Few words have been stretched as thin as "disruptive." In the last decade, it has been applied to everything from ride-hailing apps to cold brew coffee, flattened into a synonym for "new" or "exciting" or simply "tech-related." So when people ask whether AI is a disruptive innovation, they are often asking the wrong question — or at least asking it with the wrong vocabulary.
To answer it properly, you need to go back to the man who defined the term. And then you need to decide whether the definition still holds.
What Disruption Actually Means
In the mid-1990s, Clayton Christensen — a Harvard Business School professor who would go on to be called one of the most influential business thinkers of the 20th century — was studying the disk-drive industry. He noticed something that defied conventional logic: well-managed companies, doing everything right by the standards of their industry, kept getting killed by inferior products from smaller competitors.
The pattern, which he named disruptive innovation, worked like this. A new entrant would arrive at the low end of a market — the overlooked, underserved, or unprofitable segment that market leaders had little interest in defending. The new product was cheaper, simpler, often worse by every established metric. But it was good enough for a certain customer. And as it improved, it crept upmarket — until the incumbents had been made irrelevant on entirely new terms.
Critically, Christensen distinguished this from sustaining innovation — improvements that make existing products better for existing customers. Sustaining innovations are important, but they don't reshape industries. They reinforce them. The question is which category AI falls into — and the honest answer is: both, and also neither.
Framework · Three types of innovation — where does AI fit?
The Case For: Yes, AI Is Disruptive
Apply Christensen's lens to specific industries and the pattern fits in striking ways.
Consider legal services. For decades, the industry has been structured around billable hours, specialist expertise, and high barriers to entry — serving large corporations and wealthy individuals extremely well, while leaving small businesses and ordinary people largely priced out. AI entered from the bottom: contract-review tools, compliance checkers, plain-language document drafters — none of which initially threatened senior partners at white-shoe firms. They served the underserved.
The same pattern is playing out in healthcare. AI diagnostic systems are now achieving 94% accuracy in detecting certain cancers from imaging, matching or exceeding specialist radiologists in specific contexts — technology that first gained traction in under-resourced clinics that couldn't afford specialist coverage, not in elite hospitals that already had it.
By the time a Fortune 500 company approves a new AI tool, a startup has already shipped five iterations with real users.
The Christensian incumbent-rationality trap is operating exactly as predicted. Large companies adopt AI cautiously — procurement cycles, legal review, compliance, data governance, risk aversion baked into every layer. By the time a Fortune 500 company has approved a new AI tool for internal use, a well-funded startup has already shipped five iterations and has real users telling them what actually matters.
The Case Against: AI Is Something Else Entirely
But here is where the framework strains.
When McKinsey uses AI to make its consultants more productive, that is not disruption — it is efficiency. When a law firm uses AI for document review, the firm still exists, the billable hour still exists, the client relationship still exists. The product is better; the market structure is unchanged. Much more important than the AI technology in determining whether something is disruptive is the business model in which the AI is used — and its competitive impact on existing products and services.
There is also a more fundamental objection. Christensen's framework was built for a world where disruption unfolded over years. New entrants needed time to gather resources, iterate, and build distribution. AI compresses that timeline to months — sometimes weeks. When a technology moves that fast, it does not march from the bottom of a market upward in the tidy sequence the theory describes. It arrives everywhere at once. That is not classical disruption. It is something closer to a general-purpose technology shock.
Data · The speed of the shift
What History Tells Us
History is instructive here — though not always comfortably so.
| Technology | Arrived | Full impact | What people feared | What actually happened |
|---|---|---|---|---|
| Printing press | 1440s | ~200 years | Collapse of the scribal class; spread of heresy | Mass literacy, Reformation, scientific revolution — and propaganda wars |
| Electricity | 1880s | ~40 years | Fires, electrocution, moral corruption | Total restructuring of manufacturing — after factories were redesigned from scratch |
| Internet | 1990s | ~15 years | End of privacy; collapse of traditional media | New industries created (search, social, e-commerce), not just old ones destroyed |
| AI / LLMs | 2022– | Unfolding now | Mass unemployment; existential risk | TBD — but 72% enterprise adoption in 3 years is unprecedented speed |
The pattern across all of these: transformative technologies overshoot short-term expectations and undershoot long-term ones. AI is often compared to the printing press as a transformative moment — but there were also about 200 years of dark history associated with the printing press.
Electricity transformed manufacturing in the late 19th century, but the full productivity benefits took nearly 40 years to materialise, because factories had to be completely redesigned around the new technology rather than swapping steam engines for electric motors one-for-one. The disruption was real. The timeline was not what anyone predicted.
So What Is AI, Exactly?
The most precise answer is that AI is a general-purpose technology — a foundational capability, like electricity or the internet, that has no single application but instead restructures the cost and capability curves of almost every application simultaneously.
General-purpose technologies don't disrupt industries in the Christensian sense. They make entirely new industries possible while rendering the cost structures of existing ones obsolete. The AI disruption market was valued at $206.6 billion in 2025 and is projected to reach $1.5 trillion by 2030 — a 40% compound annual growth rate. These are not the numbers of a technology nibbling at the edges of specific market segments. They are the numbers of a technology reshaping economic infrastructure.
The Answer — And Why It Matters
That distinction matters because it changes how you respond. Disruption, in the classical sense, is something companies can navigate with the right strategy — identify the low-end threat early, invest in the new model, cannibalise yourself before someone else does. General-purpose technology shocks require something different: rebuilding your operating assumptions from the ground up.
If entrepreneurs and policymakers harness AI's power well, it could unleash unparalleled prosperity. If not, it will not only exacerbate inequality but also disrupt work and society in ways that benefit few.
The companies and individuals who will navigate this era best are not waiting to see which market gets disrupted next. They understand that the terrain itself is changing — that the rules of competition, the cost of expertise, and the barriers to entry are all being rewritten simultaneously.
History did not give the printing-press era to scribes who got faster. It gave it to people who understood that the question had changed.