In 2025, only 11.5% of Portuguese companies with 10 or more people were using artificial intelligence, according to INE, Statistics Portugal. That is up 2.9 percentage points on 2024, but still well short of the European Union average of around 20% (Eurostat). And the gap by company size is striking: 49.1% of large companies use AI, against 18.2% of medium sized firms and just 9.4% of small ones.
Another figure from the same survey strikes me as even more telling. Between May and August 2025, 38.7% of people aged 16 to 74 used AI tools. In other words, many people working in SMEs already use AI in their personal lives. In most cases, the company they work for does not.
I often meet business owners who see AI in one of two ways: as a fad that will pass, or as a large, expensive project for when there is time. Both lead to the same outcome, which is doing nothing. As an engineer, I suggest a third way: treat AI like any other productivity tool, with a concrete problem, a metric and a deadline.
Why SMEs fall behind
The gap between 49.1% of large companies and 9.4% of small ones is not explained by budget alone. In my experience, there are three more common reasons:
- No defined processes: you cannot automate what has never been described. If everyone does the same task differently, AI has nothing to work with.
- Nobody is responsible: in an SME, technology is often left to whoever is most handy with it, in their spare time. Without an owner and an objective, no experiment gets past the curiosity stage.
- Starting at the wrong end: choosing the tool first and then looking for a problem for it to solve.
Start with the problem, not the tool
The right question is not "which AI tool should we use?" It is "where does our team lose the most hours on repetitive work?" That matters particularly in Portugal. According to PORDATA, using Eurostat data, we work an average of 39.7 hours a week (2025), the fifth highest in the EU, while labour productivity was 47.7 thousand euros per worker in 2024, the 19th lowest in the Union. We work long hours and produce relatively little per hour. Applied well, AI targets exactly that imbalance.
In SMEs, the starting points with the best ratio of effort to results tend to be these:
- Replying to enquiries and quote requests with drafts the team then reviews.
- Summarising meetings and turning notes into task lists with named owners.
- Preparing sales proposals from templates and the customer's history.
- Sorting and routing emails from customers and suppliers.
- Extracting data from invoices and documents for the accounts.
- Producing first drafts of content for the website and social channels.
The website as an early signal
Before thinking about advanced AI, it is worth looking at the digital basics. According to INE, in 2025, 62.9% of companies had a website, but only 13.5% of those allowed customers to track orders online. That tells me many companies have a digital presence, but few have digital processes. A website that is only a shop window does not feed any system.
AI works best when there is organised data behind it: an up to date customer record, a history of proposals, a defined ordering process. If that data does not exist, your first AI project is really an organisation project.
A 90 day pilot in five steps
This is the method I recommend to any SME that wants to start without wasting time or money:
- Step 1, pick one task: just one, repetitive, which currently takes a measurable number of hours.
- Step 2, measure the baseline: how many hours a week it takes, how many errors it produces, how long customers wait for a response.
- Step 3, name an owner: someone with time set aside in their diary, not in their spare moments.
- Step 4, test for 30 days: with an accessible tool, clear rules on which data can and cannot be used, and human review of everything it produces.
- Step 5, decide on the numbers: after 90 days, compare against the baseline. If you saved hours and kept the quality, write the process down and move on to the next task. If not, drop it and learn from it.
A hypothetical example: imagine a 15 person company whose sales team spends 10 hours a week preparing proposals. If a well run pilot halves that, it frees 5 hours a week, more than 200 hours a year, to contact more leads and follow up more proposals. That links directly to the ActionCOACH 5 Ways: more leads worked and a better conversion rate.
Ground rules before you start
AI does not replace judgement. Before the first pilot, management should set out three things in writing: which information must never go into external tools (customers' personal data, sensitive financial information), who reviews the output before it reaches a customer, and how tests are recorded. They are simple rules, but they prevent the mistakes that make many owners give up at the first attempt.
Your action for this week
This week, ask each person on your team which repetitive task takes up most of their time, and how many hours a week they spend on it. Put the answers into a list ranked by hours. The top line of that list is your first AI pilot.
Moving from the 9.4% of small companies into the group already using AI does not require a large investment. It requires method and accountability. If you would like to structure that first 90 day cycle, for example through our 90 Day Planning Workshop, the ActionCOACH Porto team is happy to talk.