AI in Spanish companies: real usage, size differences, and data limits

When a company claims to use artificial intelligence, the figure alone does not indicate whether the technology is integrated into an essential operation, limited to a single task, or used by a small team. Declared usage and adoption depth are distinct measures: the former informs of presence; the latter would require knowing frequency, scope, investment, and results.

The National Statistics Institute (INE) survey offers indicators on business technologies and uses. Its percentages allow observing the diffusion of these technologies, but they do not detail how much work each tool covers or how it integrates into processes. The user company rate does not by itself measure productivity or profitability; to evaluate those outcomes, each organization needs indicators linked to the specific activity it aims to improve.

This distinction helps read figures and technological forecasts with prudence. A list of 2026 trends can describe market possibilities, but it does not replace a measurement of usage in Spanish companies. To make decisions, it is advisable to separate what a statistic counts —the declared presence of a technology— from what it does not quantify, such as its effect on costs, times, or quality.

The INE edition published on October 22, 2026, places the proportion of companies with 10 or more employees that used artificial intelligence technologies at 33.6% in the first quarter of 2026. In that same group, 39.4% used paid cloud computing services. Both data points describe usage during a specific period, not a forecast for the entire year.

Comparison with companies of fewer than 10 employees shows a difference in AI: the rate was 21.36%, compared to 33.60% in larger companies. There is also a difference in internet connection: 81.21% of small companies had a connection, compared to 98.33% of those with 10 or more employees. These are distinct indicators: internet connection describes access, while AI usage measures another business practice.

E-commerce is measured with reference to the previous year, not the first quarter of 2026. In 2025, 28.97% of companies with 10 or more employees made sales through this channel, compared to 11.81% of smaller ones; the business volume generated by electronic sales was 419,706 million euros. The temporal reference matters: these percentages do not all describe the same period.

The statistical category of AI groups different technologies: text analysis, voice and image recognition, machine learning, language or audiovisual content generation, and workflow automation, among others. Therefore, the aggregate indicator does not equate to a generative AI usage rate nor does it allow knowing how many companies employ conversational assistants.

INE also classifies application areas, such as marketing and sales, production, administration, logistics, security, human resources, accounting, and R&D. This list helps situate possible uses in business activity, but the aggregate percentage does not specify how many decisions each system automates, how much work it covers, or how many people use it.

Business trends for 2026 usually include generative AI integrated into management programs, advanced automation, visual development tools, edge computing, and digital twins. These are possibilities described in sectoral analyses, not results from the Spanish survey. A technological forecast does not prove widespread adoption: to sustain that conclusion, an indicator measuring the usage of the corresponding technology is needed.

Cloud computing is one of the technologies the survey measures separately. INE describes these services as IT resources available via the internet —for example, software, processing capacity, or storage— hosted on provider servers and adjustable to user needs. 39.4% of companies with 10 or more employees used paid cloud services in the first quarter of 2026.

That percentage does not distinguish which workloads have been migrated or what type of service each company uses. It should also not be confused with AI: the cloud can serve as infrastructure for artificial intelligence applications, but it is also used for other functions. Cloud and AI are related indicators, but not interchangeable.

The statistical operation includes variables on data usage, exchange, commerce, and analysis, in addition to ICT spending information. Spending reflects allocated resources; declared usage, the presence of a technology in business activity. Neither of these measures, considered separately, informs of the economic outcomes obtained. For a company, distinguishing between acquisition, utilization, and effect allows identifying which data to review before expanding an investment.

In the indicators published for 2026, companies with fewer than 10 employees show an AI usage rate of 21.36%, compared to 33.60% among those with 10 or more. The difference also appears in the proportion with a corporate web page —36.15% versus 82.92%— and in e-commerce sales —11.81% versus 28.97%—. Size is associated with adoption differences, although those data alone do not establish their causes.

In 2024, Adigital pointed out obstacles to technological adoption in SMEs, including lack of resources, knowledge, and training, as well as the complexity of some solutions. That analysis cited previous DESI indicators, so it does not describe current INE rates. It does provide context on barriers that can hinder the incorporation of tools: having resources, capabilities, and solutions tailored to the activity.

For an SME, the practical reading is not that it must incorporate all technologies included in trend lists. It can start by defining a specific process and assessing the cost of a solution, including its integration and necessary training. Afterwards, it must evaluate the result with metrics linked to that process. The right tool depends on the operational problem, not on a technology appearing in a trend list.

The INE survey offers comparable indicators on technologies, sizes, and economic activities. Its methodological sheet explains that ICT usage variables usually refer to the first quarter of the survey year, while e-commerce and other general information correspond to the previous year. Keeping both periods visible avoids comparing data as if they were simultaneous.

There is also a change in statistical unit since the 2021-2022 edition: in certain business groups, the “statistical company” can integrate several legal units. INE warns that the change affects results and their breakdowns; since the 2022-2023 exercise, it disseminates results with the new approach. Historical comparisons require checking that the definition is equivalent before interpreting a variation as a real change in adoption.

The data show that a relevant proportion of Spanish companies declared using AI in the first quarter of 2026, with a higher rate among those with 10 or more employees than among smaller ones. They also record the use of paid cloud and electronic sales in their respective periods. To make decisions, a company must identify the process it wants to improve and measure its results with indicators suitable for that process.