# AI Training in Companies: What Official Statistics Say

> How many companies train staff on AI? Official statistics don't measure it, but they show adoption, the SME gap and skills-related barriers.

Source: Giallo Studio (giallostudio.it) · Page: https://giallostudio.it/en/blog/formazione-ia-imprese-statistiche-ufficiali · Published: 2026-11-10 · Updated: 2026-11-10

Official statistics do not say how many companies train their staff on artificial intelligence. ISTAT, Eurostat and the European Commission measure something else: how many companies use AI, how many have evaluated it, and why they stopped. The direct answer to the question is therefore that, in the official materials we read, the data on training does not exist.

By "AI training" we mean here any activity through which a company teaches people to use these tools at work: courses, side-by-side coaching, internal rules, practice exercises. It is not the same thing as adoption, meaning the use of at least one AI technology. A company can use AI without having trained anyone, and can have trained some people without yet really using it. For this reason it is best not to infer one from the other.

There is, however, an indirect clue, and it is a solid one: lack of skills is the most cited barrier among companies that do not adopt. Among Italian companies that do not use AI, [ISTAT](https://www.istat.it/comunicato-stampa/imprese-e-ict-anno-2025/) puts lack of skills at 58.6%. In this article we read the numbers carefully, separating scope, year and interpretation, and we look at what an SME can do with them.

<div class="key-points">
<p class="kp-title">In summary</p>
<ul>
<li>None of the official sources we read measures the share of companies that train staff on AI: the data concern adoption and barriers.</li>
<li>According to <a href="https://www.istat.it/comunicato-stampa/imprese-e-ict-anno-2025/">ISTAT</a>, in 2025 16.4% of companies with at least 10 employees use at least one AI technology, up from 8.2% in 2024.</li>
<li>Again per <a href="https://www.istat.it/comunicato-stampa/imprese-e-ict-anno-2025/">ISTAT</a>, SMEs reach 15.7% and large companies 53.1%: the gap is growing.</li>
<li>Lack of skills is the most cited barrier: 58.6% according to <a href="https://www.istat.it/comunicato-stampa/imprese-e-ict-anno-2025/">ISTAT</a> and 70.89% according to <a href="https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Use_of_artificial_intelligence_in_enterprises">Eurostat</a> among EU companies that considered AI.</li>
<li>Figures from different sources have different scopes: they should be dated and not compared lightly.</li>
</ul>
</div>

## How many companies train their staff on AI?

We do not know from official statistics. None of the materials we collected reports the share of companies that train employees on artificial intelligence. Anyone who cites a precise percentage on this point, attributing it to ISTAT or Eurostat, has either read another survey or is stretching an adoption figure.

It is a useful absence to state openly. If a course provider, an article or a presentation says that "most companies train their staff", the right question is: which official survey does that come from? The sources we read do not allow us to answer, so we do not do so here either.

<div class="callout">
<p><strong>Reading rule.</strong> A figure on AI use is not a figure on training. If the direct measure is missing, we say it is missing, and we use the barriers reported by companies as a clue, not as proof.</p>
</div>

## What do ISTAT, Eurostat and the Commission really measure?

The three bodies look at the same phenomenon from different angles, and their numbers do not overlap. [ISTAT](https://www.istat.it/comunicato-stampa/imprese-e-ict-anno-2025/) measures AI use in Italian companies with at least 10 employees. [Eurostat](https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Use_of_artificial_intelligence_in_enterprises) measures use in European Union companies and states that the survey covers companies with at least 10 employees, including self-employed workers. Eurostat then splits companies by size: small, medium and large, the latter with 250 or more employees.

