Most SMEs now use AI, but few get real value yet. Here is the gap.

Here is a number worth sitting with. In the 2026 OECD D4SME survey, published in April and covering more than 2.000 smaller companies across 12 countries, 61 percent of SMEs said they now use at least one AI-enabled application. Adoption, in other words, is no longer the hard part. Most smaller companies have crossed the line.

And yet the same survey found that 76 percent of those companies are still “AI novices”, using isolated, off-the-shelf tools rather than integrating AI into how they actually work. Only 6 percent report a transformational effect, and another 15 percent a significant one. So the picture is not “SMEs are behind on AI”. It is “SMEs have adopted AI, and then plateaued”.

I see this gap in almost every company I talk to. The interesting question is not how to start. It is how to get past the plateau.

Adoption is easy. Integration is the work.

Crossing the adoption line costs almost nothing today. You pay for a ChatGPT seat, you switch on the AI features in a tool you already use, you add a chatbot to the website. That is real, and it is fine. But none of it is connected to a workflow, and none of it is measured against time saved. It sits beside the work rather than inside it.

That is exactly what “novice” use looks like in practice. A team uses AI to draft emails and polish text, which is genuinely useful, while the same team still retypes order data from email into the ERP by hand every morning. The flashy, visible use of AI is switched on. The boring, expensive, repetitive process underneath is untouched. The tool got adopted. The operation did not change.

Why companies stall here

The plateau is not a failure of effort or ambition. It usually comes down to three things.

First, scattered tools solve scattered problems. Each one helps a little, none of them changes a core process, so the gains stay small and hard to feel.

Second, nothing is measured. If you never put a number on the hours a task costs, you cannot tell which AI use is earning its place and which is just pleasant. Without that number, every tool feels equally worthwhile, and none gets the investment to go deeper.

Third, the genuinely valuable step, connecting AI to a real workflow, is harder than buying a subscription. It needs someone to look at the process, the data underneath it, and the rules around it. That is work, and it is the work most companies skip.

What moving up looks like

Getting off the plateau does not mean buying more AI. It usually means doing the opposite: picking one process that matters and going deep on it instead of wide across ten tools.

Take the order-intake example. The novice version uses AI to write nicer customer emails. The integrated version connects the intake itself: orders arriving by email or PDF get read, structured, and dropped into the ERP automatically, with a person checking the exceptions. That is the move from a tool sitting next to the work to AI doing the work, and it is where the measurable time saving lives. One integrated, measured win teaches you more than ten scattered subscriptions.

Being a novice is a normal starting point

If you recognise your own company in that 76 percent, that is not a problem to be embarrassed about. It is the normal place to be in 2026, and it is a much better place than not having started at all. You have the tools, you have some comfort with AI, and your team is no longer scared of it. What is missing is focus: one process, measured honestly, integrated properly.

That focus is precisely what the AI-waardescan is built to create. We look at your real processes, find the one where AI would save the most measurable time, check whether the data underneath it is ready, and hand you a prioritized shortlist instead of another tool to try. The survey shows most SMEs have already done the easy part. The value is in the part that comes next, and that part is very doable.