The business is what you know: the craft, the judgment, the way you treat customers. The machine does the grunt work, your way, and you decide. I start with one tool, at a fixed price, running on your real data in two to three weeks.
Twenty minutes. I'll tell you whether it's worth doing or not.
I'm Pau, from Vilanova i la Geltrú, near Barcelona. I spent five years in California, studying computer science and working on all kinds of software and AI projects, for companies small and very large. Now I do it for companies here and abroad, in English, Spanish and Catalan.
These are the jobs I most often take off a company's hands. If one sounds familiar, we already have something to talk about. It always works the same way: the machine does the first pass, your way, and you decide.
They go out late because I have to write them myself.
From a photo, a voice note or an email to a quote written your way, ready to check.
A photo, a voice note or an email arrives with what the customer wants. Out comes a draft built from your own past quotes: your template, your prices, your way of explaining things. You check it in ten minutes and you send it. The same goes for the emails and WhatsApps waiting for an answer, and for tenders, which are the same job only bigger.
There the machine drafts a whole course and the expert reviews it in 30 minutes. It used to take a day.
It's all in one person's head. If they leave or get sick, we stop.
An assistant that answers from your own documents and shows its source.
An assistant that answers the team's questions from your own manuals, price lists and regulations, and shows where each answer comes from. New people learn how you really work, not from a manual asleep in a drawer.
At Aldda it's more than 500 regulations: the assistant finds the answer with the exact citation and the expert verifies it.
We have to look at everything to find what matters.
Everything that comes in, read and ranked, with the reason written down. You decide.
Tenders, orders, CVs, supplier offers, customer requests: the system reads all of it against your criteria, ranks it and explains why. You look first at what's worth it, and you decide.
At Eskay, an analyst used to read everything; now they review what the system proposes, with the reason written down: 80% fewer hours. At Carmentor, the system gives a verdict on each listing and the buyer decides.
We have someone who spends the day typing in data.
Invoices, delivery notes and orders go in by themselves, and your programs pass data to each other.
Supplier invoices, delivery notes and orders are read automatically and go into the program or spreadsheet you already use, in the format your accountant asks for. What gets copied by hand today from one program to another (web shop to ERP, ERP to Excel) happens by itself. A person only looks at what doesn't add up.
There the system reads manuals, slide decks and PDFs as they arrive, with nobody preparing anything.
Reports get written at night, if they get written.
The technician talks and takes photos; the report comes out done.
The technician talks and takes photos on site, in the field or at the customer's. By the time they leave, the report or job sheet is done, in the company's template.
I'm preparing one for a farming company near here: voice and GPS in the field. When it's running, I'll show it.
We run on a fifteen-year-old spreadsheet because no program exists for what we do.
A small tool made to measure, with the same data inside.
A small tool made for your company, in your words and your steps. If it can be done inside what you already use, it's done there. If it needs a new tool, it's small and it's yours.
I co-founded it and built the whole product. It has paying subscribers.
That's the most common case. Twenty minutes and we look at where they go. If it isn't worth doing anything, I'll tell you.
For what others charge for a diagnostic report, I leave you one tool that works.
It always works the same way: the machine does the first pass, the person decides and signs off.
Yours could be next.
Let's talk ↓You don't need an AI strategy to talk to me. You only need to tell me where the hours go.
Twenty minutes, over coffee or a video call. You tell me where the hours go and I tell you whether it's worth doing anything about it.
We pick one thing. I build it from your real material: your past quotes, your prices, the way you write and decide. In two to three weeks you have it running. We fix the price before I start: you know what you'll pay before you say yes.
The code, the accounts and the documentation belong to the company: any developer could take it over. Maintenance is a small monthly fee, with no lock-in. If it works, we do the next one.
I take one project at a time, and I'm the one who answers. If there's an open public grant that fits, I'll tell you and help with the technical paperwork; I never count on it to make the numbers work.
The work moves from doing it to reviewing it. That's the whole idea.
One job, start to finish, by hand.
Everything passes in front of someone's eyes. What isn't right goes back. Nobody signs anything unread.
At 18 I left for California to study computer science.
Then I helped found and develop the product at Mentor126, an American AI startup.
Along the way I've worked on very different projects, at companies small and large: custom applications, AI assistants and agents, systems that understand documents, voice and images.
Now I'm back home, building the same kind of thing for companies here.
Twenty minutes, over coffee or a video call, whichever suits you. Send me two lines about what the company does and what eats your hours. I answer myself, and I'll tell you whether it's worth it or not.