AI is moving faster than anyone predicted — understand “disruption”
What can you expect from this look at disruption
The word was almost worn out — a TED-talk cliché by 2015, drained of meaning by overuse. Then AI started doing the thing the word originally described, and at a pace nobody saw coming. The term was coined by Clayton Christensen in 1995. He probably didn’t see this coming either.
For most of the last decade, disruption has been one of those words you wince at slightly when it appears in a slide deck. It signalled a particular kind of unseriousness — the management consultant promising to “disrupt” the office stationery category, the start-up announcing it would “disrupt” haircuts, the conference panel where everyone agreed disruption was happening and nobody could quite say what they meant by it. The word had become a fashion accessory. You could wear it but it no longer did any work.
This was unfortunate, because the original concept was both specific and useful. Clayton Christensen, a Harvard Business School professor, introduced it in 1995 to describe a particular pattern: new entrants quietly serve markets that existing leaders find unprofitable or beneath their dignity, and then — once the new entrants have built capability at the bottom — they move upward and displace the incumbents from above. The pattern was real. It explained why mainframe companies missed personal computers, why department stores missed discount retail, why Kodak missed digital photography despite inventing it.
What got lost in the subsequent twenty years was the specificity. Disruption came to mean any change anyone wanted to claim credit for. New product line? Disruptive. Slight pricing variation? Disruptive. The original argument — about the structural blind spot of dominant players, about how incumbents lose because their own profitable customers won’t let them notice the new thing — was buried under marketing fiction. By the late 2010s, serious thinkers had largely abandoned the word.
Then AI arrived, and the word started to bite again.
What is happening with AI is not a normal technological change. It is not the kind of efficiency gain that lets a company do the same work with fewer people, the way payroll software did. It is not even the kind of new capability that creates an adjacent market, the way smartphones did. What AI is doing — to legal research, to software development, to design, to translation, to customer service, to a long list of cognitive work — is the textbook Christensen pattern at industrial scale.
The new entrants are serving the bottom of every market simultaneously. Cheap AI is generating documents that are good enough for routine legal work, software good enough for many start-ups, translations good enough for a great deal of business communication. The incumbents — large law firms, software consultancies, translation agencies — cannot easily move down to compete, because their cost structures and their own clients won’t let them. Their most profitable work, for now, is safe. Their least profitable work is being eaten. By the time the AI capability moves up the value chain — and it is moving up faster than anyone predicted in 2023 — the incumbents will find themselves in exactly the position Christensen described: structurally unable to respond, because responding would mean cannibalising what is still working.
This is what makes the word matter again. It is not a synonym for change. It is not what every product launch produces. It describes a specific pattern of incumbent failure that is now happening simultaneously across many industries, at speed, with AI as the new entrant.
The pattern also clarifies what is not disruption, which is useful too. The AI that helps a senior lawyer write better briefs faster is not disrupting law. It is augmenting it, and the senior lawyer will probably do even better. The AI that lets a junior developer ship a small feature without senior review — that is closer to disruption, because it threatens the value structure that supported the senior. The AI that lets a non-developer ship the same feature without any developer at all — that is full disruption, and it is happening in some categories already.
The test, as Christensen would put it, is not whether AI is better than human work. It is whether AI is good enough for some part of the market, at a price the incumbents cannot match, and whether that part of the market is large enough to support the development of more capability that will then move upward. By that test, several industries are well into the early stages of a disruption they have not yet learned to recognise — because they are still thinking about AI as a productivity tool rather than as the new entrant Christensen warned them about thirty years ago.
The word is back. It still means what it meant. We just have to use it precisely again.
Christensen died in 2020, four years before the AI moment he might have found most useful for illustrating his theory. He probably didn’t see this coming either. But the framework he left behind is doing more work in 2026 than it has done at any point since he published it — not because the world has caught up to his ideas, but because the world has finally produced an example clear enough that the original concept is impossible to misread. The fashion accessory has become a tool again.
If you liked this piece, also read Will AI take your job?
Will AI take your job?
Lifestyle publictions from Leisure Media Group
