AI in Portuguese SMEs: measure value, not usage
Author: Lúcia Babo - Economist, VOICE Leadership trainer at Nova SBE, Member of the Ethics Forum at Católica Porto School of Business and Economics & Ambassador of the Agrifood VET Ambassador Network.
Tech News Portugal - Portugal is a country of SMEs: they make up 99.9% of all companies. And SMEs are where artificial intelligence has made the least progress. According to Statistics Portugal (INE), in 2025, among companies with 10 or more employees, AI was used by 49.1% of large companies, 18.2% of medium-sized companies and only 9.4% of small ones. Overall, 11.5% of Portuguese companies used AI, compared with around 20% across the European Union. The gap is a concern, but the bigger risk is not adopting slowly. It is adopting because everyone else is measuring how many people use the tools rather than how much value they create. In an SME, where every euro counts, that difference decides whether AI improves the business or simply adds to the bill:
A recent McKinsey study, The AI value curve, published in October 2026, offers a simple way to do maths. It compares, task by task, what the work costs today with what it would cost with AI. The second figure includes everything: the technology and the time people spend reviewing, validating and correcting the output. AI only creates value when that total is lower than today’s cost, at an acceptable level of quality.
In Portugal, the numbers are different:
The study makes a point that matters especially for Portuguese companies: the economics of AI are local. Technology costs much the same in Lisbon, Madrid or New York, and is often paid in dollars. Wages are not the same. In 2025, the average annual salary in Portugal was €26,202, about 63% of the EU average of €41,464, according to Eurostat.
This means some tasks where AI pays off in the United States or Northern Europe do not pay off in a Portuguese SME. Copying success stories from abroad is not enough. The numbers have to be run with local costs.
Where the value lies in a Portuguese SME:
The study shows that the largest gains per hour are not in repetitive administrative work, but in more skilled work: management, finance, legal and analysis. In a Portuguese SME, much of that work is concentrated in one person: the owner. The owner prepares the meeting with the bank, reviews proposals, watches the cash flow, answers foreign clients and still finds time for the application for public funding.
That is where AI can create the most value: preparing reports for management and for the bank, drafting proposals and content for export markets in languages the team does not master, and preparing applications for public incentives. The time freed goes to what cannot be delegated: deciding, negotiating, winning clients.
In contrast, in high-volume, low-wage tasks such as entering documents, AI can end up costing more than the current way of working, once review and correction time is counted.
A rule any manager can apply:
The study quotes a practical rule: an AI agent creates value when its probability of success is greater than the ratio between the time needed to check its work and the time needed to do the task directly. If reviewing a proposal takes a third of the time it takes to write it, the AI has to get it right more than a third of the time and still cover its own cost. With lower wages, that margin narrows.
The cost of AI also changes:
According to the study, a 30% rise in the effective cost of AI would cut the potential global value from $2.52 trillion to $1.67 trillion, roughly a third. For a Portuguese SME that often pays for licences in dollars, currency risk comes on top. A business case that only works if prices keep falling is not a good business case.
Five steps to get started:
1 - Map the tasks in your five largest cost centres: how long they take, who does them and what that hour really costs, including social security contributions and the 14 monthly salaries paid in Portugal.
2 - Run the numbers task by task, including review and correction time, not just the licence.
3 - Start with the time of those who decide. In most SMEs, that is where the greatest return lies.
4 - Test the case with costs 30% higher and 30% lower and go ahead only if it holds in both scenarios.
5 - Decide in advance where the freed time will go. Without that plan, productivity disappears into idle time.
A question of governance:
There is one more dimension the numbers do not show. Who decides where AI is used? Who validates the results? Who keeps track of the cost? In family businesses, where decisions tend to be concentrated, these questions are also an opportunity to clarify roles, delegate with clear rules and free leaders for what is truly strategic.
AI is not an end in itself. It is a factor of production, with a cost and a return, and should be managed as such. Portuguese SMEs do not need to be the ones that use AI the most. They need to be the ones that know where it pays off.
Author: Lúcia Babo, Economist at Nova SBE
Sources:
Van der Veken, L., Durth, S., Catlin, T., Alatovic, T. and Isherwood, T. (2026). The AI value curve: The economics of applying AI to existing work. QuantumBlack, AI by McKinsey, October 2026.
INE, use of AI by companies in 2025, reported by ECO, 21 November 2025.
Eurostat, use of AI by EU companies in 2025, reported by ECO, 2 April 2026.
Eurostat, average annual salary in 2025, reported by Forbes Portugal, 6 November 2026.
INE, share of SMEs in the total number of companies (2023), reported by APCMC.
