At easyJet I ran the web shop. Every ticket the airline sold went through it, and I introduced the features that now look ordinary: speedy boarding, online check-in, bundled products, and the rest. Travel agents all but disappeared along the way. I was on the winning side of that shift, and it is the lens I bring to AI in education.
Benedict Evans argued that online booking was very disruptive to travel agents but didn’t change the fundamentals of the airline business. In his newsletter of 18 October he sorts the internet’s effect on industries into three: those it changed completely, like recorded music and retail; those where it was a useful tool, like mining and construction; and a middle group, airlines and banks, where it was fundamentally incremental. He suggests asking which of the three AI will be for each industry.
I agree about the aircraft. I disagree about the shop. A travel agent sold a seat. The web shop sold the seat and then everything around it, because we owned the conversation with the customer: speedy boarding, bundles, products no counter could have sold. That was more than replacing a channel. It changed what the airline had to offer. Incremental for the planes, I think, very much so.
Two postures
I then went to MORE TH>N, the insurer, who needed someone like me because they were being forced to follow, in a very different mood. It was a classic late-majority move: we must have a presence, because the previous model has been pulled from under us. Reluctant, and unavoidable.
Same shift, two postures. At one company the new channel was a chance to build what only it made possible. At the other it was a defence: be there, because the old way is going. The two lead to very different outcomes.
Which one does AI put you in?
Learning leaders tend to talk about AI as one thing that either transforms the business or merely helps it. Evans’s test says that is the wrong question, because the same technology lands in a different bucket depending on which link of the value chain you stand on. Chegg and Duolingo show it. In its second-quarter results Chegg reported revenue down 51% year on year, with Dan Rosensweig quoted saying AI created real headwinds for this company. In its second-quarter letter Duolingo reported 58.7 million daily active users, up 23%. Same technology, opposite fortunes. Neither filing says AI is the whole cause, and I would not claim it is.
Run the sort on each link of your chain. My own hypotheses, which you should challenge: generic content has changed completely, and Chegg is the evidence; practice has too, because a patient partner at any hour is now cheap; assessment is a tool; accreditation is incremental until a body that matters changes its rules. I argued the keep-or-change half of this in AI Is Making Education Faster. It Could Make It Different.
Two randomised trials point at the tutoring link. Kestin and colleagues found that students in one Harvard physics course learned significantly more in less time with an AI tutor (I covered the headline result in The Study Nobody in EdTech Is Quoting). A World Bank team found about 0.3 standard deviations in a six-week after-school English programme in Nigeria. Both had a designed structure or a teacher behind the AI. They tell you tutoring is at least a tool. They do not tell you it replaces the person who runs the room.
easyJet’s example does not always apply
A travel agent was a channel for a seat. A teacher is not a channel for learning: a teacher is part of what is being learned. In my experience, learners look harder for a person and a place as answers get cheaper, and they want to avoid the lonely experience of learning alone from a digital interface. The human link is gaining value, not losing it.
Travel agents never got to ask in time where their customers would start looking. For learning, that may soon be an AI assistant instead of a search box or a referral. I have no data on how fast, so this is something to measure: the share of new learners who arrive that way.
Three questions for the afternoon:
- Who sits between you and the learner? A teacher, a school, a centre, an app store, a search engine.
- Is that intermediary a channel or part of the product? A channel can be replaced. Part of the product cannot, without changing the product.
- Which posture are you in? Building what the new channel makes possible, or defending a presence because the old way is going.
For each answer, write down the evidence that would change your mind. A row where you cannot say what would change your mind is a belief, not a finding.
A travel agent sold a seat. Before you decide AI is your web shop moment, find out what your teachers sell.
Leave a Reply