In 2025, 12.2% of Latvian enterprises used artificial intelligence, against 42.0% in Denmark, 37.8% in Finland and 35.0% in Sweden. The gap is real, but one big project will not close it. Pick one repeatable task, name the target and keep a person checking the result.
What the data says
The figures come from the EU survey on ICT use in enterprises, run by national statistics offices. The indicator is the share of enterprises with 10 or more employees that used at least one AI technology in 2025.
| Country | Enterprises using AI, 2025 |
|---|---|
| Latvia | 12.2% |
| EU average | 20.0% |
| Sweden | 35.0% |
| Finland | 37.8% |
| Denmark | 42.0% |
Denmark and Finland grew by 14.5 and 13.5 percentage points in a year; the EU average grew by 6.5 (Eurostat, 11 December 2025). The Latvian figure is from the Central Statistical Bureau (CSP), published on 4 November 2025, up 3.4 points on 2024 (CSP).
Two limits. Companies under 10 employees are not counted. And the indicator counts any of eight AI technology types, including text analysis and content generation, not only chatbots.
Where Latvia lags most
CSP shows a wide gap by size: 9.8% of small enterprises (10-49 staff) use AI, 19.8% of medium (50-249) and 47.1% of large (250+). Small firms pull the overall figure down.
One more number matters more than the gap. Among Latvian enterprises that considered AI but did not adopt it, 63.2% named lack of suitable expertise as the main barrier. The EU pattern is the same: among enterprises that considered AI and did not adopt it, 70.89% cite lack of expertise, 52.52% unclear legal consequences and 48.83% data protection concerns (Eurostat, 2025).
So the obstacle is rarely money or technology. It is more often not knowing where to start or what is allowed with customer data.
Why chasing Denmark is the wrong goal
“42%” does not mean Danish firms earn 42% more. It only says a tool is in use, and the indicator cannot tell serious process automation from staff occasionally asking a chatbot to draft text.
A better goal is concrete: “answer an enquiry within an hour” or “stop keying in invoices by hand”. We covered where AI pays off and where it does not in AI automation for small business. The short version: if the task repeats, the output can be checked and a mistake is not a disaster, AI fits.
Three first processes
This is our recommendation from project work, not a statistic. A small company usually has three places where AI helps with little risk.
- Incoming e-mail and enquiries. AI sets topic and urgency and drafts a reply; a person approves it. Our helpdesk product FDesk works this way: AI triages tickets, a person sends the answer.
- Invoices and documents. AI reads data from a PDF or scan; an accountant checks and approves. For this we have Invoa, which reads invoice data.
- Repeated customer questions. A bot answers from your own content and hands anything difficult to a person. See AI customer service with human hand-off.
In each case a person stays in the loop. It is slower than full automation, but a mistake costs little and trust grows.
Before you start: data and law
The three main barriers are expertise, legal consequences and data protection, so settle them before you choose a tool.
- Data. Find out where the service processes data, whether it is used for model training and whether a data processing agreement exists. We offer solutions where data stays in the EU.
- Transparency. If a customer talks to an AI assistant, say so. What applies and from when is in our article on the EU AI Act for small companies.
- Literacy. Staff who use AI should know what they are doing and where the model fails. It is the cheapest step and addresses the expertise gap directly.
How to know the start worked
Before you begin, write down two numbers: how long the task takes now and how many errors it has. Measure again after 4-6 weeks. If time dropped and errors did not rise, move to the next process. If not, you bought a cheap conclusion instead of a year-long project.
Conclusion
Statistics show the gap, not what to do about it. The rule: if you have a weekly repeatable task whose output can be checked, start there and keep a person in the loop; if you have no such task or the process is not written down, fix the process first instead of buying a tool. In the first case the project can be small and quick.
We build AI automation and assistants and choose with you the first process where AI will pay off. Discuss your project.
Sources
- Eurostat, Use of artificial intelligence in enterprises, 2025 data, dataset isoc_eb_ai, December 2025
- Eurostat, AI use in EU enterprises (ddn-20251211-2), December 2025
- Central Statistical Bureau of Latvia, ICT use in enterprises, November 2025
