AI Is Not Your Competitive Advantage
Part 2/5 of the essay series “Your Competitive Advantage Is Human”
When I founded Astu Labs, I expected every organisation that came to us to accelerate its AI transformation to be its own kind of puzzle. It turned out otherwise. Regardless of organisation, sector, or size, everyone has the same challenges with AI and roughly the same benefits on offer. That observation is uncomfortable, but it is also liberating — and it explains why, according to the research, the great majority of companies’ AI initiatives produce no measurable result, even though the technology works.
In the first part of the series I described how, during my IT years, I came to see that hardware, software, and processes are commodities: everyone has the same tools and the same rules, so technology is not where you stand out. I also wrote that AI takes this old truth to a new scale. Now I’ll make good on that claim.
My work is helping organisations adopt AI responsibly, in a controlled way, and toward clear goals — so that the effects are also measured. We don’t deliver technology. We are a change partner for leadership teams. That is exactly why we see something the technology vendor does not, because its gaze is on the product: we see what happens in an organisation after the tools have been bought.
And what we see is astonishingly similar everywhere.
Every organisation thinks it’s unique
Almost every client begins the same way: “Our sector is a bit special.” Or: “Our organisation is the exception on this one.” And I understand the feeling — every organisation is unique in its history, its relationships, and its culture. But from the point of view of adopting AI, the claim doesn’t hold.
It really makes no difference which organisation I walk into. Everywhere the challenges are the same: uncertainty about where to start; confusion about which tools may be used, and for what; worry about data protection and about skills; the gap between enthusiastic early adopters and sceptical wait-and-see types; pressure from the top to “do something with AI” without a clear goal. And everywhere the available benefits are roughly the same: more efficient text work, information retrieval, summaries, customer communication, and routine processes — that is, the same promise every competitor gets too.
This is important to take in: if your challenges are the same as everyone’s and the available benefits are the same as everyone’s, then acquiring the technology cannot be your competitive advantage. It’s merely a chip that gets you a seat at the table. The game is decided by how your organisation’s people learn to play — and how fast.
So your organisation is not unique in its challenges. But your people can be unique in their solutions.
What the research says: 95% fail, and the reason isn’t the technology
My observation is not just a consultant’s anecdote; the same picture emerges from the largest study done on the subject so far. The MIT NANDA initiative’s report The GenAI Divide: State of AI in Business 2025 reviewed hundreds of corporate AI roll-outs, interviewed executives, and gathered survey responses from employees. The result was blunt: about 95% of companies’ generative-AI pilots produced no measurable impact on results. Only a small minority managed to scale a pilot all the way to a business effect.
What’s most interesting, though, is not the failure rate but the reason for the failure. It was not the quality of the models, not regulation, not data security — even though those were exactly what the executives in the study blamed. The researchers named the real reason the learning gap: organisations could not integrate AI into their workflows, their structures, and their culture. In other words, the technology worked, but the organisation did not learn to use it. The failure happened on the human side — where the investments were not directed.
The same report revealed another distortion: the lion’s share of AI budgets flowed into visible pilots in sales and marketing, even though the largest measurable return was found at the more mundane end — automating back-office processes, cutting outsourcing costs, and smoothing operations. The money went where it looked good on management’s slides, not where it would have paid off. This too is a human phenomenon, not a technological one.
And if you still suspect this is about the immaturity of the technology, the report’s third finding pulls the ground from under that as well: the researchers described a “shadow AI economy” in which only about 40% of companies had official language-model subscriptions — but in over 90% of companies, employees reported using AI tools regularly on their own. Read that again. People have already adopted AI past the official initiatives, on their own logins, because it makes their work easier. So adoption does not fail because people resist the technology. It fails because the organisation does not lead what is already under way.
Same technology, different results — where does the difference come from?
