An AI agent is a system that takes an objective, decides which steps to take and executes them against your real tools: your CRM, your inbox, your database. Unlike a model that only generates text, an agent acts: it looks things up, cross-checks them, writes into your systems and flags anything that needs a human decision. It is the piece that turns AI into work actually completed.

How is that different from a chatbot?

A chatbot answers questions inside a conversation and that is where its job ends. An agent has access to tools and permissions to run tasks end to end, even when nobody is watching. A chatbot tells you how to issue an invoice; an agent issues it, records it and sends it.

How much does an AI project cost?

It depends on scope: automating one workflow is not the same as building a full platform with several integrations. We do not publish fixed prices because quoting without knowing your systems would be guesswork. After the 30-minute call we send a proposal with scope, timeline and a fixed price, with no surprises later.

How long until it is up and running?

The opportunity audit takes one to two weeks. The build usually runs three to eight weeks depending on the number of integrations and the state of your data. We prioritise getting something useful into production early and extending it afterwards, rather than disappearing for six months.

Do I need a technical team in-house?

Not to get started. We handle design, build and deployment, and we leave the system documented. It does help to have someone on your side who knows your processes and can grant access. If you later want your own team to maintain it, we train whoever you nominate.

Where is the system deployed and who owns the data?

Everything is deployed on your infrastructure: your cloud or your server. The data is yours and stays where you decide, and it is never used to train third-party models. The code is yours too, delivered and documented, so you do not depend on us to keep operating.

How do you handle GDPR and security?

We work on a principle of minimum exposure: only the data needed for the task gets processed, and always inside your environment. When the information is especially sensitive we use open models running locally so nothing leaves your servers. Access rules, activity logging and retention policy are agreed with you before anything is deployed.

Which tools does it integrate with?

Practically any system with an API: CRM, ERP, email, calendar, spreadsheets, databases, WhatsApp or your own internal tools. Where there is no API, we assess the alternatives before committing to anything. During the audit we review your stack and tell you plainly what can be connected and what cannot.

What happens if the agent gets something wrong?

It is designed on the assumption that it will get things wrong sometimes. That is why sensitive actions go through human validation, everything is logged, and the system flags uncertainty instead of improvising. We measure failures and tune, exactly as you would with a new process run by people.

Do you provide AI training for our team?

Yes. We run ChatGPT courses for companies, Claude training, AI fundamentals and fully tailored programmes built on your own processes. Sessions can be online or on site, and the materials stay with the company. Pairing training with an automation project is common, so the team understands what they are about to use.

There's work in your company that no longer needs a person

Custom software and automations, deployed on your infrastructure. In production in weeks, not quarters.

Services

We build the software that grows your business

Grouped by what you're after: finding where the opportunity is, building something new, taking work off your plate or understanding your numbers.

Start here

AI adoption audit

We look at how your company works from the inside and tell you where AI will make you money and where it isn't worth it. You leave with a list of opportunities ranked by impact, with the cost and timeline of each one.

We build

New software for your business

We automate

The work a person does today

We analyse

Your data, turned into decisions

How we work

The path to a system in production

We don’t ship prototypes: we ship systems that survive real work. We start with a diagnosis, and if we don’t see a clear return, we tell you before you spend.

1-2 weeks

Diagnosis

We look at your processes and come back with a list of opportunities ranked by impact and risk.
1 week

Demo of your case

We show you the case you care about most, working, before you commit to anything.
3-6 weeks

Pilot on your data

A scoped trial on your real data, with success criteria agreed up front.
Depends on scope

Production

Deployed on your infrastructure, integrated into operations and documented. The code is yours.
Ongoing and optional

Continuous improvement

We measure, tune and extend. Your team can take it over, or we stay on.

Projects

This is already in production

Not mock-ups: these are platforms with clients using them.

What it costs you today

What that manual work is costing you today

Move the three controls using your real numbers. It doesn't calculate what we'd save you: it calculates what you're already paying for leaving it as it is.

Hours a week spent on the manual task 6 h

Add up the time of everyone who touches that process. Include the time spent checking and fixing it.

Real hourly cost of those people 28 €

Gross salary plus social security, divided by hours worked. If you don't know it, €25 to €35 is the usual range for admin and operations roles.

Share of the process that is purely mechanical 70 %

Copying, pasting, reconciling, formatting, forwarding. Whatever needs no judgement. That's the part worth automating; the rest stays with a person.

6115 €

Annual cost of the mechanical work in that process

Total annual cost of the process: 8736 €
Mechanical hours per year: 218 h
That's the same as: 5,5 semanas

Over three years: 18.346 €

How it's calculated and what this figure doesn't say. Weekly hours x 52 x hourly cost, applying the mechanical percentage. It doesn't include the cost of mistakes, of decisions made late because the data wasn't there, or of the person who leaves and takes the process with them. Nor does it subtract the price of the project or assume automation covers 100% of that mechanical part. It's the floor, not the ceiling. In the diagnostic we put numbers on your specific case and tell you whether it's worth it.

FAQ

What people ask us before signing

An AI agent is a system that takes an objective, decides which steps to take and executes them against your real tools: your CRM, your inbox, your database. Unlike a model that only generates text, an agent acts: it looks things up, cross-checks them, writes into your systems and flags anything that needs a human decision. It is the piece that turns AI into work actually completed.
A chatbot answers questions inside a conversation and that is where its job ends. An agent has access to tools and permissions to run tasks end to end, even when nobody is watching. A chatbot tells you how to issue an invoice; an agent issues it, records it and sends it.
It depends on scope: automating one workflow is not the same as building a full platform with several integrations. We do not publish fixed prices because quoting without knowing your systems would be guesswork. After the 30-minute call we send a proposal with scope, timeline and a fixed price, with no surprises later.
The opportunity audit takes one to two weeks. The build usually runs three to eight weeks depending on the number of integrations and the state of your data. We prioritise getting something useful into production early and extending it afterwards, rather than disappearing for six months.
Not to get started. We handle design, build and deployment, and we leave the system documented. It does help to have someone on your side who knows your processes and can grant access. If you later want your own team to maintain it, we train whoever you nominate.
Everything is deployed on your infrastructure: your cloud or your server. The data is yours and stays where you decide, and it is never used to train third-party models. The code is yours too, delivered and documented, so you do not depend on us to keep operating.
We work on a principle of minimum exposure: only the data needed for the task gets processed, and always inside your environment. When the information is especially sensitive we use open models running locally so nothing leaves your servers. Access rules, activity logging and retention policy are agreed with you before anything is deployed.
Practically any system with an API: CRM, ERP, email, calendar, spreadsheets, databases, WhatsApp or your own internal tools. Where there is no API, we assess the alternatives before committing to anything. During the audit we review your stack and tell you plainly what can be connected and what cannot.
It is designed on the assumption that it will get things wrong sometimes. That is why sensitive actions go through human validation, everything is logged, and the system flags uncertainty instead of improvising. We measure failures and tune, exactly as you would with a new process run by people.
Yes. We run ChatGPT courses for companies, Claude training, AI fundamentals and fully tailored programmes built on your own processes. Sessions can be online or on site, and the materials stay with the company. Pairing training with an automation project is common, so the team understands what they are about to use.

Contact

Let's talk about your process

We take on few projects at a time. Tell us which process is costing you time and on the first call we'll tell you whether automating it makes sense.

The more specific the process you want to solve, the more useful our answer will be.

We work with companies in Barcelona and across Spain, on site or remotely.

There's work in your company that no longer needs a person