In 2025, people in Portugal worked an average of 39.7 hours a week, against 37 in the European Union as a whole. In 2024, each worker generated around 47.7 thousand euros of wealth, which puts us 19th in the EU, according to PORDATA. We work longer hours and produce less per person. It is in this context that many business owners ask me whether artificial intelligence is the answer. The honest answer, based on the serious studies we have, is: it can be, but not in the way most people are using it.
What the studies say, without the hype
As an engineer, I am wary of promises without data. Fortunately, there are already rigorous studies, with control groups, and they are worth knowing.
- Customer support: Brynjolfsson, Li and Raymond followed 5,179 support agents. With an AI assistant, they resolved on average 14% more issues per hour. Customers were happier and staff turnover fell.
- Professional writing: Noy and Zhang, in a study published in Science in 2023 with 453 qualified professionals, measured writing tasks done around 40% faster, with quality around 18% higher according to independent evaluators.
- Consulting: Dell'Acqua and colleagues at Harvard Business School tested 758 BCG consultants. On tasks suited to AI, those using it completed 12.2% more tasks, 25.1% more quickly and with over 40% higher quality.
So far, it all sounds wonderful. But the same Harvard study delivered the most important lesson.
The jagged frontier: where AI helps and where it hurts
The researchers called it the "jagged technological frontier". AI is very good at some tasks and surprisingly poor at others, and the line between them is not obvious. On a task designed to sit outside that frontier, consultants using AI were 19 percentage points less likely to reach the correct answer than those working alone. They trusted an answer that was convincing, but wrong.
The 2025 METR study points the same way. Sixteen experienced developers, working on projects they knew well, completed 246 tasks. Beforehand, they expected AI to make them 24% faster. Afterwards, they believed they had been 20% faster. In reality, with AI they took 19% longer. Perception and measurement said opposite things.
For an SME, the conclusion is simple: it is not enough to "give the team AI". You need to know which tasks it helps with, and to measure rather than trust a feeling.
The least experienced gain the most
There is a pattern that keeps repeating. In customer support, new or less skilled agents improved by 34%, while the most experienced barely gained at all. At BCG, below-average consultants improved by 43% and above-average ones by 17%. In the writing study, those who started weakest benefited most.
This is excellent news for an SME with 5 to 50 people. AI can act as a way of passing the knowledge of your best people on to those just arriving. A new salesperson can prepare proposals closer to the level of your best seller. A newly hired administrator can answer customers with the quality of someone who has been there ten years. The learning curve gets shorter.
Why the gains do not show up in the P&L
This is the point I see fail most often. Suppose, as a hypothetical example, that your team saves 30 minutes a day with AI. If that half hour dissolves into more emails, more meetings or longer breaks, your bottom line does not change by a single cent. Time saved is only worth money when it is deliberately redirected to something that generates revenue or cuts costs.
Productivity gains show up in hours. Business gains show up in sales, margin and customer retention. Between the two sits a management decision: what do we do with the time freed up? Without that decision, AI is just another nice tool.
A practical four step method
This is the method I suggest to the business owners I work with:
- 1. Choose three specific tasks. Repetitive, heavy on text or information, and where a mistake is easy to spot: replies to quote requests, meeting summaries, first drafts of proposals. To begin with, avoid critical decisions or calculations nobody will check.
- 2. Measure before and after. For two weeks, record how long each task takes without AI. Then two more weeks with AI. Use the clock, not your impression. The METR study shows that the feeling is misleading.
- 3. Redirect the hours saved. Write down where the freed time will go: more calls to customers, more follow up on proposals, visits to key accounts. Assign an owner and a weekly number.
- 4. Review quality. Once a week, someone experienced checks a sample of the work done with AI. If quality drops, that task is outside the frontier and should go back to the old process or be redesigned.
After a month, you have real data from your own business, not supplier promises. Keep what works, drop what does not, and extend it to the next tasks.
This week's action
Choose one single task that repeats every week in your business. Ask the person who does it to time the next five occurrences, without AI. Write down where the time would go if you could save half of it. Only then try AI. Starting with measurement, not with the tool, is what separates those who reap results from those who merely experiment.
If you would like help identifying the right tasks in your business and turning saved hours into sales and margin, the ActionCOACH Porto team can help, whether through a Free Business Analysis or One to One Coaching sessions.