How many Dutch SMEs use AI, and what for?
How many SMEs use AI depends mostly on how big you are. In 2025, 13.8 percent of Dutch micro-enterprises used AI, against 45 percent of businesses with 50 to 250 employed persons (CBS, 12 December 2025). Across the EU it was 17% of small, 30.36% of medium and 55.03% of large enterprises (Eurostat, 2025). Of the businesses that use AI, 35 percent do so for marketing or sales. These figures only say whether a business uses AI, not how often, how well, or in which channel.
Published on
A lot gets said about AI in Dutch small and medium-sized business, and very little of it gets measured. This article does the opposite. Everything below comes from figures published by Statistics Netherlands (CBS) or Eurostat themselves, with the year and the source attached. No estimates, no adjusted percentages, no retelling of a retelling. Where a number is missing, you will read that it is missing instead of getting an approximation.
That is not caution for its own sake. If you want to know whether your business is ahead or behind, you need a yardstick that actually holds. A percentage that came from somewhere, with a denominator nobody can name, is not a yardstick but a feeling. And where AI is concerned, the feelings are strong enough already.
How many businesses use AI, and how fast is that moving?
The sharpest growth figure comes from CBS, published on 12 December 2025. Among businesses with 50 to 250 employed persons, the share using AI rose from 20 percent in 2023 to 45 percent in 2025. That is more than a doubling in two years, and it is the strongest increase across all the size classes CBS measures.
Put the other half of that sentence next to it right away. Behind that 45 percent sits a majority that still was not using AI in 2025. If you are in that group, you are not the exception. You are in the majority, and that majority is shrinking fast. Both observations come out of the same number, and both are true. Quote only the first half and you are selling urgency; quote only the second half and you are selling calm. The figures do neither.
Notice what the number does not say. CBS is measuring here whether a business uses one or more AI technologies. Not how often, not how central it is, not whether it works. A business where one employee occasionally has a text rewritten counts exactly as heavily in this statistic as a business that has rebuilt its entire customer contact around an AI system. I come back to that later, because it is the single most important limitation of the whole figure.
Why does "SMEs" as one group tell you nothing here?
If you take one thing away from this article, make it this: AI use tracks company size more closely than almost anything else. A figure about "SMEs" throws a two-person shop and a two-hundred-person company onto the same pile, while in this statistic those two live in different worlds.
Eurostat shows the same thing across the whole EU for 2025: 17% of small enterprises, 30.36% of medium enterprises and 55.03% of large enterprises used AI. From small to large, that is more than a tripling. It is the same staircase CBS sees in the Netherlands, and it points the same way in almost every European country.
The practical conclusion is simple. Compare yourself against your own size class, not against an average that has two-hundred-person companies in it. Otherwise what you are mostly measuring is how big you are.
| Group | Area | Share using AI | Year | Source |
|---|---|---|---|---|
| Micro-enterprises (2 to 10 employed persons) | Netherlands | 13.8 percent | 2025 | CBS |
| 50 to 250 employed persons | Netherlands | 20 percent | 2023 | CBS |
| 50 to 250 employed persons | Netherlands | 45 percent | 2025 | CBS |
| Small enterprises | EU | 17 percent | 2025 | Eurostat |
| Medium enterprises | EU | 30.36 percent | 2025 | Eurostat |
| Large enterprises | EU | 55.03 percent | 2025 | Eurostat |
What are the smallest businesses doing, and why do you almost never see them?
Here is a gap that most articles on this subject quietly skip. The regular ICT statistics on businesses only start at ten employed persons. Everything below that falls outside by default. And that is exactly where the vast majority of Dutch businesses sit.
CBS studied that group separately. In 2025, 13.8 percent of Dutch micro-enterprises used one or more of seven listed AI technologies. That was an increase of more than 3 percentage points compared with 2024. Set against larger businesses, that share is clearly lower.
Put that 13.8 percent next to the 45 percent for businesses with 50 to 250 employed persons and you see the gap in a single glance. Same country, same year, same statistical office. The only difference is the number of people on the payroll. Anyone claiming that "SMEs are switching to AI en masse" has looked at the top of that range and left the bottom out.
One more detail to know before you apply this figure to yourself: micro-enterprises here means businesses with 2 to 10 employed persons. If you work on your own, you fall outside even this separate measurement. For the self-employed person without staff, this series simply contains no percentage. That is not evidence that the group does not use AI. It means it was not counted here.
What do businesses use AI for most often?
Of the businesses that use software or systems based on AI, 35 percent do so for marketing or sales. That is the largest application. Next comes business administration or management tasks at 32 percent, under which CBS places things like risk assessment and the selection of job applicants. Both figures are from CBS and cover 2025.
Watch the denominator, because it shifts here. This 35 percent is not about all businesses. It is about the share of businesses that already use AI. In a group where, say, 45 percent uses AI, 35 percent of those is therefore not 35 percent of all businesses. This is the most common mistake made when these figures get retold, and it systematically makes the picture look broader than it is.
There is another limit on what you may read into this. "Marketing or sales" is a wide category. Customer contact is in there, but so is writing ads, translating texts and scoring leads. The figure does not tell you how many businesses actually let an AI system speak up in email, WhatsApp or on the phone. That breakdown is not published, so I do not quote one here.
| Breakdown | Share | Of which group | Year |
|---|---|---|---|
| Marketing or sales | 35 percent | businesses that use AI | 2025 |
| Business administration or management tasks | 32 percent | businesses that use AI | 2025 |
| Information and communication sector | 54 percent | all businesses in that sector | 2025 |
| Information and communication sector | 28 percent | all businesses in that sector | 2023 |
In which sector is AI the most ordinary?
