Portugal’s firms use less AI than Europe’s — but its workers use more
Only 11.5% of Portuguese firms use AI, versus 20% in the EU — yet 38.7% of Portuguese people already use AI tools. What explains the paradox?
The paradox in numbers
In 2025, only 11.5% of Portuguese enterprises with 10 or more persons employed used artificial intelligence technologies. The figure comes from Statistics Portugal (INE) in its highlight of 21 November 2025, based on the annual enterprise survey carried out between February and June that year, and it is 2.9 percentage points higher than in 2024. That is real growth — but slow growth, especially against the European average: according to Eurostat’s 2026 statistical report, 20.0% of EU enterprises with 10 or more persons employed used AI in 2025, up from 13.5% in 2024 and 8.1% in 2023. On the other side of the same picture, Portuguese people, as individuals, are ahead. INE’s 2025 household survey on information and communication technology use found that 38.7% of residents in Portugal aged 16 to 74 had used AI tools in the three months before the interview, mostly for personal use. The comparable European figure, published by Eurostat in the same 2026 report, is 32.7% for the use of generative AI by individuals. Among young Portuguese people aged 16 to 24, the share rises to 76.5%; among students, to 81.5%. In other words: the country where workers have already tried AI is the same country where firms adopt it slowly. There is no contradiction in that, but there is an uncomfortable question: if people already know how to use these tools at home, why does the workplace lag behind? The answer, as we will see, has less to do with the technology and more to do with firm size, the lack of knowledge inside small firms, and the cost of getting started.
The gap that opened in two years
The most worrying part is not the 2025 snapshot, but the film of the last two years. In 2023, Portugal was practically level with the European Union: 7.9% of Portuguese enterprises used AI, against 8.1% for the EU average — a gap of just 0.2 percentage points. In 2025, the gap is 8.5 points. Portugal grew by 3.6 points in two years; the European Union grew by 11.9 points over the same period. Europe accelerated, and Portugal did not keep pace. That does not mean Portuguese firms are standing still. It means the European pace has shifted up a gear: between 2024 and 2025, the EU average jumped 6.5 points in a single year, the largest jump in the series. When the rest of Europe starts using AI in administrative processes, marketing and text analysis at the same time, a Portuguese firm that postpones the decision does not stay in the same place — it falls further behind competitors, suppliers and customers who already work with these tools. There is one sector where Portugal is closer: in information and communication, which includes technology firms, 52.8% of Portuguese enterprises use AI, according to the 2026 Radar on AI use and supply in Portugal by Via Analítica, against 62.5% for the EU average. The gap exists even in the most digital sector, but it is smaller. The Portuguese problem, then, is not a lack of technology firms. It is the rest of the economy: services, retail, light industry and firms that do not live from technology but still need it to stay competitive.
It is an SME gap
When the national figure is broken down by size, the story becomes clear. In 2025, only 9.4% of Portuguese enterprises with 10 to 49 persons employed used AI. For enterprises with 50 to 249 persons employed, the share was 18.2%. For large enterprises, with 250 or more persons employed, it reached 49.1%. In other words: almost half of large firms already use AI, and fewer than one in ten small firms do the same. The national average of 11.5% therefore hides two very different realities. And there is a detail that makes the picture even more demanding: INE’s enterprise survey only covers enterprises with 10 or more persons employed. Micro-enterprises, with fewer than 10 people — the large majority of Portugal’s business fabric — do not even enter into that 11.5%. If they did, the national share would almost certainly be lower. The paradox, then, is even larger than the headline figure suggests. Why do small firms fall behind? The data on barriers points, above all, to knowledge. According to Lynx Agency’s summary of the INE survey, 74.4% of enterprises that considered AI and did not adopt it cited a lack of relevant knowledge as a reason. Cost comes next: around 60% of Portuguese enterprises in that situation cite high costs, according to Via Analítica’s Radar — a figure that INE’s highlight does not publish in its barriers section, so we present it here as a Radar figure, not as an official INE number. For the EU average, Eurostat records a similar pattern with different weights: 70.3% of enterprises that considered AI without adopting it cite a lack of expertise, and 38.4% cite costs. In Portugal, cost weighs more heavily than in Europe; in both cases, knowledge weighs more heavily than cost. This makes sense when you think about how a small firm works. In a 15-person business, there is no technology department and nobody with spare time to test tools. The person deciding is often the same person serving customers, dealing with suppliers and closing the month. Learning a new tool is not just paying a subscription: it means stopping to understand what it does, whether the firm’s data stays safe, and whether the result repays the time spent. Without that knowledge inside the firm, postponing is the most rational decision — and it is exactly that rational decision, repeated thousands of times, that opens the gap.
| Enterprise size | Use AI in 2025 |
|---|---|
| 10 to 49 persons employed | 9,4% |
| 50 to 249 persons employed | 18,2% |
| 250 or more persons employed | 49,1% |
What firms that already use AI do with it
Looking at what firms that use AI actually do with it helps clear up a misconception: AI in Portuguese firms is not, in most cases, advanced robotics or research projects. It is office work with assistance. Among firms that use AI, the most common application, according to INE, is written language analysis, used by 59.4%. Next come the generation of images, video or sound, at 50.9%, and the generation of text or code, at 45.6%. The areas where these tools enter confirm the same picture: the organisation of business administration processes or management leads, at 40.8%, followed by marketing and sales, at 36.9%. In other words, AI’s way into firms is repetitive, text-based work: summarising documents, drafting replies, organising information, producing content. It is exactly the kind of task a small firm recognises in its daily routine — and it is also the kind of task where a mistake is spotted quickly and fixed without drama. That is good news for anyone starting out, for two reasons. First: no big project is needed to get in. The firm that already uses AI did not generally start by reinventing the business; it started with one concrete task. Second: the most common uses are also the easiest to measure. If a tool starts drafting the first versions of replies to customers, the time saved per week can be measured. If it generates product descriptions, the number of descriptions per day can be measured. Measuring is what separates an experiment from an investment — and it is what is missing from many postponed decisions, which are postponed precisely because nobody knows what would be gained.
Leiria: the response starts with training
If the biggest barrier is a lack of knowledge, the most direct response is training — and that is exactly where the Portuguese government placed its first step. On 20 October 2026, a pilot programme to train workers from micro, small and medium-sized enterprises in artificial intelligence began in Leiria, promoted by IAPMEI with the support of NERLEI, the region’s business association, and delivered by Shaken Academy through the «AI Dive for Business» cycle. The first phase covers around 50 workers; the pilot will have two editions, with 100 participants in total. The programme sits within the PME Digital+ project and is, according to the Ministry of Economy’s statement, the first step of a wider programme that aims to train around 100,000 workers from micro-enterprises and SMEs across the country over three years. The calendar gives the topic immediate relevance: according to The Portugal Post’s coverage, the first edition continues on 3, 4 and 5 November at the NERLEI premises in Leiria — dates that the official statement does not detail, so we cite them here with that attribution. The content, by contrast, is in the statement: equipping participants for the practical use of generative AI tools, focused on their application to the concrete challenges of their own firms, not on generic theory. It is worth keeping the proportions. One hundred participants in a pilot do not close a national gap — and the government itself presents it as a pilot, not a solution. What the pilot tests is the model: short, practical training, tied to real problems of small firms, with materials and use cases the programme aims to disseminate in open access. If the model works in Leiria, the question becomes scale: how to take the same format to thousands of firms without turning it into a generic course. For an SME outside the Leiria region, the practical lesson is different: there is no need to wait for the national programme to start training one person — and the next section suggests how.
What an SME can do on Monday morning
None of this requires a six-month strategic plan. It requires a Monday morning and one small, well-measured decision. The method the data itself suggests has four steps. First, choose a process, not a tool. The ideal process is repetitive, text-based and easy to count: answering customers’ frequent questions, writing product descriptions, summarising long emails, preparing meeting notes. If you cannot count how many times a week the process happens, it is not ready to be automated yet. Second, measure before changing. For one week, note how much time the process consumes: how many minutes a day, done by whom. Without this baseline, any «saving» announced afterwards is talk. Third, try a single tool, for 30 days, with a single person responsible. A tool nobody masters is worse than no tool: the goal of the month is not to impress, it is to learn whether the process gets faster without losing quality. Fourth, and most important: keep sensitive decisions under human control. Hiring, firing, granting credit, deciding about customer data or answering a delicate complaint are not tasks to delegate unsupervised. AI that enters through repetitive tasks should free up time for decisions — not make them. At the end of the 30 days, the question is simple and numerical: does the process take less time, at the same quality or better? If yes, keep it and move to the next process. If no, switch it off without drama. That is how, process by process, a 15-person firm closes a gap that no report will close for it.