The [Country Report on Italy for the Digital Decade](https://digital-strategy.ec.europa.eu/it/factpages/italy-2025-digital-decade-country-report) is a summary report by the Commission: it collects indicators on businesses, skills and specialists, and reads them against the European average. Finally, there is another Eurostat series, on [individuals](https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Use_of_artificial_intelligence_by_individuals), which covers individual use of generative AI and not use by companies.

| Source | Who it measures | 2025 figure |
|---|---|---|
| [ISTAT](https://www.istat.it/comunicato-stampa/imprese-e-ict-anno-2025/) | Italian enterprises with at least 10 employees | 16.4% use at least one AI technology |
| [Eurostat, companies](https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Use_of_artificial_intelligence_in_enterprises) | EU companies | 19.95% use AI technologies |
| [Eurostat, individuals](https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Use_of_artificial_intelligence_by_individuals) | People aged 16-74 in the EU | 33% have used generative AI tools |

The 19.95% from [Eurostat](https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Use_of_artificial_intelligence_in_enterprises) and the 16.4% from [ISTAT](https://www.istat.it/comunicato-stampa/imprese-e-ict-anno-2025/) should not be pitted against each other to say that Italy is "below" or "above": scopes and definitions do not coincide. The third figure, moreover, concerns citizens and not companies.

## How is AI use growing in Italian companies?

It is growing, from a low base. [ISTAT](https://www.istat.it/comunicato-stampa/imprese-e-ict-anno-2025/) reports that in 2025 16.4% of companies with at least 10 employees use at least one AI technology, against 8.2% in 2024 and 5.0% in 2023. The same release notes that 83.6% of companies do not adopt any AI technology, and reads this as a level of penetration that is still very limited.

A warning about dates. The [Commission's Report](https://digital-strategy.ec.europa.eu/it/factpages/italy-2025-digital-decade-country-report) cites 8.2% of Italian companies having adopted AI, without giving the year in the passage we read, while ISTAT attributes that 8.2% to 2024. If you quote it, you must date it, otherwise it looks like a current figure.

The same Report notes that most Italian SMEs, 70.2%, have reached at least a basic level of digital intensity ([European Commission](https://digital-strategy.ec.europa.eu/it/factpages/italy-2025-digital-decade-country-report)). The two things fit together: a company can be digitalized at an elementary level and still not use AI.

## How wide is the gap between SMEs and large companies?

It is wide and it is widening. According to [ISTAT](https://www.istat.it/comunicato-stampa/imprese-e-ict-anno-2025/), large companies go from 32.5% in 2024 to 53.1% in 2025, while use among SMEs doubles, from 7.7% to 15.7%. The release notes that for most indicators size gaps are narrowing, whereas for AI the opposite is happening.

The difference in intensity of use between large companies and SMEs goes from about 20 percentage points in 2023 to 25 in 2024, up to 37 in 2025, again according to [ISTAT](https://www.istat.it/comunicato-stampa/imprese-e-ict-anno-2025/). Also, [Eurostat](https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Use_of_artificial_intelligence_in_enterprises) sees a scale by size in the EU: in 2025, 17% of small companies, 30.36% of medium ones and 55.03% of large ones use AI.

| Measure | Value | Source |
|---|---|---|
| Italian SMEs using AI in 2025 | 15,7% | [ISTAT](https://www.istat.it/comunicato-stampa/imprese-e-ict-anno-2025/) |
| Italian large companies using AI in 2025 | 53,1% | [ISTAT](https://www.istat.it/comunicato-stampa/imprese-e-ict-anno-2025/) |
| Small companies in the EU | 17% | [Eurostat](https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Use_of_artificial_intelligence_in_enterprises) |
| Medium companies in the EU | 30,36% | [Eurostat](https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Use_of_artificial_intelligence_in_enterprises) |
| Large companies in the EU | 55,03% | [Eurostat](https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Use_of_artificial_intelligence_in_enterprises) |

The gap is wider where specialist skills or complex structures are needed. [ISTAT](https://www.istat.it/comunicato-stampa/imprese-e-ict-anno-2025/) cites data analysis, used by 41.9% of SMEs and 83.6% of large companies, and ERP and CRM management software, where SMEs are at 48.8% and 21.1% and large companies at 85.9% and 56.5%. For an SME the message is practical: AI works best where data and processes are already in order.

## Why do many companies not adopt AI?

Skills come first. [ISTAT](https://www.istat.it/comunicato-stampa/imprese-e-ict-anno-2025/) writes that lack of adequate skills holds back adoption in almost 60% of the companies that evaluated AI investments but then did not make them. Among companies that do not use AI, 11.5% have considered using it, against 4.6% in 2023.

