baseline risk scores and averages. For example, any underwriter will tell you that there is a vast difference between an earthquake hazard score of "6" in California versus a "6" in New Jersey, but baseline hazard scores tend to put them both in the same bucket, which doesn’t exactly exemplify the ‘intelligence’ in artificial intelligence.
If we really believe that our uniquely human creativity and knowledge has a valuable role to play in insurance (the ‘art’ in the underwriting equation), doesn’t that mean we’ve inherently fallen short? Could it also mean that the best of technology and the best of AI in underwriting still lie ahead?
In some ways it would be easy to feel defeated after stumbling into this change in perspective a decade into working with AI – admitting to oneself that we’ve spent more time dumbifying than we have adding value. But if we look to the past, there is ample reason to be optimistic: While the internet started in 1983, it wasn’t until 1994 that the Mosaic/Netscape browser made things real, spurring thirty years of innovation. Most would agree that contemporary AI (i.e., Deep Learning) has its roots somewhere around 2011. And yet here we are, just about a decade later, and ChatGPT has finally made it real and ubiquitous. History would suggest, insurance’s “AI Moment” is only just beginning.
Countless individuals are just now feeling, touching, and experiencing these technologies, a decade in the making, for the first time. Creativity and human ingenuity have been unleashed in new ways! And I couldn’t be more excited to keep forging boldly ahead on AI capabilities that just make sense for the underwriting community. Onwards!
Federato, together with Google Cloud, will demonstrate these new underwriting capabilities live at ITC Vegas, the world’s largest gathering of insurance innovation, October 31 – November 2, 2023, Mandalay Bay, Las Vegas. Visit Federato at Booth #3341 and plan to attend the breakout session “Tomorrow's Underwriter – Building Trust in AI, Generative and Beyond” on November 1, 11:10 am – 12:00 pm PT (Location: Mandalay Bay Ballroom J).
Not attending ITC Vegas this year? No need to feel that fear of missing out! Schedule a live 1:1 demo at your convenience to see the full suite of AI capabilities in our RiskOps Underwriting Platform: https://www.federato.ai/demo.