We have the playbook. But are we ready to recommit to learning the lessons of past technological upheaval?

I didn’t realize I was walking into a firestorm back in January 2012 when I joined an insurtech company. But that’s what it felt like. My job was to win over insurance firms by demonstrating the power of predictive analytics to transform underwriting by combining data science with human expertise. 

As the insurance landscape cycles through the hype and hope of artificial intelligence, I feel we’re walking into that same kind of combustible situation I faced then. The resistance we encountered from the industry 13 years ago at Valen Analytics was fierce. Similar patterns are forming again, but now the term “predictive analytics” has been replaced with “AI.” 

The expressions go like this: “You’re trying to take my job!”, “Our data is garbage anyway”, or “It’s all garbage in, garbage out, right?” My favorite: “This is Moneyball, and I’m a talent scout (a.k.a. underwriter) that doesn’t need Jonah Hill.” It’s a reference to the 2011 movie starring Brad Pitt as Oakland A’s manager Billy Beane. Hill portrayed a Yale-educated assistant operations manager for a rival team whose advice on how statistical analysis can supercharge the scouting of talented players at a lower cost was minimized until Pitt’s character realized that this was a (pun intended) game-changing approach. 

To me, the industry’s current struggles and uncertainties over AI is good news. After all, we’ve learned these lessons before. And there’s a proven playbook for getting AI right.

Why insurance companies still wrestle with innovation

Before we dive into solutions, let’s acknowledge an uncomfortable truth about the insurance industry: it doesn’t face competitive pressure the way most industries do.

In tech, retail, or manufacturing, being average is a death sentence. In insurance, it’s called stability. Property casualty insurers are so well-capitalized and make so much money on investments that they are not forced to innovate. The entire world needs insurance to own a home, drive a car, or run a business. We actually need these companies to be reliable, resilient, and therefore conservative;  experimentation is easy when lives and livelihoods are not at stake, and innovation is a luxury afforded only in safety.

But this philosophy, which pits safety against innovation, creates a laggard culture. That’s brutal for anyone trying to introduce new technology. I’ve told C-suite executives countless times, “You don’t understand how much you’re missing by not supporting a thriving tech ecosystem.” 

Meanwhile, startups sit in front of their boards explaining why, despite having an amazing product, they can’t crack the insurance market. And investors don’t want to hear “the insurance industry is slow to adopt,” they want results. This structural resistance means that anyone attempting to transform insurance operations needs to understand what they’re really up against. And it’s not just technical integration.

Three types of resistance facing insurtech players

When we introduced predictive analytics at Valen, we encountered three distinct forms of resistance. Today’s AI implementations face precisely the same challenges.

First, there’s the expertise threat. Underwriters are proud professionals who’ve spent careers developing judgment and intuition. When you introduce AI-powered risk scoring, they hear, “Your expertise doesn’t matter anymore.” 

One underwriter told me flat out: “My brain is my tool. You are insulting my expertise and coming after my job.” Here’s the Moneyball problem in action: scouts who believe talent evaluation is an art, not a science. Insurance and baseball, they’re not so different.

Second, there’s the data quality deflection. Even underwriters intrigued by the possibilities would say, “Our data is such a mess, there’s no way you could possibly make sense of it.” Sometimes this was a genuine concern. More often, it was a convenient excuse. 

Third, there’s the sacred cow problem. If you ask insurance companies about their identity, they’ll talk about underwriting. That’s where their pride lives. By daring to transform underwriting with technology, we were attacking the very heart of what made these companies special. In a heavily regulated industry, that’s an enormous threat and one not easily assuaged.

Understanding these three resistance patterns is essential, because you can’t overcome objections you don’t fully understand.

The math that changes everything

The best way to reduce fear and uncertainty is to provide clear, trustworthy information. In Valen’s case, we published research showing that underwriter expertise plus predictive analytics was 3x better than an underwriter working alone.

