Today, we’re 130 people. We’ve worked across industries, survived multiple hype cycles, and watched the AI landscape reshuffle itself faster than most traditional businesses can update their strategy decks. And one of the things I get asked most often (usually by someone ambitious and younger than me) is “What should I read if I want to build something in AI?”

My honest answer: not what I read, but not entirely different either. The fundamentals still matter, and some classics are classics for a reason. But the specific books that shaped my thinking may be liabilities now as they reflect a world where “AI-native” meant something completely different from what we’re seeing today.

If you’re aiming to build an AI company in 2026, you need a reading list that reflects where we are headed in the future.

The conversations I’m having with clients today are almost unrecognisable from 2020.

Back then, a big chunk of any client engagement was educational. You were explaining what machine learning could and couldn’t do. You were managing fear and scepticism in equal measure.

Now I walk into meetings where the client’s internal team has already built something with an LLM. They have opinions about RAG architectures, they’ve had a bad experience with a vendor and they want to know how to fix it. The conversation starts further along, which sounds like a good thing, and often is, but it also means your positioning has to be sharper, your differentiation more concrete, and your delivery track record more credible.

The opportunities for young builders have also multiplied in ways that didn’t exist before. You can ship something meaningful with a small team in a way that would have taken an order of magnitude more resources five years ago. If you’re early in your career and thinking about building something in AI, you have a structural advantage: you can move fast.

Co-Intelligence, Ethan Mollick

Mollick is probably the most balanced voice I’ve encountered on human-AI collaboration, and that balance is exactly what makes this book useful for a builder. He’s not a booster and he’s not a doomer. Ethan is trying to honestly describe what these tools can and can’t do for knowledge work, and he backs it up with actual experimentation.

Why does this matter for building a company? Because your clients are going to come in with wildly miscalculated expectations and part of your job is to reset those expectations without losing the room. This is a good economics meets AI book to start.

The Innovator’s Dilemma, Clayton M. Christensen

Well, this one is from 1997, but i’m putting it on this list, because I think it applies to AI consulting specifically in a way that most people haven’t thought through.

The dilemma Christensen describes, where established players can’t respond to disruptive entrants because the new thing initially serves worse customers with an inferior product at a lower margin, maps exactly onto what’s happening in professional services right now. AI-native boutiques are taking work that large consultancies can’t profitably do because their cost structures and client expectations don’t allow for it. It’s the core strategic opportunity for anyone building in this space today.

Obviously Awesome, April Dunford

A positioning book that clarifies your offer and sharpens how you sell it.

You’ve built something real (and it’s easy these days), you know it works but you have absolutely no idea how to explain what you do to someone who isn’t already familiar with your work.

Dunford’s framework (figure out what you uniquely are, for whom, against what alternative) is simple. I’d make this required reading before anyone at a young AI company talks to a potential client or investor.

The Cold Start Problem, Andrew Chen

If you’re building any kind of AI product that depends on data, users, or network effects to get better over time, this is the book that explains why the first phase is so brutally hard and what to actually do about it.

Chen’s core insight is that products with network effects don’t fail because the technology is bad; they fail because nobody solved the bootstrapping problem. Getting from zero to the point where the product is good enough to retain users is a different discipline entirely from scaling what already works.

The Innovation Stack: Building an Unbeatable Business One Crazy Idea at a Time,

McKelvey co-founded Square and this book is his attempt to explain something that most business strategy frameworks can’t quite capture: what happens when you solve one hard problem and discover it forces you to solve five others, and then those solutions become the actual moat.

He calls this an “innovation stack”, a cluster of interdependent inventions that competitors can’t copy because they’d have to copy all of them simultaneously. It sounds abstract until you read the case studies, and then it becomes one of those ideas you can’t unsee. The defensible AI businesses won’t be the ones with the best model, they’ll be the ones where the product, the data flywheel, the customer relationship, and the operational approach are so entangled that unpicking any one piece breaks the whole thing.

The cost of building has collapsed. The appetite for AI-native solutions has accelerated, so go build!