ICT-gebruik bij bedrijven 2025 (CBS)
TL;DR
Chapter 3 of the CBS report Digitalisering en kenniseconomie 2025. It reports how Dutch firms with 10 or more employed persons use ICT, including AI.
- One in three firms uses AI. In 2025, 33% of firms with 10 or more employed persons used at least one of eight specified AI technologies. That was 23% in 2024 and 14% in 2023. The 2025 figure is provisional.
- Language technology leads. Text mining is used by 22% of firms and natural language generation by 16% (2024: 14% and 12%).
- Sectors differ fourfold. Information and communication 70%, financial services 59%, specialist business services 55%. Hotels and restaurants, transport and storage, and construction: 16% to 21%.
- 10% considered AI but did not use it (2024: 9%).
- The 2025 definition is wider. An eighth technology, generating images, video or sound, was added in 2025; until 2024 seven were asked.
- Size matters (StatLine). Use rises from 25% of firms with 10–20 employed persons to 73% of firms with 500 or more. Health and welfare (sector Q) is at 28%, health care (SBI 86) at 26%.
Key data
AI use by Dutch firms with 10 or more employed persons (table 3.4)
| Year | Use AI | Considered, not used |
|---|---|---|
| 2020 | 13% | – |
| 2021 | 13% | 5% |
| 2022 | 16% | 5% |
| 2023 | 14% | 6% |
| 2024 | 23% | 9% |
| 2025 (provisional) | 33% | 10% |
2022–2024 are labelled “nader voorlopig” (revised provisional).
By size class, 2025 (StatLine 86118NED): 10–20 persons 25% · 20–50 33% · 50–100 43% · 100–250 52% · 250–500 61% · 500+ 73%.
By sector, 2025 (StatLine 86120NED; total covers sectors C–N and Q): information and communication 70% · insurers 75% · banks 50% · specialist business services 55% · manufacturing 29% · trade 30% · health and welfare 28%, of which health care 26% and care and welfare 29% · construction 21% · transport 20% · hotels and restaurants 16%.
Method and definitions
- Unit. The enterprise, with 10 or more “werkzame personen” (persons employed). Each firm counts once, whatever its size.
- What counts as use. At least one of eight technologies: text mining, speech recognition, natural language generation, image recognition, generating images, video or sound, machine learning, AI-based robotic process automation, and service robots or autonomous vehicles. The eighth, image, video and sound generation, was added in 2025 (pop-up definition on the chapter page).
- Status. 2025 figures are provisional. StatLine’s 2025 tables note that “vanaf dit jaar is de wegingsmethode verbeterd, waardoor de cijfers licht kunnen verschillen ten opzichte van eerdere jaren” (“from this year the weighting method has been improved, so figures may differ slightly from earlier years”).
- Survey. The chapter names no survey and gives no sample size. Its 2021–2025 totals equal Eurostat’s Netherlands series for the same indicator to the rounded percent (13.1, 14.1, 23.06, 33.21). This is therefore the Dutch part of the EU survey on ICT usage in enterprises.
Scope and reliability
- Official statistics. A national statistical office, a fixed list of technologies and a survey harmonised across the EU. Executive surveys of large firms sit at the other end of the scale.
- The 23% to 33% step is not fully like-for-like. Three things changed in 2025: an eighth technology was added, CBS changed its weighting, and Eurostat flags the Dutch 2025 figure as a break in series. The chapter mentions only the first.
- Firm-weighted. Small firms dominate the count, so the total sits far below the rate for large firms (73% at 500+).
- Breadth only. “Use” means at least one technology in use. The survey says nothing about how deeply, how widely within the firm, or with what result.
- Sector coverage. CBS’s total includes financial services and health care. Eurostat’s EU total excludes both. The Dutch figure in Eurostat’s tables uses Eurostat’s own scope.
Dynamic-capabilities reading
digital-sensing/digital-scouting: an official, repeated scan of how many Dutch firms use which AI technologies, by sector and size.contextual/external-triggers: a third of Dutch firms with 10 or more employed persons, and nearly three quarters of the largest, already use some AI. That is the competitive field a firm operates in, whether or not it adopts AI itself.
How it connects
- enterprise-ai-adoption and the thread ai-maturity-measurement-comparison: an official firm-level adoption figure for the Netherlands.
- Eurostat, Use of AI in enterprises: the EU-wide version of the same survey, 19.95% in 2025.
- RaboResearch (Rabobank): AI use by business function in Dutch IT and business services, from the same CBS survey.
- Rendtorff-Smith (OECD): the gap between big and small firms in official statistics.
- Stanford DEL Adoption Monitor: 69% of firms in an executive survey of four high-income countries.
- Hatzius (Goldman Sachs): the US Census firm figure and its wording change.
- M&I AI Monitor Ziekenhuizen 2026: hospitals are large health-care organisations; CBS gives 26% for health-care firms as a whole.
What was actually ingested
The full chapter as served on 7 November 2026: four sections and four tables (internet access, telework, e-commerce, AI), and the pop-up that defines the eight AI technologies. The report’s erratum page corrects chapters 2 and 7 only. The size-class and sector split are not in the chapter. They were read from CBS StatLine tables 86118NED and 86120NED (published 12 December 2025) through the CBS open-data API and are appended to the raw file.
Linked entities and concepts
- enterprise-ai-adoption: concept.
Dangling (single-source mention, deferred): Centraal Bureau voor de Statistiek (CBS; also named in the body of the RaboResearch source), Eurostat.