AI adoption has a small-business gap

NEWS & ANALYSIS · AI, MACHINES & THE ECONOMY

OCTOBER 16, 2026 · THE WEEKDAY BRIEF

A useful AI product can still be a difficult purchase. Someone has to connect it to the work, teach people how to use it, and keep it working when the inputs change.

That makes implementation a business worth examining alongside the models themselves.

In Eurostat’s 2025 survey, 17% of small EU enterprises used AI, compared with 30.36% of medium enterprises and 55.03% of large enterprises. “Small” means 10–49 employees and self-employed people; “large” means at least 250. These are 2025 observations, not a live reading of October 2026. Source: Eurostat.

The measure covers multiple AI technologies, not just chatbots. It records use, not profitability. Its enterprise and sector coverage also means these figures should not be treated as a count of every European business, particularly businesses with fewer than ten people. Survey methodology.

Among nonusers that had considered AI, 70.89% cited a lack of relevant expertise. That is a conditional group, not 70.89% of all businesses. Source: Eurostat.

Our interpretation: there may be room for businesses that make a narrow workflow easy to adopt. The survey does not prove that implementation services cause adoption or that any particular service will be profitable.

Consider a hypothetical distributor that wants help turning incoming quote requests into draft responses. A useful offer would include a clean product catalogue, a connection to the existing inbox, staff training, and a process for ambiguous requests. The customer would buy a working quoting routine with defined support.

For the supplier, the central question is how much of that work can be reused. If every customer needs a different integration, different data repair and hours of individual coaching, recurring revenue may come with recurring labour. A subscription label does not settle the economics.

Installation effort. Record the hours needed to get the first customer operating, then compare them with the next similar customer. Reuse should show up in delivery time.

Independent use. Check whether the customer can complete the workflow without the installer standing beside them. Training that ends in dependence creates an ongoing service obligation.

Support burden. Track how often ordinary changes—a new product, a revised document, a new employee—require specialist intervention.

For buyers, these measures help reveal what the ongoing bill is buying. For suppliers, they test whether a repeatable product is emerging from the service work.

The opportunity to investigate is the distance between a successful demonstration and a customer who can run the system on an ordinary Tuesday.

Machines and Money Team