Businesses in information and communication use one or more AI technologies most often, at 54 percent. In 2023 that was still 28 percent. There too the share nearly doubled in two years, and that sector leads all others by a distance.
This is less surprising than it sounds, which is precisely what makes it useful as a reference point. These are the businesses that build software and deliver digital services. They have the knowledge in house and the threshold is low. If you run an installation company, a clinic or a garage, 54 percent is not your yardstick. It is the upper bound of what gets measured in the Netherlands, and the distance between that upper bound and the average tells you how skewed the picture is that you meet in the media.
CBS publishes its own percentages for the other sectors. Those are not among the sources listed under this article, so I do not name them here. If you want the figure for your own sector, look up the CBS publication of 12 December 2025 and read the breakdown there. That is one click, and it beats a number I would have to round off.
How does the Netherlands compare with Europe?
It is tempting to divide the Dutch figures by the European ones and conclude that the Netherlands is ahead or behind. Do not do that without thinking. The CBS series and the Eurostat series are different surveys, with their own definitions and their own size classes. A difference of a few percentage points between two such series can just as easily be a measurement difference as a real one.
What you can do is compare the shape. And the shape is strikingly similar. In both series AI use scales hard with company size, in both the sharpest growth sits in the mid-sized segment, and in both 2025 is the year the mid-sized segment moved towards half. In the Eurostat figures the medium segment sits at 30.36% in 2025 and large enterprises at 55.03%; in the CBS figures the Dutch group of 50 to 250 employed persons sits at 45 percent in that same year. The trend is the same; the levels do not lay side by side that easily.
Where the European figures do add something is on the question of why businesses are not doing it. That is what the next section is about.
Why do more than half stall on the legal side?
In 2025 Eurostat asked EU enterprises that had ever considered using AI technologies why they ultimately did not. The most common reason was a lack of relevant expertise, at 70.89%. Next came lack of clarity about the legal consequences, at 52.52%.
That second number is the most interesting one in this entire article for you as a business owner. More than half of the businesses that seriously considered AI keep the brake on because they do not know what applies legally. Not because the technology cannot do what they want. Not because they do not need it. Because the rules feel unclear.
That is an obstacle of a completely different kind from the other two. Missing technology has to be bought and missing knowledge has to be built up, but uncertainty about the rules is largely solvable by working it out properly once. The obligations around AI in customer contact do exist: you need a legal basis for the personal data flowing through it, you have to record what your supplier may do with that data, and since the AI Act it has to be clear that the customer is talking to a machine. Those rules are worked out topic by topic elsewhere in this knowledge base. The point here is statistical: that uncertainty is not a fringe issue, it is the second-largest reason it does not happen.
What do these figures not measure?
This is the section most pieces on AI adoption leave out. The figures above are reliable for what they measure, and there are quite a few things they do not measure.
- They do not measure intensity. "Uses AI" is a yes or no. One employee occasionally having an email rewritten produces the same yes as a fully automated customer service desk. So from a rise of 20 to 45 percent you cannot conclude that twice as much work is being done by AI.
- They do not measure results. Nowhere in these series does it say whether the AI use delivered anything, saved money or helped customers. Adoption is not success.
- They do not measure compliance. A business using AI without a single agreement on paper counts in the statistic exactly the same as a business that has arranged everything properly. Nothing about how well it is going follows from these figures.
- They do not measure your channel. The breakdown covers "marketing or sales" and "business administration", not email, WhatsApp or phone separately. Anyone quoting a percentage for AI in customer contact took it from somewhere it was not stated.
- They do not measure the one-person business. The regular series starts at ten employed persons; the separate micro-enterprise study starts at two. The owner working alone does not appear in it.
- They measure what is on the list. CBS asks about a fixed set of seven AI technologies. If you use something that does not fall under those, you do not count in this figure, however clever it is.
- They lag. A publication from December 2025 describes the year 2025, and in a market that moves by the quarter that is a snapshot. It is the best one available, but it is one from yesterday.
How do you read a percentage like this without fooling yourself?
Three questions strip out most of the nonsense before you repeat a figure in a meeting or a proposal.
- What is the denominator? 35 percent of AI users is something entirely different from 35 percent of all businesses. If the denominator is not stated, you do not know what you are reading.
- Which size class? Between 13.8 percent for micro-enterprises and 45 percent for businesses of 50 to 250 people there is no disagreement, only a different headcount.
- Which year, and who measured it? With these figures, 2023 and 2025 are two different worlds. And a percentage without an author is a percentage you cannot check.
Ask these three questions of every AI figure that crosses your desk and most of them filter themselves out. Not because the figures are wrong, but because they lost their context somewhere along the way.
What can you do with this as a business owner?
Three things, and none of them is advice to buy or to wait.
First: hold your own situation against your own size class. If you employ five people, 13.8 percent is your reference, not the 54 percent of the ICT sector. If you already do something with AI, you are in that small group of front-runners. If you do nothing, you are in the broad majority. Both are normal, and neither is a reason in itself to act or to hold off.
Second: the growth figures point one way. From 20 to 45 percent in two years for the mid-sized segment, from 28 to 54 percent over the same two years in information and communication. What is still the exception at your size today is probably less so in two years. That is not a prediction of mine; it is the direction of the measured series.
Third: the biggest brake is not technology. Knowledge and legal uncertainty top the list of reasons not to start, together producing a picture in which most businesses do not stop because it cannot be done, but because they do not know where they stand. That is exactly the kind of obstacle that disappears once you work it out and write it down. Which legal basis, which agreements with your supplier, how long you keep things and what you tell the customer: these are all questions with an answer. That is more than you can say for most business risks.