This is a good place to stop and sit with a logical conclusion. If everyone has access to the same models and tools, but only a fraction gets a measurable benefit from them, then by definition the differentiating factor cannot be the technology. It has to be something that varies between organisations even though the technology is constant.
And that something is people: their skills, their willingness to experiment, their permission to fail, the quality of their leadership, and the organisation’s ability to change how it works. In the first part I introduced strategy research’s VRIN test — sustained competitive advantage requires a resource that is valuable, rare, and hard to imitate. A language-model subscription doesn’t pass the test: it is neither rare nor hard to imitate. But an organisation whose people learn new ways of working faster than competitors passes it clearly. The ability to learn cannot be bought as a licence, and a competitor cannot copy it, because it lives in people and in the trust between them.
This is also why I called the observation of sameness liberating in the opening. If the challenges are the same for everyone, you don’t need to invent a unique AI strategy. You only need to win at one thing: how well and how fast your people learn. And over that you have more influence than over any technological variable.
“But doesn’t the first mover get a head start?”
The most common objection goes: even if the technology is the same for everyone, the first mover gets a head start. This is partly true — and that’s exactly why it’s a dangerous idea.
Yes, the early mover gets to learn before the others. But notice what the head start rests on: not on the technology acquired, but on the learning achieved. If your organisation acquires the tools first but does not change how it works, no head start is created at all — you’ve merely paid for licences longer. MIT’s data is merciless on this: the number of pilots did not predict success, and large companies led in the number of pilots but lagged in successful roll-outs. The head start is not created at the moment of purchase but by the speed of learning, and the speed of learning is a property of people and leadership.
A second point: a head start based on technology erodes faster than ever. The next model version levels the playing field in months. A head start built into people — a culture of experimentation, learned ways of working, accumulated judgement — instead compounds, because it also speeds up the exploitation of the next wave of technology.
What this means in practice — do this tomorrow
- Stop counting pilots; start counting changed ways of working. The number of pilots is a vanity metric. Ask instead: how many of our employees have permanently changed their weekly way of working with the help of AI? That figure predicts the impact on results better than any list of initiatives. If you don’t know the figure, that too is an answer — then this is exactly where measuring should begin.
- Map shadow use without blame. Somewhere in your organisation someone is already using AI on their own — probably many people. Don’t begin by forbidding it; begin by asking: in which tasks, for what need, with which tools? Shadow use is a gold mine for two reasons. It reveals where genuine need and the energy for change already are — that’s where the official roll-outs should start. And it reveals the risks that are better managed than driven underground. Responsible adoption does not mean blocking use; it means leading it.
- Give every AI initiative a human metric alongside the euro metric. The euro tells you whether the initiative paid off. The human metric tells you why it did or didn’t: how many use the tool weekly three months on, how many feel able to judge AI’s output, how many have trained a colleague. When you track both, you learn to tell the working initiatives from the showy ones — and, on the evidence of MIT’s data, that is exactly what separates the succeeding minority from the failing majority.
In closing: a chip is not skill at the game
I’ll condense this part into one thought: AI is a chip that gets you a seat at the table — skill at the game lives in people. Your organisation is not unique in its challenges, and that is good news, because you don’t have to solve a unique problem. You only have to take people’s learning as seriously as you take technology purchases. Few do. There is your opportunity.
But if technology doesn’t differentiate, why should AI matter any more than any earlier wave of technology? Because this shift is, in scale and in kind, of a different species than any before it — for the first time in history we can grow the output of thinking work without growing payroll costs, and for the first time office technology cuts across every function at once. I’ll tell you about that in the next part of the series.
I speak about these themes in keynotes, leadership-team coaching, and panel discussions. If you’d like to challenge your organisation’s thinking about AI and competitive advantage, get in touch → lenni@lennilaukkanen.fi
Previous part: You Can’t Copy People, and AI Won’t Become Your Competitive Advantage
Next part: Why the AI Transformation in Working Life Is Unlike Any Earlier Technological Shift