By /

Why the AI Transformation in Working Life Is Unlike Any Earlier Technological Shift

Part 3/5 of the essay series “Your Competitive Advantage Is Human”

Earlier office technologies arrived one function at a time: the spreadsheet for finance, CRM for sales, ERP for production. The AI transformation arrives everywhere at once, touches everyone at once. For the first time in human history we can increase the output of cognitive tasks without increasing payroll costs — and for the first time a single technology affects every knowledge worker at the same time. In this part I’ll argue why these two sentences mean that a leader should not treat AI like one IT project among others.

In the first two parts of the series I’ve argued that technology doesn’t differentiate, because everything copies, and that AI isn’t a competitive advantage either, because the same tools, the same challenges, and the same benefits are within everyone’s reach. From this someone might draw the conclusion that AI can be taken calmly — “one wave of technology among others, this too shall settle”.

That would be the wrong conclusion. My claim is not that AI is insignificant. My claim is that AI is exceptionally significant, but that its significance is realised not as a technology purchase but as a change in people and the organisation. And for the scale of that difference to open up, you have to understand why this shift differs from every earlier one. I see in it three historical firsts.

The first first: the output of thinking work detaches from payroll costs

Throughout industrial history a simple equation has held: if you want more thinking work — more analyses, reports, plans, customer replies, translations, code — you hire more people. Between cognitive output and personnel costs there has been a practically linear link. Machines multiplied the productivity of muscle work a couple of hundred years ago, and information technology has made moving and storing information more efficient, but the thinking work itself has always demanded a human’s time, hour by hour.

Now that link is breaking for the first time. The output of cognitive tasks can be increased without payroll costs rising in the same proportion. A draft appears in minutes, a summary in seconds, a hundred customer replies for the same effort as one.

The economic significance of this is hard to overstate, and so it’s easy to draw the wrong conclusion from it: that this is primarily about cost savings, about doing the same work with fewer people. I find that thinking both short-sighted and, in the light of history, unlikely. When the price of a factor of production collapses, its consumption usually doesn’t fall but explodes — cheap electricity, for instance, didn’t reduce the use of electricity; it gave rise to entirely new electricity-using industries. The essential question, then, is not “who do we need fewer of”, but “what is now worth doing that wasn’t worth doing before”, and above all: where do we direct the time and thinking that people are freed up to give? That is a question of leadership, not of technology.

The second first: every function, every knowledge worker, all at once

Earlier office technologies were precision weapons. The spreadsheet revolutionised the finance department. CRM changed sales. ERP organised production and logistics. Each of them was a big change, but for only one function at a time, and the rest of the organisation could carry on as before. The roll-out was a project with an owner, a budget, and an end.

AI doesn’t work like this. The AI transformation cuts across every function in working life and affects everyone doing knowledge work — broadly and all at once. The same technology changes the work of the lawyer, the marketer, the bookkeeper, the coder, the customer-service agent, and the CEO in the same year, often in the same month. There is no function you could rule out of the change, and no project manager to whom “adopting AI” could be delegated the way an ERP project once was.

Two things follow from this. First, the burden of change leadership is different in kind: this is not about one department adapting but about the whole organisation learning at the same time. Second — and this often goes unnoticed — AI transformation forces you to rethink processes and operating models. When individual stages of work speed up tenfold, the bottlenecks in the process move: an approval chain that used to be tolerable turns absurd when the work before it is finished in an hour instead of days. An old process with new technology is like a motorway with a horse-and-cart speed limit. That’s why simply handing out tools isn’t enough, and that’s why, as I described in the previous part, 95% of pilots produce no measurable result.

The third first: no one knows where this bends

Earlier technological shifts had one merciful feature: the technology stayed roughly still for as long as the organisation was learning it. An ERP system was, in its essentials, the same at the end of the roll-out project as at the beginning.

With AI, that assumption fails too. The technology is developing so fast that no one knows where it will bend in six months or a year. I say this fully aware that most of the field’s forecasts are hype — and so I’m not asking you to believe any single forecast. I’m asking you to notice something more important: uncertainty itself is a permanent feature of this AI transformation, not a passing phase. And uncertainty has a direct strategic consequence. If you cannot know what the technology will be able to do in a year, you cannot build your strategy on any single technological bet. The only bet that pays off in every scenario is an organisation that learns fast, because it benefits from every step of development, regardless of which direction the development takes.

What the research says about how the benefit is distributed

The three firsts explain the scale of the AI transformation. But the research tells one more thing that ties this part to the series’ central claim: the benefit is distributed not according to the technology but according to the people.

