88% of companies are using AI. 6% are getting results from it. The math is not complicated.
Somewhere in your LinkedIn feed right now, a CEO is announcing that their company is going “AI-first.”
They have not defined what that means. They have not named a specific problem it will solve. They have not budgeted for operations beyond the pilot. They have just decided, with the conviction of someone who read a McKinsey summary on a flight, that the announcement itself constitutes a strategy.
Their peers will like the post. Their board will nod. Their engineering team will quietly open a new Jira board and start calling meetings.
And somewhere between six and eighteen months from now, the initiative will be quietly shelved, rebranded as a “learnings” period, and replaced with a fresh announcement about their next transformation.
This is not a prediction. It is a description of what is already happening.
The Fear That Ate the Strategy
In September 2025, Censuswide and Thoughtworks surveyed 3,500 IT decision-makers and C-suite executives across seven global markets. They found that 56% of executives globally say they feel competitive pressure to adopt AI quickly. In Singapore the number was 66%. In India, 62.8%.
A separate ABBYY survey of IT leaders found that 63% report they are worried their company will be left behind if they do not use AI.
Left behind. That is the phrase. Not “unsolved operational problem.” Not “competitive gap in a specific capability.” Just a generalised terror of being in the wrong place when history moves.
This is FOMO dressed in a business suit. And unlike the kind you feel scrolling through someone else’s holiday photos, this one comes with a budget approval and a project manager.
The question nobody is stopping to ask is: left behind in what, exactly?
What the Numbers Actually Say
Here is what AI adoption looks like in practice right now, according to people who actually measure it.
McKinsey’s 2025 State of AI survey covered 1,993 respondents across 105 countries. It found that 88% of organizations now use AI in at least one business function. Usage is up from 78% a year earlier. AI is everywhere.
The same survey found that only 6% of those organizations qualify as AI “high performers” — meaning AI has contributed more than 5% to their EBIT. One percent of C-suite respondents describe their generative AI rollouts as mature.
Sit with that for a moment. 88% claim adoption. 1% are doing it properly.
MIT’s NANDA initiative published a 2025 report based on 150 interviews with business leaders, a survey of 350 employees and an analysis of 300 public AI deployments. It found that 95% of generative AI pilots at large companies are failing to produce rapid revenue results. Custom enterprise AI tools are reaching production at a rate of about 5%.
42% of companies abandoned most of their AI initiatives in 2025. That figure was 17% in 2024. The abandonment rate nearly tripled in one year.
These are not the numbers of an industry thoughtfully adopting a transformative technology. These are the numbers of an industry running after a bus.
“Step One: Use LLMs. Step Two: Figure Out What For.”
Marina Danilevsky, Senior Research Scientist at IBM Language Technologies, put it more directly than most: “People said, ‘Step one: we’re going to use LLMs. Step two: What should we use them for?’”
That sentence is funnier than it has any right to be. It is also an accurate description of how the majority of enterprise AI projects begin.
A Gallup survey from late 2024 found that only 15% of US employees report that their workplaces have communicated a clear AI strategy. Not a detailed strategy. Not a good strategy. Just any strategy at all.
The EY 2025 Work Reimagined Survey covered 15,000 employees and 1,500 employers across 29 countries. It found that 88% of employees use AI at work. It also found that only 5% are using it in ways that genuinely transform how they work. The other 83% are mostly using it for search and document summarisation — tasks they were already doing, slightly faster.
92% of executives in McKinsey’s survey plan to increase their AI spending in the next three years.
The pattern is not ambiguous. Companies are buying the technology before identifying the problem. They are building capability before establishing need. They are scaling investment while the evidence for return is still missing.
Klarna: The Cautionary Tale That Nobody Learned From
In 2023, Klarna CEO Sebastian Siemiatkowski announced that his company had stopped hiring. AI was the future. By 2024, Klarna had partnered with OpenAI, slashed its headcount from roughly 7,000 to around 3,000 and declared publicly that its AI chatbot was doing the work of 700 customer service agents. The CEO told the world that “AI can already do all of the jobs that we as humans do.”
The board loved it. Tech media loved it. Every LinkedIn thought leader who had recently discovered the word “disruption” cited Klarna as proof that the future had arrived.
By mid-2025, Klarna was quietly hiring again.
