AI Moves Fast. People Don't.

I hear people talk about AI a lot lately, mostly in passing rather than in any organised way. At a bus stop, in a shop queue, at the school gate. And it's almost always dismissive. It's a threat, or it's beneath their interest, or it will never really touch their life. Which is fine. None of that changes anything for me directly.

But it does make me think about timing. Not whether AI will change things, that argument is basically settled, but when the change people are already predicting will actually reach the people I overhear at the bus stop.

I created a Facebook account in early 2008, which even at the time felt fairly late. I remember watching it move slowly outward from high school networks to workplaces to extended family, one relative at a time, over years rather than months. It took roughly fifteen years from that account creation to the evening my brother and I looked at each other across a dinner table and sighed, in unison, at our parents for being on their iPads instead of talking to us. Fifteen years for a technology to travel from "the thing university students are doing" to "the thing that interrupts family dinner." That's the real adoption timeline for something that, by most accounts, changed the world.

McKinsey's report on generative AI puts mass adoption somewhere around 2028. I don't doubt the modelling. What I doubt is that anything written by AI commentators, most of whom are already deep inside the technology, reflects how the average middle-aged person on the street actually feels about changing their habits. There is no version of the next two years in which fifty-nine-year-old Susanne has a digital companion. She might at seventy-four, fifteen years from now, on roughly the same curve Facebook took to get from me to my parents. But not on the curve the reports imply.

The same pattern shows up every time, once you start looking for it. Smartphones existed years before most people replaced their old phone with one. Contactless payment was available long before most people trusted it enough to stop carrying cash. Streaming didn't kill the cable subscription the year it launched, it took the better part of a decade, and there are still households paying for both. In every case, the technology arrived on one timeline and the behaviour change arrived on a much slower one, dragged along by habit, trust, and the simple fact that most people don't reorganise their lives around something new until they've watched enough other people do it first and nothing bad happen to them.

There's a lot of language right now about rapid disruption. Cannibalisation, people blindsided, entire categories of jobs gone within a couple of years. I don't think that timeline survives contact with how humans actually behave. AI can, in principle, do a great deal within two years. Whether it will is a different question entirely, and that question depends on human decision-makers who move at a completely different speed than the technology underneath them. Capability and adoption are not the same clock.

What I keep coming back to is that AI requires participation, not just adoption. Anyone can install a tool. Fewer people will actually change how they work, learn, or spend an evening because of it. Adoption is a checkbox: does the technology exist in your life in some form. Participation is a behaviour change: are you actually doing something differently because of it. Most forecasts measure the first and quietly assume the second follows automatically. It rarely does. We are consistently quick to overestimate technological change and just as consistently slow to account for how much people resist changing what they actually do day to day.

This isn't unique to AI. It's the same gap institutions run into whenever they roll out a new digital tool and expect people to just start using it because it's there. A museum can launch an app, a chatbot, an interactive kiosk in the lobby, and the technology can be genuinely good, and still sit mostly unused for years, not because people rejected it outright but because using it requires a behaviour change nobody has actually made yet. The tool existing and the tool being adopted into how people actually behave are two separate timelines, and institutions plan for the first one far more often than the second.

None of this means the change isn't coming. It clearly is. It just means the timeline being sold, the one where entire professions vanish within two years and everyone has a digital companion by the end of the decade, is describing the technology's capability, not people's willingness to participate in it. Susanne isn't wrong to be dismissive right now. She's just early in a curve that will eventually catch up with her, the same way it caught up with my parents and their iPads, on its own schedule and nobody else's.

AI moves fast. People don't.