Almost everyone already "uses" AI. And that's the problem.
The question that mattered two years ago — "will SMEs adopt AI?" — has been answered. In 2025, according to Eurostat, 17% of small European enterprises (10-49 employees) and 30% of medium ones (50-249) used AI tools, against 55% of large firms. A year earlier the figures were 11%, 21% and 41%. The curve is rising fast across every size band.
But the adoption number hides what matters. The OECD, in a 2025 report, reaches a far broader measure: 61% of SMEs use at least one AI-enabled application. The problem is what follows in the same sentence — 76% of those firms are "novice" users, with isolated, superficial uses and no integration into the business.
The distance between "17%" and "61%" isn't an error. It's the whole story: using AI has become easy and almost universal; capturing value from it remains rare.
Adoption is easy. Value capture isn't.
McKinsey, in its 2025 State of AI report, measures the same phenomenon globally with numbers that should stop any manager: 88% of organisations use AI in at least one function, but only 39% report any impact on operating profit (EBIT) — and of those, most say less than 5% of EBIT is attributable to AI. Only about 5% of organisations are "high performers" with material impact. And just 7% have scaled AI across the enterprise.
In other words: the bottleneck isn't access to the technology. It's the ability to turn it into results. The overwhelming majority of companies — large ones included — are stuck between "we already use AI" and "it changed something in the business."
What AI Changes — and Where It Doesn't Replace Judgement
Where value capture happens, AI changes three things in SMEs: speed of analysis, operational consistency, and the ability to scale without scaling headcount.
Speed means going from weeks compiling information to hours — not because AI replaces the analyst, but because it removes the work that stops them from thinking. Consistency means documenting and replicating what depended on one person's tacit knowledge: onboarding, proposals, reports. Scale means doing more without hiring proportionally.
There is, however, territory AI doesn't cover: judgement in anomalous context. When the client is unusual, when the market shifts unexpectedly, when the decision involves a long-term relationship — there, AI informs, it doesn't decide. The useful boundary isn't simple versus complex. It's stable context versus context that shifted in ways nobody anticipated.
What Separates Winners from Those Stuck in a Pilot
Here is the most actionable finding of all. McKinsey tested 25 organisational attributes to see which weighs most on EBIT impact. The winner wasn't budget, size, or the tech stack: it was workflow redesign. And only 21% of organisations have fundamentally redesigned at least some of their workflows.
This is why so many AI projects die in the pilot. A 2025 MIT study made headlines by finding that roughly 95% of generative-AI pilots produce no measurable bottom-line impact — a figure contested on methodology, and one to read as a signal, not a verdict. But the same study's robust findings are revealing: the biggest return was in the back office, not the marketing where most budgets were spent; and buying from specialised vendors succeeded about twice as often as building internally.
Translated for an SME: the advantage doesn't come from buying the most advanced tool. It comes from clarity (which process, which indicator, which stopping criterion) and cadence (implementation doesn't end at go-live — it ends when the team uses the new process without thinking about it).
Portugal Is Behind — and That's What Opens the Window
In 2025, 11.5% of Portuguese enterprises used AI — below the European average of 20%. It's easy to read this as bad news. It's the opposite. At a moment when almost everyone "uses" AI but almost no one captures value, a market where adoption is still low is a market where the advantage of executing well is larger, not smaller. The window to build competitive distance with AI remains wide open in Portugal.
How THE iCON Approaches This
THE iCON's work with SMEs on AI adoption always follows the same sequence: diagnosis, prioritisation, supervised implementation.
Diagnosis means mapping where team time goes and where output quality is most variable — not where AI is most impressive, but where the problem is most real. Prioritisation means choosing the first process by three criteria: high impact, low adoption friction, and a measurable result in under 60 days. Supervised implementation means staying until the process works, the errors are documented, and the team can iterate without us.
It's no accident that this matches what the data points to as the differentiator: process redesign and execution, not buying a licence. The result isn't an SME with more technology. It's an SME with a new capability — and the structure to expand it.
Conclusion
The competitive advantage of AI doesn't belong to whoever adopts it first, nor to whoever has the biggest budget. It belongs to whoever turns it into results — and the numbers show that, today, that is still a minority. For SMEs that grasp this quickly, and that treat AI as a capability to build rather than a tool to install, the window is still open. For the rest, it's closing.