It is an uncomfortable thing, to counsel patience in a room giddy with possibility. This is not an argument against AI. Some of us cut our teeth in those heady internet years when everything seemed possible. But we also remember the bust.
There is something unsettling about watching whole industries rush headlong into a transformation without pausing to ask whether anyone actually wants what is being built. The better question was never how can we. It was should we.
We have been here before
The pattern is not new, only the vocabulary. In the dot-com years, everyone was suddenly an internet strategist, speaking with certainty about things no one had heard of a year earlier. The confidence ran far ahead of the understanding.
It is happening again. Everyone is an AI expert. Everyone has an opinion and a strategy. Dig beneath the surface and you find the same mix of excitement, heavy spending and quiet confusion that ran through the nineties.
That era built some of the largest organisations alive today. It also burned through fortunes on ideas that never resonated. The failures were not only financial. They were confusing experiences for the people who met them, systems that made simple tasks hard, digital services that left customers longing for the days of picking up the phone and reaching a person.
Does this sound familiar?
The amnesia
What is striking about both waves is how quickly the last lesson is forgotten. In the dot-com years, organisations taught a generation of customers and staff to dread the technology by giving them good reason to. People spent more time fighting new systems than doing their work. And then everyone seemed surprised when they pushed back, resisted, or quietly opted out.
We are at risk of doing it again. Every poorly built solution adds a little more frustration and a little more distrust. We are teaching people to distrust automation by handing them automation worth distrusting.
The sameness trap
Here is the part the rush hides. When the answer comes out of a box, everyone gets the same answer.
An off-the-shelf AI solution gives you what it gives everyone else. The same capability, the same limits, the same experience your competitor just bought. Whatever edge you imagined you were purchasing dissolves the moment the tool is deployed as intended, identically, everywhere. Sameness is not an advantage. It is the short road to commodification, where the only thing left to compete on is price.
Complex organisations do not have off-the-shelf problems. They have particular people, particular histories, particular constraints a general tool cannot see. Meeting those needs is not a matter of better prompts. It is a matter of judgement, and judgement is the one thing the box does not come with.
Ask the ones who rushed
The organisations that moved fastest are already quietly walking parts of it back.
Ford is the clearest case. It spent three years and a great deal of money on AI tools meant to lift production quality, confident the technology would do the work. It did not. The company ended up rehiring around three hundred veteran engineers, the ones Ford itself calls its “gray beard” engineers, to fix what the automated systems could not.
Its own engineering leadership named the mistake without flinching.
“Mistakenly, we thought that by just introducing artificial intelligence and ingesting the design requirements that we had, that would produce a high quality product.”
Charles Poon, Ford VP of vehicle hardware engineering – source https://www.foxbusiness.com/technology/ford-rehires-experienced-engineers-after-ai-misses-mark
There’s a detail in the Ford story that should give any leader pause. Many of their most experienced technicians had already left the company before their knowledge could be captured, so the AI was trained on everything except the expertise that mattered most. The machine learned the process. It never learned what the veterans knew.
“We recognized that for us to enhance some of our automation and machine learning and artificial intelligence tools, we needed to ensure that they were trained by the most experienced individuals.”
Charles Poon, Ford VP of vehicle hardware engineering – source https://www.foxbusiness.com/technology/ford-rehires-experienced-engineers-after-ai-misses-mark
The rehired engineers helped Ford top an industry quality ranking it had not led in sixteen years. The expertise it had tried to automate away turned out to be the thing that fixed the problem.
Klarna said the quiet part out loud
Klarna went further and faster. The fintech replaced some seven hundred customer service agents with an AI assistant, froze hiring, and spent the better part of two years held up as the poster child for what AI could do to a headcount. Then quality fell, and the story changed.
Its chief executive, Sebastian Siemiatkowski, was unusually candid about why.
“As cost unfortunately seems to have been a too predominant evaluation factor when organizing this, what you end up having is lower quality.”
Sebastian Siemiatkowski, CEO, Klarna – source https://www.aol.com/klarna-ai-replaced-700-workers-210647896.html
The reversal was not a retreat from AI. Klarna still leans on it heavily for routine work. What changed was the recognition that stripping the humans out entirely, to chase a cost saving, cost the company something it valued more.
Neither Ford nor Klarna abandoned the technology. Both learned the same lesson the hard way. AI is only as good as the human expertise behind it, and pulling that expertise out to save money is a false economy that bills you later, with interest.
This is not two unlucky companies. It is the pattern.
IBM’s study of two thousand chief executives1 found nearly two thirds, sixty-four percent, admit that the fear of falling behind drives them to invest in technologies before they understand the value those technologies will bring. The rush is not a strategy. It is an anxiety.
Design the right thing, then design it right
Human-centred design offers a way through, built on two embarrassingly simple principles the rush routinely ignores.
Design the right thing. Start by understanding real needs, not the needs you assume from a distance, not the ones that would make the technology look impressive, but the actual, messy needs people have in their real contexts. Skip this and you spend heavily on things nobody wanted.
Then design the thing right. Build small, prototype with real people, learn from what fails, improve as you go. This is not only about avoiding expensive mistakes. It is how you find the solutions people actually want, the ones you could not have imagined from the boardroom.
What survives the bust
Many of us have seen this film, and we know how it ends. The excitement will meet the hard facts of human behaviour and economic gravity, and many of today’s AI initiatives will fail. But some will not, just as the internet eventually produced genuinely transformative companies. The work is telling the signal from the noise. The organisations that thrive will not be the ones deploying AI fastest. They will be the ones deploying it most thoughtfully.
This is the work we do. We help leaders ask should we before how, understand what people genuinely need before committing to a solution, and build the particular rather than buy the generic. Not to slow you down, but to make sure that what you build is worth having built.
The future is not made better by meeting it faster. It is made better by meeting it wisely. Just because AI can build something does not mean it should. But understand what people actually need, rather than what impresses other technologists, and you might build something extraordinary.