The barriers reported by these companies, again according to [ISTAT](https://www.istat.it/comunicato-stampa/imprese-e-ict-anno-2025/), are six and should be read together, because they are not mutually exclusive.

| Barrier | Share | Source |
|---|---|---|
| Lack of skills | 58,6% | [ISTAT](https://www.istat.it/comunicato-stampa/imprese-e-ict-anno-2025/) |
| Lack of legal clarity | 47,3% | [ISTAT](https://www.istat.it/comunicato-stampa/imprese-e-ict-anno-2025/) |
| Data unavailable or of poor quality | 45,2% | [ISTAT](https://www.istat.it/comunicato-stampa/imprese-e-ict-anno-2025/) |
| Privacy and data protection | 43,2% | [ISTAT](https://www.istat.it/comunicato-stampa/imprese-e-ict-anno-2025/) |
| High costs | 43,0% | [ISTAT](https://www.istat.it/comunicato-stampa/imprese-e-ict-anno-2025/) |
| Ethical considerations | 25,7% | [ISTAT](https://www.istat.it/comunicato-stampa/imprese-e-ict-anno-2025/) |

In the European picture the order is similar. According to [Eurostat](https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Use_of_artificial_intelligence_in_enterprises), among EU companies that considered AI in 2025 the most common reason for not using it is lack of relevant skills, at 70.89%, followed by unclear legal consequences, at 52.52%, and data protection concerns, at 48.83%. In the same picture, 14.21% of EU companies that used no AI technology had nonetheless considered it.

These numbers do not say that training would solve everything, nor that companies are not already training some people. They say that companies themselves name skills as the first brake. That is a good reason to treat training as part of the project and not as an extra.

## What about people's digital skills?

The context does not help. The [Commission's Report](https://digital-strategy.ec.europa.eu/it/factpages/italy-2025-digital-decade-country-report) indicates that only 45.8% of the Italian population has basic digital skills, with gaps affecting people with lower levels of education but also young people. Information and communication technology specialists account for 4% of total employment in 2024, below the EU average.

On the citizens' side, [Eurostat](https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Use_of_artificial_intelligence_by_individuals) reports that in 2025 33% of people aged 16-74 in the EU used generative AI tools in the three months before the survey. The lowest levels of use are recorded in Romania (18%), Italy (20%) and Bulgaria (23%). The most common reason for not using it is not feeling the need, at 64%.

It is a figure about people, not companies, but it is a reminder of a simple thing: many people working in an SME have often not yet tried these tools even outside the office. Starting from zero is the norm, not the exception.

## What does this mean for an SME? Two examples

Here are two typical situations, described as hypothetical examples and not as client cases.

**An accounting firm with a few employees.** The owner reads that the gap with large companies is widening and thinks they must buy a tool right away. The data, however, say that the most cited brake is skills, not software. It is better to choose a single task, for example drafting replies to recurring client requests, and define who checks the result before it is sent. Training is the rule of use for that task, not a generic course.

**A craft business with a three-person sales office.** It uses management software but the quotation data are scattered across different files. [ISTAT](https://www.istat.it/comunicato-stampa/imprese-e-ict-anno-2025/) lists among the barriers also unavailable or poor-quality data, at 45.2%. Before talking about AI, it makes sense to put the data in order and write down who can enter what. Training helps people understand why that order matters for the tool too.

In both cases the point is not to chase large companies, but to remove the concrete obstacle in front of you. To learn more about setting up the first steps you can read the [practical guide to AI for SMEs](https://giallostudio.it/en/blog/ai-for-small-business-practical-guide) and the path to [train your team on AI](https://giallostudio.it/en/blog/ai-team-training).

## Which mistakes should you avoid when reading these data?

Five mistakes are the most frequent when commenting on AI statistics.

### Mistaking adoption for training

It happens because the two topics are close and a newspaper headline merges them. You avoid it by always saying what the source measures: ISTAT measures the use of AI technologies, not courses taken.

