Artificial intelligence is already inside your company. It probably did not ask you first.

Two published figures, placed side by side, tell an uncomfortable story.

According to national and European statistics, 38.7% of people in Portugal used artificial intelligence tools in 2025. That is above the European average of 32.7%, and ahead of France and Germany.

Over the same period, only 11.5% of Portuguese companies with ten or more employees were using these technologies. The European Union average sits closer to 20%.

People adopted. The companies those people work for did not.

This does not mean artificial intelligence is absent from your business. It almost certainly means the opposite: it is already there, your staff brought it in, and it never crossed your desk.

What that looks like in practice

Someone on your team pasted a client proposal into a free tool to rewrite it faster. Someone translated a contract. Someone asked for help summarising a contact list.

Nobody did anything malicious. They got the work done faster, which is exactly what you ask of them.

The problem is that information left the company with no record, no agreement with whoever received it, and no idea where it ended up. If a client asks tomorrow where their data is, you have no answer.

And note this: it is not an IT problem. It is a management problem, and it is solved with a decision, not with software.

Why companies fell behind

Businesses were also asked why they do not use these technologies. The answers are revealing:

Reason given Companies
Lack of in-house knowledge 74.4%
Costs appear too high 60.4%
Unclear legal consequences 57.3%
Data protection and privacy concerns 55.4%

Notice that none of these answers says "it is not useful for my business". They say "I do not know where to start", "I think it is expensive" and "I am worried about what could go wrong".

Three legitimate fears. All three solvable.

Cost is the easiest one to dismantle

Six in ten companies think it is expensive. Yet the same data shows that 67.4% of those already using it adopted through off-the-shelf commercial software, and that this was the only adoption model that grew in 2025.

In other words, most of them did not commission anything. They started using tools that already existed, many of them inside software the company was already paying for.

The usual mistake is to imagine a large project, budget tens of thousands of euros, take fright and do nothing. The approach that works runs the other way: pick one tedious, repetitive task, solve that, measure the time saved, and only then decide whether to go further.

There is already a date in the calendar

From 2 August 2026, the transparency obligations of the European artificial intelligence regulation start to apply.

In plain terms: if you have an automated assistant answering clients on your website, that person must understand they are speaking to a machine rather than to a member of staff. And artificially generated content has to be identifiable as such.

The heavier obligations, those covering uses classified as high risk, were pushed back to December 2027. That gave many people the mistaken impression that nothing needs doing this year. The transparency piece still stands, and it is precisely the part that affects anyone running automated support on a website.

As always with legal matters, confirm your specific situation with your own advisers. But do not assume this is a problem for 2028.

Where a decision maker starts

Five steps, none of them technical:

  1. Ask what is already being used. Without reprimanding anyone, or you will not get the truth. You will find more than you expect.
  2. Write one page of rules. What may and may not be shared with outside tools. One page is enough, and it beats a manual nobody reads.
  3. Pick one concrete task. Quote requests, meeting summaries, product descriptions. Just one.
  4. Buy before you build. Commissioning something bespoke only makes sense once you know exactly what you want, and you only know that after trying.
  5. Check what reaches the client. Nothing generated automatically should leave the company without a human reading it first.

And when not to use it

A note you will rarely hear from anyone selling this.

If a task carries legal responsibility, if an error costs money or reputation, or if nobody in the company is competent to check the output, then the right answer is not to use it. Automating a broken process only means doing it badly, faster.

The intelligent decision is not always to adopt. It is knowing where to adopt.

How Digito helps

The gap between 38.7% and 11.5% is not a technology problem. It is the result of nobody stopping to make a decision.

That is where we come in. We start by finding out what is already being used in your company and by whom, help you write the internal rules that do not yet exist, choose the first task to tackle with you, and make sure what your clients see meets the transparency obligations arriving in August.

No six-month projects and no promises of transformation. One decision at a time, with a measured result.

If you recognised your company anywhere in this article, let's talk.

Sources