Many companies in Andorra will not replace employees. They will simply stop hiring the next person.
That is the real shift, even if it does not make headlines in the Diari. It is not "AI is coming for your job". It is something quieter: companies in the Principat that, before hiring their fourth admin, their second sales person or their weekend receptionist, pause and ask whether that part of the work can be turned into a system. And often, the answer is yes.
That is why the question in the title shows up. It sounds provocative, but it is a conversation I have more and more often with real business owners in Andorra. Does it make sense to pay €3,000 a month for a person if most of what they will do can be set up with €500 a month in tools? Let's look at it properly, because in a market as small and expensive as ours, the answer has nuances worth understanding.
The trap of comparing €500 with €3,000
The average monthly salary in Andorra in January 2026 was €2,678.30, according to CASS statistics. When we talk about the €3,000 in the title, we are not talking about a junior admin salary: we are talking about a qualified profile, with experience or responsibility. The kind of person you actually consider hiring because they bring something.
On top of that €3,000 gross, the company pays an additional 15.5% to CASS (the employee pays another 6.5% out of their own payslip). Result: that profile costs the company €3,465 a month in salary and contributions alone. Across the year, close to €42,000. And that does not include holidays, sick leave, training, equipment, office space or the cost of replacing them when they leave.
Andorra has the advantage, compared with Spain or France, of much lower employer contributions. Even so, the real cost is always above gross. And it is not only money: the cost of a wrong hire in a small labour market, where finding qualified profiles is already hard, is much greater than the gap between one salary and another.
But that is not even the main problem with the comparison.
When you hire, you are not just buying hours. You are buying judgement, responsibility, the ability to react when something falls outside the script, and someone who owns the outcome. When you pay €500 a month for AI tools, you are buying something else: speed, consistency, scalability and automatic execution of specific tasks.
You are not comparing an AI with a person. You are comparing a system with a position.
The useful question is not which is cheaper. It is what problem you are trying to solve and which part of that problem can be turned into a repeatable process.
My take, no fluff
I am saying this early so you don't have to read between the lines.
Most companies in Andorra will not choose between hiring and using AI as if they were two symmetric options. What they will actually do is this:
- Use AI to avoid hiring too early for repetitive functions.
- Hire better when the role genuinely brings judgement, responsibility or growth.
Boiled down to a napkin rule:
Automate before hiring when the process is repeatable, measurable and allows for reasonable human review.
Hire before automating when you need to sell, lead, negotiate, improvise well or carry responsibility.
The rest of this article is the defence of that thesis applied to an Andorran SME.
Why Andorra is the right place to do this well
One data point I find relevant. The latest DESI report places Andorra above the European average in 19 of the 29 digitalisation indicators, and as a European leader in digital infrastructure and skills. Where the Principat still sits below the average is precisely in business digital transformation: the digital maturity score of the business fabric stands at 4.31 out of 10, with a target of reaching 8 by 2030, according to the Government of Andorra.
Translated: we have the plumbing of a digital country, but companies are not yet using it well. That, for anyone starting now, is good news. The gap has been identified, the cost of getting on board is low, and the Andorran business ecosystem is still taking its first serious steps in automation.
Whoever acts now finds less competition. Whoever puts it off until 2028 will be fighting in a market where automation is the norm.
What AI is starting to do (and quite well)
AI starts to seriously compete with human work when four ingredients show up at once: high volume, more or less structured inputs, repeatable criteria and a low cost of error (or cheap human review).
When all four are in place, automating stops being a promise and starts being a management decision.
Real examples from the Andorran business fabric.
A real estate agency that manually produces listings, descriptions and translations for every new property. In Andorra this multiplies: a single property needs a listing in Catalan, Spanish, French and English at minimum. Every captation involves the same six or seven admin tasks in four languages. High volume, structured inputs, reviewable error. That now runs in an automated flow in minutes, not in a junior sales person's afternoon.
A gestoría that spends hours classifying incoming emails, moving documents to the right client folder and preparing draft replies. Most of that can already be automated with supervision. The gestor still signs. But they stop starting the day with 80 unclassified emails and can take on more clients without hiring another person.
A restaurant or hotel receiving bookings, menu questions and group requests through three different channels (web, Instagram, phone), in several languages depending on the season. A well-built system answers the repetitive parts in seconds and escalates to a human only what really matters. The maître still runs the floor. But they don't reply at 11pm to confirm a Thursday table for a French group.
A freelancer or microbusiness launching their first product alone because first-level support, lead management and admin are all held up by an AI layer underneath. In the past, opening certain businesses in Andorra meant hiring from day one. Not anymore.
That last image is the one I care about most, especially here. 94% of the Andorran business fabric is microbusinesses or small companies. As parts of the operation become software, more people will build smaller, faster, more profitable companies. Not companies without people. Three-person companies doing the work that used to take ten.
And two international references: in studies with thousands of customer support agents, AI assistance improved productivity by around 14%, with an especially strong benefit among less experienced staff. In professional writing tasks, another experiment measured 40% less time and 18% higher quality. We are not talking about "it might help one day". We are talking about numbers that already change the P&L.
What AI does not do well (yet)
AI gaining ground in certain tasks does not mean it wins in every position.
The moment ambiguity, internal politics, consultative sales, negotiation, leadership or decisions with serious consequences enter the picture, the comparison changes completely. If the job demands carrying consequences, "AI gets it right a lot" is not enough. You need someone who signs.
And there is something else that gets said too rarely: bad automation is expensive. Probably more expensive than not automating at all.
Bad automation creates silent errors, rework, decisions made with incomplete data and a false sense of efficiency that takes months to detect. I have seen automated processes that, instead of saving time, drain it, because someone has to check them one by one because they don't trust the system.
Practical rule: if you can't measure whether an automation is working, it is not ready for production.
The right question
That is why a company should not ask "person or AI?". That question is a trap.
The right question is: which part of this position can I turn into a system without breaking the service?
A sales person closing complex deals does not stop being a sales person because AI prepares proposals, summarises meetings and prioritises leads. What happens is that they deliver twice as much. An operations lead does not stop making decisions because AI classifies, alerts and generates reports. What happens is that they no longer need three interns to keep the operation running.
Where AI brings the most value today is not replacing work that requires judgement. It is multiplying the people who already have it.
What a €500/month budget actually buys
A note so the number does not float in the air. €500 a month in AI does not mean "a fancy ChatGPT". It means a mix of subscriptions for the team, an automation platform to chain flows, variable API usage depending on volume, and a margin for human review and maintenance. It is not a black box. It is a small piece of infrastructure connected to your CRM, your inbox, your documents and your business rules.
It can cost less. It can cost significantly more. It depends on how many people use it, what tasks are covered and how much supervision is needed. But this is the range where AI stops being an individual toy and starts competing with real working hours.
Conclusion
The future is probably not a company without people. The future looks more like a three-person company doing the work that used to take ten.
AI is not going to wipe out positions overnight. It is going to drastically reduce how many are needed to build certain businesses. And in a small country like Andorra, where finding talent is already a structural bottleneck, that difference shows up earlier than in larger markets.
The next hire you make, whether you see it that way or not, is already competing against an automation.