In April 2026, at the Chief Architect Network event, I sat on a live Architect Tomorrow panel, alongside Whynde Kuehn and hosted by Oliver Cronk. The conversation ranged across The Five Human Lenses of Transformation™, business architecture and sustainability. But one exchange near the end has stayed with me more than the rest.
Charles Betz, a VP and Analyst at Forrester, raised a paper I hadn’t come across before. Lisanne Bainbridge. Ironies of Automation. A UK human factors thought leader who had flagged, decades before any of us were talking about AI copilots, a fundamental issue with the way we tend to design automated systems.
Her argument goes something like this. We automate the routine work, and we leave the humans in the loop to handle the critical issues. That sounds sensible. Right up until you realise that the very act of removing the human from the routine slowly erodes their focus, their context, and their readiness. So when the automation breaks and at some point, it always does the human who is meant to step in no longer has the muscle memory, the context, or the confidence to act.
I’ve been sitting with that idea since the panel, and I keep coming back to something I said in the moment that now feels more important than I realised.
Around $1.5 trillion is currently being spent on AI. And if you ask honestly how much of that investment is going into the technology and how much is going into the humans on the other side, the people who are supposed to adopt it, adapt with it, and eventually be the safety net when it wobbles, my best guess is 98% on the technology and 2% on the human side.
If that split is even roughly right, we have a problem.
That problem has a name. It’s the human gap: the distance between the plan and what people actually believe, experience and do.
Because what we are building, at pace, is a generation of organisations where the technology is sprinting ahead and the humans are being asked to sprint behind it. Same meetings. Same workloads. Same decision rights. Just with more outputs to review, more outputs to trust, more outputs to be accountable for, when nobody has quite had time to build the capability or the confidence to do any of that well.
Which leads me to something I want to say quite directly to leaders of AI transformations.
Enterprise architects have a role to play in slowing this down.
I know. That is not the most fashionable thing to say in a room full of technologists. We are usually the ones being accused of being too slow, too cautious, too attached to governance. And there is some truth in that. At times we’ve slowed down the wrong things; we’ve used a hammer to crack a nut. But there is a version of our role right now that is badly needed. Not to block AI. Not to fight it. But to give the human side a fighting chance of catching up with the technical side. If EAs fuel the tech scenarios without amplifying the need for the people side of the house to step up, we are in trouble.
Oliver picked up the same thread from a different angle. What he called augmentation fatigue. The quiet risk that roles drift from being engaging and creative to being something much closer to staring at a production line looking for anomalies. Imagine a software engineer whose day used to involve the dopamine hit of solving a tricky problem. Now imagine that same engineer whose job is largely to debug code an AI has produced. Same job title. Very different experience.
Multiply that across a function, across a business, across an industry, and something subtle but serious begins to happen. The parts of work that made people feel capable, creative and alive quietly disappear. The parts that remain are the ones that require vigilance without much reward.
That is a design decision we are making, whether we name it or not.
And I think this is where the most important reframe has to happen. Because most AI transformation conversations I hear at the moment are happening at what I’d call the micro level. Which tool to deploy. Which use case to prioritise. Which governance layer to sit on top. Which metrics to report.
They are all reasonable questions. But they are not the question.
The question is macro. Does where we are going actually make sense? Not for this quarter. Not for this platform. For the people, the teams and the ecosystems that will be living inside the systems we are designing right now. If we only keep answering the micro questions, we risk building something technically impressive but socially unliveable (I think Whynde coined that phrase in another session) that nobody recognises themselves in. A transformation on paper. A workaround in practice.
And here is the part I’ve become less willing to ignore. We already know how to do the macro work. It is not a new capability. It is the part of the architect’s craft that involves stepping away from the diagram, walking across the office floor, spending a day in the life of the people who will actually use what we are building.
When I was at WPP, I had a researcher on my team whose entire discipline was human behaviour. Not a business analyst. Not a technologist. Someone who watched how people opened their laptops in the morning, what they did first, what they avoided, how they felt by the end of the day. You cannot capture that in a requirements document. You have to get out there and see it. And even if you only do it once a year, I promise you, you will think about your technology decisions differently afterwards.
Who on your AI programme is doing that work right now? Observing the impact of the tools being rolled out, and the signals of the consequences, some intended, some unintended.
Not the sponsor. Not the business owner. Not the change lead running comms. Someone whose actual job it is to notice how this is landing in the lived experience of people’s Monday mornings.
If no one in your AI programme has that brief, then you are almost certainly in the 98%.
There is a version of AI transformation that ends with the human dimension treated as an afterthought. A change workstream bolted on at the end. A training plan. A comms pack. There is another version that starts from the human first and uses that as the compass for where and how the technology goes next.
Both are possible. The choice is not forced on us by the technology. It is being made, every day, in the decisions about where the money goes and who gets to be in the room.
So to anyone leading an AI transformation right now, this is probably the quietest but most important question I can offer.
Are we spending 98% on the technology and 2% on the people, and calling that progress?
Or are we willing to slow the technology down just enough for the humans to come with us?
Because one of those futures sticks. The other one is Bainbridge’s warning, turning up again in a new decade.
Whatever next?