Let me repeat that. If an underwriter made decisions solo, we’d get baseline results. If we let the model decide completely, it performed better than the human. But when we combined human expertise with the model’s insights, the results were three times better than the underwriter alone. This is the equation that matters. AI amplifies human intelligence, it doesn’t replace it. That statement has quickly become a cliché across all industries adjusting to AI. But just because everyone’s saying it, doesn’t make it less true. In this case, the numbers prove it.

In devising our study or predictive analytics’ impact, we provided a risk score from one to 10, where one meant low risk and ten meant high risk. You could still write high-risk business; you just needed the right price. 

For a score of eight, we provided a pricing indicator, say 1.3. That meant this policy was expected to perform 30% worse than the book average. The underwriter could then make an informed choice: charge a 30% debit on the premium to write it profitably, or grow the price over time if the market wouldn’t bear an immediate adjustment.

The critical part? We had underwriters send us every example where they disagreed with the score. As companies contributed years of historical data plus ongoing changes, we could compare predicted versus actual results. The predictions improved continuously, creating a self-reinforcing cycle of better data, better models, and better decisions. Models aren’t perfect. But they’re more accurate than not, and they get better over time. 

Let the math do the math. That simple principle tears down resistance, and allows underwriters to turn their brains up, not off. It allows technical teams to get closer to the business and the customer. And it creates entirely new roles, like underwriting managers whose job is managing bots, not just people.

Who says ‘yes’ first, and why it matters

At Valen, we learned that two types of customers would overcome their resistance to adopt predictive analytics first.

The first group was companies in genuine trouble. When workers’ compensation insurance had combined ratios of 113 or 115 – as bad as commercial auto today – those carriers had urgency. They had to try something different. Performance problems made them willing to share data and experiment. And there was so much low-hanging fruit, we could demonstrate dramatic results pretty quickly.

The second group was true early adopters. These were companies determined to stay on the leading edge. These organizations were open to experimentation. And they were especially willing to collaborate on building solutions. We learned enormously from them, and they became powerful case studies.

Once we built our data consortium to a certain size, something shifted; let’s call it the “FOMO factor” (“Fear of Missing Out”). 

The middle of the market – all those companies that had been sitting on the sidelines – suddenly became interested. Our dataset had become a genuine competitive advantage. They realized that not playing in this space would put them at a disadvantage.

That’s the perfect pattern for AI adoption too. Target the motivated, deliver results, build network effects. And let FOMO build momentum.

What both sides need to get right

For insurance companies embracing AI, success requires more than just a sound budget and good intentions. You need a champion at the top. It can be either the CEO or the head of a department with a P&L. But it has to be someone with enough conviction to weather the inevitable resistance. When people come at them insisting “this doesn’t work” or “the score is wrong,” that leader has to hold steady.

You also need to start with the specific business problem you want to solve, think beyond the attractions of the latest “shiny new technology.” Where are your pain points? What problems are you having trouble solving? Answer those questions before you ask whether AI is the right tool for you.

And here’s a counterintuitive tip: involve your biggest skeptic early. 

Find the influential person on the floor that everyone watches, the one who’ll poke holes in everything. Bring them in. A doubter who becomes a convert will be your strongest advocate.

Finally, for tech companies trying to serve insurance, domain expertise is non-negotiable. You can’t just be general technologists anymore. You need to speak the language of insurance fluently. You need to deeply understand the constantly changing regulatory environment. And you need to prove beyond all doubts that you’re squarely addressing real business problems, not just promoting features. We invested heavily in education at Valen. It worked because we became genuine partners to the industry, not just vendors.

The urgent reality

If you’re an underwriter reading this and thinking, “I’ll wait and see,” I have news for you. If you fought predictive analytics back in the day, you didn’t win that one by waiting and seeing what your competitors were approaching it.

Every underwriter I know complains about “busy work” that keeps them from “real work.” AI can give them back that time. It can help them make more consistent decisions backed by better data. It can let them focus on complex cases that genuinely require human judgment.

The question isn’t whether AI will alter underwriting operations. It’s whether you’ll be part of shaping that inevitable transformation or whether it will just happen to you. We’ve all been here before, and not just in 2012. And the playbook has existed for as long as any industry has dealt with tech advances. The only question now is whether we’ll learn from it.