ZapDesk
One concrete first process: AI reception on WhatsApp, in a live demo with no signup.
The honest reading: what these numbers do not prove
Before closing, it is worth stating what this article does not say. First, the 11.5% does not describe all Portuguese enterprises: it describes enterprises with 10 or more persons employed that answered INE’s survey. Micro-enterprises are outside the picture, and any headline that generalises the figure to «Portuguese firms» without that qualification is oversimplifying. Second, not every 51% is the same as the 11.5%. In recent weeks, news circulated of a Strand Partners study for AWS according to which 51% of Portuguese firms have already adopted AI. The figure is not wrong — it measures something else: a different sample, a different definition of «using AI» and different selection criteria. Comparing that 51% with INE’s 11.5%, as if they were the same race, is a reading error. The AWS study’s value lies elsewhere: in the distance between having access to a tool and genuinely integrating it into processes — only a minority of firms that use AI have reached an advanced level of integration. Third, not every figure from Via Analítica’s Radar appears in this article. The Radar, which could not be located in an open version, is cited in the trade press with its own counts — such as 486 AI startups or 2,847 approved projects under the «IA nas PME» line — which we could not verify from a primary source and therefore do not use as facts. The same applies to the Portuguese cost figure: we present it as a Radar figure, not an official one, because INE’s highlight does not publish that barriers section. The paradox, even so, survives all these cautions, because it rests on the two official pillars: INE and Eurostat. Portuguese firms adopt slowly, and Portuguese people use quickly. The good news hidden in the paradox is that the country does not need to convince people that AI is good for something — they have already discovered that on their own. It needs to help firms, above all small ones, take the first measured step. And that step is not taken with a report. It is taken on a Monday morning, with one process, one measurement and one person responsible.
Frequently Asked Questions
Why do Portuguese people use more AI than the EU average while Portuguese firms use less?
Because they are two different measurements. Individual use (38.7%) counts people who have tried AI tools, often free ones and for personal use. Enterprise use (11.5%) counts firms that have integrated an AI technology into their processes, which requires a decision, training and budget. Trying something at home is free and immediate; adopting it in a firm requires someone who knows how, time to learn and money to pay.
My firm has fewer than 10 workers. Do these figures say anything to me?
Yes, but by absence. INE’s enterprise survey only covers enterprises with 10 or more persons employed, so micro-enterprises — the majority of the business fabric — do not appear in the 11.5%. If your firm is micro, the official picture does not measure you; what does measure you, indirectly, is the main barrier: in small firms, lack of knowledge (74.4%) weighs more than cost. Starting with one small, measured process is the advice the data supports.
What holds adoption back more: cost or lack of knowledge?
Lack of knowledge. Among enterprises that considered AI and did not adopt it, 74.4% cite a lack of relevant knowledge, according to Lynx Agency’s summary of the INE survey. Cost comes next: around 60% in Portugal, according to Via Analítica’s Radar, and 38.4% for the EU average, according to Eurostat. The order matters: a cheap tool nobody knows how to use stays idle.
What is the Leiria pilot, and how can my firm take part?
It is a pilot programme to train workers from micro, small and medium-sized enterprises in AI, promoted by IAPMEI with NERLEI and Shaken Academy, which began on 20 October 2026 in Leiria with around 50 workers in the first phase and two editions totalling 100 participants. It is a pilot, not an open enrolment programme: the aim is to test a practical training model that would later be extended to around 100,000 workers over three years. Firms outside the region should follow IAPMEI’s notices rather than count on an immediate place.
Where should a firm start without spending much?
With a repetitive, text-based process that already consumes time every week: answering frequent questions, writing product descriptions or summarising documents. Measure how much time the process takes for one week, try a single tool for 30 days with one person responsible, and decide on the numbers: if the process takes less time at the same quality, keep it; if not, switch it off. Sensitive decisions — hiring, credit, customer data — always stay under human control.