What Classic Business Thinking Still Teaches Insurance Leaders About AI
Editorial | James W. Moore
Key Takeaways
- A small shelf of business books, written mostly in the 1990s, quietly became the operating system for a generation of technology executives, insurance included.
- Most of those frameworks have held up remarkably well. This isn’t an argument that they were wrong.
- AI is exposing places where the frameworks need to expand, not places where they need to be discarded.
- The more useful exercise isn’t rereading the books. It’s noticing which of their assumptions you’re still carrying without realizing it.
If you came up through technology in the 1990s, your bookshelf probably looked a lot like mine.
Crossing the Chasm. The Innovator’s Dilemma. Good to Great. Blue Ocean Strategy. The Long Tail came a little later, but it joined the same shelf. These weren’t just business bestsellers passing through an airport bookstore. For a lot of us who spent those years building technology into insurance, in IT departments, in agencies, at wholesalers trying to figure out what the internet meant for how we sold and serviced policies, they became something closer to an operating system. You didn’t just read them once. You absorbed them, and then you made decisions for the next twenty years without necessarily remembering which decisions came from where.
I’ve been pulling a few of those books off the shelf again lately. Not because they’ve become obsolete. Because AI made me curious whether I’d been carrying assumptions from them without realizing it.
Geoffrey Moore and the Chasm That’s Still There
Start with Crossing the Chasm, because it’s the one that still gets quoted in product meetings without irony, and it earns that.
Moore’s core argument was simple and it has aged well. There’s a dangerous gap between the early adopters who buy a vision and tolerate a broken product, and the early majority, the pragmatists, who buy a finished solution and need proof before they’ll touch it. Selling to one group does not teach you how to sell to the other. Most technology efforts don’t fail because the product doesn’t work. They fail in that gap, when a team keeps selling a vision to people who were never going to buy a vision in the first place.
Two pieces of the framework are doing real work in insurance AI conversations right now, whether people know they’re borrowing from Moore or not.
The first is whole product thinking. Moore’s point was that pragmatists don’t buy a piece of technology. They buy a solution to a painful problem, and that solution has to arrive wrapped in everything required to make it actually usable, not just impressive in a demo. In insurance, that whole product has gotten considerably bigger than it was even five years ago. It’s not just governance and audit trails anymore, though those matter enormously. It’s explainability that survives a regulator’s questions. It’s security postures that survive a reinsurer’s due diligence. It’s organizational trust, which is the hardest ingredient of all because you can’t buy it off a vendor’s roadmap. Moore wasn’t wrong about whole product. The product just got bigger than he had reason to imagine in 1991.
The second is the beachhead strategy, sometimes called Moore’s D-Day approach. Don’t try to cross the chasm everywhere at once. Pick a niche specific enough to dominate, win it completely, and expand from there. Be a big fish in a small pond before you try to be a fish at all in the ocean. I’ve watched more than one AI initiative inside a carrier or an agency stall for exactly the reason Moore predicted three decades ago, an attempt to transform everything at once instead of picking one workflow, proving it decisively, and letting the proof do the selling.
Where the model creaks a little is in how linear it assumes the crossing is. Moore wrote it as something you do once. Cross the chasm, and you’re in the mainstream market. AI doesn’t behave that way. You cross a small chasm with one capability, and six months later there’s a new one waiting, because the technology itself hasn’t finished arriving yet. Even Moore seemed to sense this eventually. His later books, Inside the Tornado and Zone to Win, were attempts to wrestle with a market that doesn’t sit still long enough for one clean crossing.
Christensen and a Surprise I Didn’t Expect
The Innovator’s Dilemma is a harder book to summarize honestly, because most people remember the wrong half of it. The common shorthand is that incumbents get disrupted because they ignore change. That’s not quite Christensen’s argument, and the actual mechanism matters here. Incumbents get vulnerable because they over-serve their best, most demanding customers, and in doing so they ignore the low end of the market, or an entirely new market, where a cheaper, worse product is quietly getting good enough. By the time the incumbent notices, the disruption has already climbed the ladder and is coming for the customers who used to be safely theirs.
For a long time I expected AI to follow that exact script in insurance. Nimble AI-native startups entering at the low end, undercutting the incumbent policy administration platforms, climbing the ladder the way Christensen’s model predicts.
That’s not quite what I’m seeing.
Instead, the existing ecosystem seems to be getting stronger. Duck Creek didn’t get disrupted from below, it built an agentic platform into its own core system and then went out and acquired an AI-native underwriting orchestration company to bolt onto it. And it’s telling that a specialty AI vendor like Convr chose to build its underwriting workbench on top of Guidewire and Duck Creek, not around them, explicitly betting that carriers will keep the platforms they have rather than rip them out for something newer. AI is being layered into decades of existing workflow rather than replacing that workflow wholesale.
The model is rentable. The data isn’t. That’s likely why the incumbents are proving more resilient than Christensen’s classic disruption model would have predicted.
