The short version
The Brand AI Tool is a custom AI application built for your brand in 2 to 4 weeks: a product advisor on your site, an internal assistant for a team, lead handling that never sleeps, or a generator of personalised materials. Fixed fee, quoted within 24 hours of a scoping call. Built on the same stack that runs our own products. Hosted in the EU, in accounts in your name. No personal data ever goes into model training. You own the code, the content and the data, and if you never call us again, the tool keeps working.
Not a chat widget with your logo on it: a tool that knows your products, your rules and your tone, and does one job well.
Four patterns
Every tool starts from one of these four. Combining two costs more, and we say so.
1. The on-site product advisor
What it does. Helps a visitor choose: "which plan fits a team of eight", "does this fabric survive a dog", "what goes with the serum I already use". It answers from your catalogue and your guidance, recommends, and hands over to a person or a form when it should.
What it needs. Product data in one clean place (names, attributes, prices, availability, recommendation rules), the content that explains the products, your tone of voice, and a decision on where a conversation ends: a cart, a form, a booking, a human.
What it must not do. Invent products or prices: it answers from your data or says it does not know. Keep a conversation longer than you decide. Send anything a person wrote into model training. Pretend to be human.
2. The internal assistant for a team
What it does. Answers the team's questions from your own documents: procedures, policies, specs, past proposals, the onboarding folder nobody reads. Each answer comes with its source, so people can check it.
What it needs. The documents, and someone who can say which version is current. Who may see what, because an assistant that shows HR files to the whole company is a data breach with a friendly interface. