The Customer That Isn't Yet

What happens to disruption theory when the customer is being disrupted too

Clayton Christensen built disruption theory on a stable assumption: you know what a customer is. The disk drive buyer, the steel mill operator, the university student. They might be underserved or overserved, they might have access or lack it. But you could point to them and say: there. That’s who we’re building for.

His framework gives you three moves. Sustaining innovation: make the product better for existing customers. Low-end disruption: make it cheaper for customers who are overserved. New-market disruption: make it accessible for non-consumers who couldn’t participate at all.

I’ve been involved in building a product. The Innovator’s Dilemma maps neatly onto the market we’re entering. Incumbents serve established needs. AI tools make the same service cheaper. Our product opens access for non-consumers who currently get nothing. Textbook new-market disruption.

And then you notice the floor moving.

Here’s the problem with applying a 1997 framework in 2026: the non-consumer isn’t static.

The person we’re building for is simultaneously being transformed by the same forces that enabled our product. AI doesn’t just change what you can build — it changes who does the buying. Today’s solo operator might be tomorrow’s agent-augmented business running multiple revenue streams with no employees. Or an autonomous entity that needs services through an API, not a dashboard. Or an organizational form that has no natural person to serve at all.

Christensen’s framework assumes the customer category is the stable ground you build disruption on. The disk drive market changes; the PC buyer stays recognizable. But what happens when the technology disrupts the customer and the product simultaneously?

Three things break.

First, “non-consumer” loses its boundary. Non-consumption implies a stable market you could serve but aren’t. It assumes the person exists, wants the product, but can’t access it. When AI reshapes what operating a business looks like, you’re not serving a known population that lacks access — you’re serving a population that is itself emerging. You can’t survey non-consumers who don’t exist yet.

Second, the disruption timeline collapses. Christensen showed disruption taking years or decades — the entrants improve from the bottom, the incumbents respond too late. But when the customer transforms at the same speed as the product, there’s no comfortable march upmarket. You launch for today’s non-consumer and within eighteen months the category might mean something fundamentally different. The product you built for non-consumers needs to serve a customer that didn’t exist when you started building.

Third, business model assumptions destabilize. Disruption theory says the entrant wins because the incumbent’s business model can’t serve the new market profitably. But if the future customer is an autonomous agent consuming services via API calls, the pricing model isn’t human subscription economics at all — it’s per-inference, per-decision, per-transaction. The business model advantage of today becomes irrelevant against a customer that doesn’t think in months.

None of this means Christensen was wrong. The pattern he identified — incumbents failing because they optimize for existing customers while non-consumers go unserved — is playing out exactly as described in market after market. The incumbents are pivoting to enterprise. The smaller segments are ignored. The entry points are open.

But the framework treats this as a snapshot: identify the non-consumer, build for them, improve over time, eventually displace the incumbent. What it doesn’t account for is a world where the snapshot itself is in motion. Where the customer you identified in January has different capabilities by July, different form by December, and might not be recognizably the same category by the year after.

This isn’t hypothetical. Agent-run businesses, autonomous entities, API-consuming agent-customers — these aren’t science fiction. They’re design considerations that anyone building for small operators is already encountering.

So what do you do? You can’t wait for the customer to stabilize — that’s the incumbent mistake Christensen already identified, just applied to a moving target instead of a static one. And you can’t plan for every possible customer form — that’s speculative architecture, the trap of building for hypothetical requirements.

What you can do is separate the knowledge from the interface. Deep understanding of a domain doesn’t care who consumes it. That’s the stable asset. How you deliver that understanding — dashboard, API, feed, whatever comes next — is the variable. Build the knowledge. Let the interface follow.

Maybe this is the update Christensen’s framework needs for the AI era: disruption happens not just to products and markets, but to the customer categories themselves. The fourth move isn’t sustaining, low-end, or new-market. It’s building for non-consumers who are themselves in the process of becoming something you can’t fully predict. Not building for the future customer, but building so that you can serve whoever the customer becomes.

The Innovator’s Dilemma says to serve the underserved. The update says: the underserved are also transforming. Build the knowledge. Let the interface follow.