### Comparing numbers with different scopes

The 19.95% from [Eurostat](https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Use_of_artificial_intelligence_in_enterprises) and the 16.4% from [ISTAT](https://www.istat.it/comunicato-stampa/imprese-e-ict-anno-2025/) look comparable but are not. You avoid it by placing them side by side only as two distinct measures, each with its own scope.

### Citing a figure without a year

The 8.2% appears in several sources with different dating: [ISTAT](https://www.istat.it/comunicato-stampa/imprese-e-ict-anno-2025/) attributes it to 2024, while the [Commission's Report](https://digital-strategy.ec.europa.eu/it/factpages/italy-2025-digital-decade-country-report) gives no year in the passage we read. You avoid it by always writing the year next to the figure and the name of who publishes it.

### Turning a barrier into a recipe

If lack of skills is the first brake, it does not follow that a course will eliminate it. You avoid it by treating the figure as a clue and checking with your own people what is really missing.

### Ignoring the constraints of rules and data

Companies also cite legal clarity and privacy. Whoever trains only on the use of the tools neglects these points. You avoid it by including a reading of [GDPR and AI in business](https://giallostudio.it/en/blog/gdpr-e-ai-aziende) and of [AI Act obligations for an SME](https://giallostudio.it/en/blog/ai-act-sme-obligations) in the training path.

## How can you use these data in your company?

These data are useful for deciding where to start, provided you turn them into a few actions.

<ol class="steps">
<li>Write the question you want to solve in one sentence, for example which repetitive task you want to lighten.</li>
<li>Check which skills are really missing, by asking the people who perform that task, not whoever buys the tool.</li>
<li>Define a short rule of use: which data can be entered, who checks the result, what is not done. A starting outline is in the guide to the <a href="https://giallostudio.it/en/blog/company-ai-use-policy">company policy on AI use</a>.</li>
<li>Start a limited trial, following the approach of a <a href="https://giallostudio.it/en/blog/ai-pilot-project">pilot project</a>, with a small group.</li>
<li>Decide in advance what you will measure, drawing on the <a href="https://giallostudio.it/en/blog/ai-project-kpis">indicators of an AI project</a>, and revise the training based on what emerges.</li>
</ol>

For the organizational side of ongoing training, the reference resource is the article on [AI for internal training](https://giallostudio.it/en/blog/ai-for-employee-training).

## Where to start, in practice

Official statistics show growing adoption, a widening size gap and skills named as the first brake. They do not show how many companies train their staff, and this should be stated clearly. For a small company, the sensible step is to train on one specific task, with written rules and human oversight, and only then to widen the scope.

If you want to work out which task to choose and how to support your people, you can start with an [AI consulting session](https://giallostudio.it/en/ai-consulting): we reason through the company's processes, with no commitment and no promises of results.

## Frequently asked questions

### How many companies train their staff on AI?

The official statistics we read do not measure it. ISTAT, Eurostat and the European Commission report AI adoption and barriers, not courses taken. For this reason it is not correct to attribute a percentage on training to these sources.

### Is lack of skills really the main barrier?

Yes, in the sources we read it is the first barrier cited by companies that do not use AI. Both ISTAT for Italy and Eurostat for the European Union point to it. It remains a self-declaration by companies, not a measure of actual skills.

### Can I compare the ISTAT figure with the Eurostat one?

With caution. The ISTAT figure covers Italian companies with at least 10 employees, while the Eurostat one covers companies in the European Union. The scopes do not coincide, so it is best to present them as two distinct measures and not as a ranking.

### Is the gap between SMEs and large companies narrowing?

Not for AI, according to ISTAT. The release observes that for many digital indicators size gaps are shrinking, while for AI the difference in intensity of use between large companies and SMEs increases from 2023 to 2025.

### Where should an SME that wants to train its staff start?

From a specific task and a written rule of use. Identify the people involved, clarify which data can be entered and who checks the result. Only after a limited trial does it make sense to extend training to the rest of the company.

### Does an adoption figure tell us whether people are prepared?

No. It says that the company uses at least one AI technology, not how well it does so or whether people have been trained. To find out, you need to ask directly the people who work with those tools.