Researchers at Harvard and Boston Consulting Group (Dell’Acqua et al. 2023) gave AI to hundreds of BCG consultants to use in real work tasks. The results were two-sided in the literal sense. On tasks that fell within AI’s zone of capability, the consultants performed clearly faster and produced higher-quality work. But on tasks that looked the same yet fell outside that zone, those who used AI performed worse than the control group, because they trusted output that sounded convincing but was wrong. The researchers called the phenomenon the jagged technological frontier: AI’s capability boundary does not run neatly between task types but jaggedly within them, and you can’t see the boundary from the outside. You learn it only by using the technology — and that is exactly why the decisive skill is not using the tool but judgement: when do I trust, when do I check, when do I do it myself?

The second result is even more interesting from the point of view of leadership. Researchers at Stanford and MIT (Brynjolfsson, Li & Raymond) followed a little over five thousand customer-service agents who were given an AI assistant. Productivity rose by about 15% on average, but the average hid the essential. The largest benefit went to the least experienced workers, whose performance improved by around a third, while the most seasoned veterans benefited only a little. The researchers’ interpretation is worth noting: AI in effect distributed the veterans’ tacit knowledge to the novices and narrowed the gaps in skill.

Stop and sit with this, because it changes the logic of standing out. If AI lifts everyone’s baseline performance close to good, then “good enough” — which I wrote about in the first part of the series — becomes cheaper and more common still. Then competition is decided even more firmly where AI cannot reach: in judgement, in the encounter with the customer, in trust, and in leadership. Technology levels the playing field at the task level and shifts the standing-out to the human level.

“Electricity and the internet were just as big as the AI transformation”

This part’s objection is historical: every technology has, in its time, been considered a unique upheaval. Electricity changed everything. The internet changed everything. Why would AI be different?

My answer is that electricity and the internet really did change everything — but they were aimed at a different thing. Electricity multiplied physical work and made new modes of production possible. The internet collapsed the cost of moving and distributing information. Neither, however, did the thinking work itself: analysis, writing, planning, and reasoning remained entirely with humans, and so their output stayed tied to payroll costs. AI is the first general-purpose technology aimed at cognitive work itself. What’s more, earlier shifts advanced as a transition spanning decades — electrification took factories a generation, because the benefit was realised only once the whole mode of production was redesigned on electricity’s terms. The same lesson applies now, but the time span is measured in years, not decades. In the age of electricity there was time to look to your neighbour for a model. Now there isn’t.

What this means in practice — do this tomorrow

  1. Swap the annual plan for 90-day learning cycles. Don’t try to write a three-year AI strategy — it goes out of date before the ink dries. Define a direction and principles instead, and proceed in 90-day cycles: each cycle a clear experiment, an owner, a human metric and a euro metric, and at the end of the cycle an honest review — what did we learn, what do we scale, what do we drop. Define and put into words why the AI transformation concerns your particular organisation.
  2. Draw your organisation’s jagged frontier. Gather from your teams a living map of where AI is already a reliable help, where it is useless, and — most important — where it is dangerously convincing but wrong. Update the map every cycle, because the frontier moves with each model version. This map is genuinely your own: it can’t be bought or copied, because it arises only from your people’s experience in your work. Notice what I just said — it passes the VRIN test.
  3. Train judgement, not buttons. Tool training goes out of date in months; judgement never does. Build the training around three questions: When can I trust AI’s output as it is? When must I check it? When am I better off doing the work myself? And make sure your most experienced specialists are involved in the teaching — in the light of the research, it is precisely their tacit knowledge that AI spreads to others, so they had better be spreading it deliberately rather than by accident.

In closing

The three firsts — thinking work detaches from payroll costs, the change touches everyone at once, and uncertainty is a permanent condition — together mean one thing: the AI transformation is too big to delegate to the IT department and too fast to settle with purchases. It is settled where people learn, and people learn where they are led well.

And that is exactly why I argue that the single most important success factor in the AI shift is not technological but a matter of leadership doctrine. In the next part I’ll tell you why I believe servant leadership is the leadership doctrine of the AI era — and what getting at root causes, hearing the quiet signals, and meeting employees mean in practice, tomorrow morning.

I speak about these themes in keynotes, leadership-team coaching, and panel discussions. If you’d like to challenge your organisation’s thinking about the scale of the AI shift, get in touch → lenni@lennilaukkanen.fi

Previous part: AI Is Not Your Competitive Advantage

Next part: Leading the Competitive Advantage — Servant Leadership in Practice

Sources

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top