Customer satisfaction had dropped. Users complained about robotic responses, inflexible scripts and the experience of explaining their problem to a bot, watching it fail, and then explaining it all over again to the human who came after. Siemiatkowski told Bloomberg: “As cost unfortunately seems to have been a too predominant evaluation factor when organizing this, what you end up having is lower quality.”
The company’s grand AI-first transformation is now an “Uber-style” gig workforce model. The human support staff whose jobs were eliminated are being invited back, not as employees, but as contractors with flexible hours and no stability.
The $10 million in savings Klarna celebrated? The costs did not disappear. They moved. Customer churn costs money. Reputation damage costs money. Hiring back the workforce you just eliminated, at speed and under pressure, costs money. Research from Bain and Harvard Business Review consistently puts the cost of acquiring a replacement customer at anywhere from five to twenty-five times more than retaining the one you had.
Klarna is not a uniquely bad actor. It is just the company that was loud enough to be specific. The same calculation happened in quieter offices across every industry.
An Orgvue survey found that 55% of companies that executed AI-driven workforce reductions now regret them.
Duolingo Did It Faster
In April 2025, Duolingo’s CEO Luis von Ahn sent an internal memo declaring the company would be “AI-first.” Teams wanting to request new headcount would need to prove the work could not be done by AI. Contracts with 10% of the company’s contractor workforce were terminated.
The backlash was immediate and severe. Users threatened to cancel subscriptions. The story was everywhere within days.
One week later, von Ahn published a post on LinkedIn walking it back. “To be clear: I do not see AI as replacing what our employees do,” he wrote. “I see it as a tool to accelerate what we do.”
From “AI replaces workers” to “AI is just a tool” in seven days. That is not a nuanced recalibration. That is a company discovering, at speed, that the announcement it made for investor optics had consequences in the real world.
The Real Cost Is Not the Budget Line
The number in the failed AI project write-off gets a lot of attention. It should not get all of it.
When a company rushes AI adoption because it is afraid of looking behind, the costs are not only financial. They are structural.
People who were hired for a skill are told that skill is now worthless, before that claim has been tested against reality. They update their sense of their own value accordingly, and they do it accurately, because the company just told them outright that they are replaceable. Some of them leave. The ones who stay are watching.
Processes that were working adequately get disrupted to accommodate a tool that is not yet ready to replace them. The disruption is real. The gain is not.
Vendors who have noticed that “AI” unlocks procurement budgets adjust their positioning accordingly. The product has not changed. The deck has. Research from Gartner estimates that of the thousands of vendors currently claiming to offer agentic AI capabilities, only around 130 actually deliver them. The rest updated their marketing materials. That is agent washing, and it exists because FOMO creates buyers who do not ask hard questions.
Trust, once lost, is expensive to rebuild. Klarna’s CEO can rehire contractors. He cannot unhire them from the workers’ memory of what the company decided they were worth.
What Actual AI Adoption Looks Like
The 6% of companies that McKinsey identifies as genuine high performers share a pattern. They identify a specific problem first. They quantify what success looks like before they start. They redesign the workflow, not just the tool that sits on top of it. They invest in people and process alongside the technology.
McKinsey found that workflow redesign has the biggest measurable effect on whether AI produces EBIT impact. Not the model. Not the vendor. The willingness to rethink how work is done.
This is not exciting. It does not make for a good LinkedIn announcement. There is no moment where a CEO gets to declare their company “AI-first” and watch the engagement roll in.
There is just a team that spent three months mapping a specific process, identified where AI could handle the predictable parts, kept humans for the parts that require judgment, measured the outcome and adjusted based on what they found.
It is the difference between buying a drill because everyone else is buying drills and buying a drill because there is a specific hole that needs to be in a specific wall.
The drill is the same. The result is completely different.
The Question That Cuts Through the Noise
Before the next AI investment gets approved, before the next “AI-first” strategy gets announced, before the next vendor gets invited in to demonstrate their product, one question should be asked and answered.
Not “what AI tools should we buy?” Not “which competitors are using AI?” Not “what does our AI roadmap look like?”
The question is: what specific problem are we solving, and how will we know when it is solved?
If that question does not have a clear answer, the project is not ready. The budget should not be approved. The vendor should not be let in the door.
Competitive pressure is real. AI is genuinely transformative in the hands of organizations that use it deliberately. The 6% who are getting results did not get there by being afraid of being left behind. They got there by being specific about what they needed.
The other 94% were very busy. They had meetings. They bought tools. They made announcements. They are, at this moment, preparing their next initiative.