That’s the part his original framework doesn’t quite have a slot for. AI isn’t always disruptive from below the way his model expects. Sometimes it reinforces the center instead.
The Rest of the Shelf
I don’t want to give the impression these were the only two books that mattered, because they weren’t, even if they’re doing the heaviest lifting in this piece.
The Long Tail is the one I’ve actually tried to emulate with this publication, more than any of the others. Chris Anderson’s argument, that the aggregate value sitting in the niche and the overlooked can rival or exceed the value concentrated at the top of the curve, is more or less the bet behind writing for a genuinely underserved corner of insurance media instead of chasing the same handful of headline topics everyone else is already covering.
Blue Ocean Strategy is still in there somewhere every time I catch myself asking whether a piece of AI adoption is really competing on the same terms as everything around it, or whether the more interesting question is what an uncontested space might look like. Good to Great is in there too, mostly in the stubborn belief that discipline, the right people, and a clear sense of what you can actually be the best at still beats cleverness most of the time. I doubt I’m alone in this. Everyone who read that shelf probably has one or two books they’d add that I haven’t mentioned. That’s sort of the point. This wasn’t one author’s influence on one executive. It was an entire generation’s intellectual scaffolding, built mostly by five or six people, most of whom never met each other.
The Questions None of Them Could Have Asked
None of these books contemplated foundation models, natural language as the interface itself, or agentic systems that take action rather than just producing output. None of them had a framework for intelligence that’s become cheap enough to embed everywhere, for software that improves on a monthly cadence instead of an annual release cycle, or for AI that increasingly governs and evaluates other AI.
Christensen’s framework doesn’t tell you what to do when the disruption is probabilistic. Moore’s whole product concept doesn’t have a native slot for “can we actually explain how this recommendation was reached.” None of them wrote for a world where the thing you bought last quarter has meaningfully different behavior this quarter, without you changing anything on your end.
So insurance executives are left holding genuinely new questions the old shelf never asked. Can an AI recommendation be trusted if nobody, including the people who built it, fully understands how it arrived there? Who owns an AI-generated underwriting decision when something goes wrong, the carrier, the vendor, or some undefined space in between? Can governance itself become a competitive advantage rather than just a compliance cost? How does a regulator evaluate software that’s a different piece of software by the time the exam is finished?
I want to be careful about how I frame this, because it would be easy to read it as an indictment of the old frameworks, and it isn’t one. These are entirely new questions, the kind nobody could have written a framework for in 1991 or 1997 or 2001, because the thing generating the questions didn’t exist yet.
The Mistake I Keep Making
Here’s where I’ll indulge myself a little, because this is an editorial and I’ve earned the right after forty years of watching this pattern repeat.
Every generation of technology arrives with the same unearned confidence that it has invalidated everything that came before it. The internet was going to make Peter Drucker irrelevant. It didn’t. Cloud computing was supposedly going to make Geoffrey Moore’s frameworks quaint. It didn’t. Mobile was going to retire Christensen. It didn’t. And I promise you, AI is not going to be the exception that finally proves the rule, no matter how many keynote speakers insist it changes everything.
What actually happens, every single time, is narrower and less dramatic than the hype suggests. The frameworks don’t get proven wrong. They get exposed as incomplete, because the world underneath them changed shape while nobody was looking directly at the framework itself. That’s not a criticism of Moore or Christensen or Collins. It’s an honest description of what a framework is, a simplification of reality that was true enough, for long enough, to be worth building a career on. Reality eventually moves. It always does.
The frameworks worked so well, for so long, that they stopped feeling like frameworks at all. They started feeling like facts.
I’ve made that mistake. I suspect most of us who came up on that shelf have made it more than once.
Where This Leaves Us
I’m not going to pretend I know which AI vendors win this decade, and I’d be suspicious of anyone in insurance who tells you they do with confidence. But a few patterns feel like they’re already forming underneath the noise.
Owning the workflow, not just supplying a tool that touches it, seems to be turning into the more durable position. Trust, the kind that takes years to build and one bad quarter to lose, seems to be worth more than it used to relative to raw capability. And proprietary data, especially the unglamorous decades of it that nobody thought to throw away, looks more valuable with each passing quarter, not less.
I’ll hold those loosely. Naming winners this early has burned smarter people than me.
Back to the Bookshelf
I don’t know which AI vendors will dominate the next decade in insurance, and I’ve made my peace with not knowing. What I do know is that the books on that shelf shaped how I think about technology and business more than almost anything else I encountered across forty years in this industry.
They still do.
AI isn’t proving them wrong. It’s helping me see the assumptions I’d stopped noticing I was making.
I have a feeling I’ll be pulling a few more old books off the shelf before this is over.
Books Referenced
- Crossing the Chasm — Geoffrey A. Moore
- The Innovator’s Dilemma — Clayton M. Christensen
- The Long Tail — Chris Anderson
- Blue Ocean Strategy — W. Chan Kim and Renée Mauborgne
- Good to Great — Jim Collins
