The names Pets.com, Webvan, Boo.com, WorldCom, NorthPoint, Global Crossing, MP3.com, and PeopleSound may ring vaguely familiar to people of a certain age. They’re just some of the companies consigned to the dustbin of tech history by the popping of the dot-com stock market bubble in 2000. Everyone with an investment portfolio lost money, except Warren Buffet, the Sage of Omaha, who didn’t like the look of tech stocks.  

When he pronounced against the internet-driven market mania, people said he had lost it. But on 11 March 2000, when the Nasdaq had fallen 78%, he once again seemed extraordinarily wise.

The AI investment frenzy aside, even those with only a passing interest in markets are eager to see if Greg Abel, the man called upon to fill Buffett’s size 50 triple-X Brooks trainers, can perform with equal panache. Meanwhile, Berkshire Hathaway underwriting guru Ajit Jain continues to lead the insurance side of the company which, according to AM Best, is the world’s largest re/insurer, with non-banking assets of $1.15 trillion in 2024. Jain’s second, Charlie Shamieh, is mooted to take over in due course, but in the interim the impressive Mr Jain has the conn.

So when Jain told his shareholders this week that AI is “very fashionable right now” among insurers, but is unlikely to be useful for anything more than “routine repetitive things” anytime soon, many will have listened. I suffered a physical spasm brought about by internal contradiction. I know it’s a fool’s choice to doubt Buffet, even when embodied by Jain and his incredible underwriting track record. I also know that on this occasion, he’s wrong.

Let me start by saying Mr Jain is right. AI is very good at tackling routine repetitive tasks. But he seems to have swept aside the benefits of that win, when in fact they are enormous. Any ideas how much humdrum processing work currently costs the global insurance industry? Hint: it’s a lot. How much does it cost Berkshire Hathaway? Hint: it’s a large share of a lot.

Worse, though, Mr Jain doesn’t seem convinced that AI will have applications beyond the basics. Perhaps he’s stuck at AI Level 1. “I don’t think AI will reach a point where you can make a [human vs. AI] trade-off on things like pricing, settling a claim,” he said. “That is still years away.”

To take the last point first (settling a claim), I invite Mr Jain to take a meeting with Reserv, the AI-driven TPA which has just enjoyed a $125m Series C fundraising led by KKR, “to advance AI-driven transformation of insurance claims.”

Reserv says it is “automating the mundane and humanizing the complex for adjusters to focus on the most important tasks” using its proprietary, AI-powered claims technology which “delivers rigorous analysis in partnership with our adjusters” through the claims lifecycle. What they do works. They’ve made claims dramatically better (ie faster, cheaper, and better matched to actual loss) for everyone.

Reserv has accomplished what the original insurTech ‘disruptors’ thought would be easy: a revolutionary approach. They have succeeded where other failed because they have reinvented from the ground up the processes which drive only a single function, making them suited to the use of AI-driven, data-first technology, rather than attempting to upend the whole sector in one go.

Meanwhile, on the consumer side AI has already been widely deployed in claims-fraud detection. That has stifled illegality and created a significant saving which allows carriers to reduce premiums.

Similar gains are being made on all fronts, from underwriting and pricing to client service and product design. Successes occur when processes are redesigned to allow AI and other tech tools to do things efficiently, rather than simply to do the things the same way people are already doing them, just better and faster.

A huge insurance survey by CapGemini was published last week. It found that “the use of AI [by P&C insurers] is commonplace and indeed, 40% of P&C leaders say AI is meeting their expectations.” However, it also found that “many insurers report only marginal gains in cost savings, revenue growth, and time to market.”

To make the big gains, we have to do more than file the corners off the square pegs. AI implementations must be bold. Processes must be reinvented to take advantage of the abilities of AI, not simply bent slightly to fit with existing workflows. Even before AI, we called this the ‘data first’ approach. Fiddling at the edges will not put data at the forefront, where it needs to be.

I am hopeful that’s what Ajit Jain meant. AI will not allow human vs AI trade-offs on pricing and claims settlement as long as processes remain unchanged. I am sure Mr Jain understands that when we adopt data-first processes, huge amounts can be sliced from expense ratios, and, even better, claims ratios can be reduced because new human-driven, AI-enhanced underwriting processes are better at rating.

If I am mistaken, perhaps Charlie Shamieh gets it.

* Like every Insurance Technology Diary entry about AI, this one is accurate only to the best of my knowledge at the time of writing. The pace of AI progress is so great that I cannot guarantee it remains so now that it’s finished, let alone when you read it.

Guillaume Bonnissent is CEO of Quotech.