Ask around your company who uses artificial intelligence and you will get two surprises: more people than you thought, and none of them knows what is allowed. Somebody keeps it open all day, somebody else has never opened it, and somebody pasted a client's details into a public chat to settle a quick question. Artificial intelligence in your company does not fail for lack of tools: it fails for lack of criteria.
What is actually happening in small companies
Subscriptions get signed up "to try" and then stay. Each department ends up on a different tool, sometimes paid with somebody's personal card. Nobody has written down what can be uploaded and what cannot, and nobody checks whether the thing bought in January is still used in July.
The result is not a disaster, it is worse: it is a draw. You pay, you half-use it, and it shows up nowhere. That deadlock is what needs breaking, and buying another tool does not break it.
There is a second pattern worth naming, because it explains most of the frustration we hear. Companies compare themselves against headlines about what AI supposedly does, then judge their own results against that. Meanwhile the useful gains are far more boring: twenty minutes saved on a report, an inbox that sorts itself, a quote drafted before the client hangs up. Boring is exactly what you want here — it is repeatable, it survives a busy week, and it does not depend on anyone being enthusiastic about technology.
The three mistakes that keep repeating
- Starting with the tool instead of the task. "Let's use AI" is not a goal. "Let's stop writing repetitive email replies by hand" is, and it can be measured.
- Piloting somewhere that does not matter. If the pilot is a task nobody cares about, nobody defends the change. Start with something that hurts every week.
- Skipping the training. Handing out access is not training. Without an afternoon of practice on real company cases, half the team drops the tool within a fortnight.
The next-month test
Before signing up for anything, pick the task and ask what would have to be true in a month to call it a success. If the answer is vague — "move faster", "modernise" — it is not ready. If it is concrete — "email replies go from twenty minutes to five" — you have both a starting point and a way to measure it.
Want to know what it would cost you?
We reply the same day. No commitment, no sales pitch.
Where to start: two tools and three tasks
Less is more. Two tools for the whole company, chosen deliberately, work better than six scattered subscriptions. And three concrete tasks to begin with:
- Summarising and organising information. Meeting notes, long email threads, technical documentation. Lowest risk, quickest to convince people.
- Drafting internal text. Standard quotes, frequent replies, first drafts a person then reviews. Nothing goes out unread.
- Searching inside your own documents. Contracts, manuals, client history. This is where the most time comes back and where data needs the most care.
Five cases that already work in small companies
To bring it down to earth, five uses we see genuinely working in local businesses, nothing spectacular about any of them:
- Answering repetitive emails. A draft in five seconds that somebody reviews and sends. It returns the most time and is the easiest to defend to the team.
- Writing up meeting notes. Record, transcribe, and out comes the summary with the tasks and who owns them. It used to be done from memory, which is why it often was not done at all.
- Searching your own documents. "What did this client's contract say about renewals?" answered in seconds instead of half a morning opening folders.
- Preparing standard quotes. A first draft built from previous jobs; the numbers and conditions are always set by a person.
- Sorting what comes in. Emails, forms or tickets labelled by type and urgency so nobody has to triage by hand every morning.
None of them replaces anybody: they remove the mechanical part and leave the judgement where it was. And all of them can be trialled in a week with what you already pay for.
What is still not worth handing over
Three areas call for caution. First, any text going out with your name on it without a person reading it: AI invents facts with remarkable confidence. Second, decisions affecting people — hiring, personalised pricing, complaints — where judgement and responsibility have to stay human. Third, anything involving confidential information uploaded to tools whose terms nobody has read. If your company qualifies for Spain's Acelera pyme programme, there is free guidance there on responsible digitalisation too.
The rules worth writing down
You do not need a forty-page manual. One sheet is enough, and it has to answer three questions: which tools are used, for which tasks, and what information never leaves the company under any circumstances.
The third one matters most. Client data, contracts, supplier prices and anything with names attached stay inside. Spain's data protection agency publishes clear guidance for small companies on handling personal data with these tools, and since 2024 there is also a European AI framework worth keeping on the radar, even if most of its obligations do not apply to basic internal use.
We'll tell you in a ten-minute conversation
We reply the same day. No commitment, no sales pitch.
Training the team: the part almost everyone skips
One well-run afternoon changes the whole outcome. Not a theory talk: cases from the company itself, with the real documents, with people writing their first instructions and seeing where the tool gets it wrong.
That is where the things that actually matter get explained: that AI invents facts perfectly confidently, that whatever comes out has to be checked, and that the quality of the answer depends on how you ask. None of that is learned from an internal email.
What it costs and what it saves
The subscriptions are the cheap part: most sit between twenty and thirty euros per user per month, and plenty of companies pay for more seats than they use. The cost is the tidying work: reviewing what is contracted, choosing, writing the rules and training people.
We do it in two pieces: an AI audit to see what is used, what it costs and what is surplus; and integration and training to leave the team working with clear criteria. It is quoted per project rather than as a fee, and the pricing page publishes the recurring rates.
How we work
We are three partners based in Palamós: Ivan handles development and integrations, Jordi marketing and campaigns, Izan cybersecurity and data — which weighs more here than people expect. We work with companies across Girona and Barcelona.
If artificial intelligence is already being used in patches at your company and you want it sorted, tell us which tools you are paying for and which tasks eat your week. That is usually enough to see whether you need an audit, training, or simply to cancel two subscriptions and make proper use of the one that remains.
Tell us what you need and we'll give you a price
We reply the same day. No commitment, no sales pitch.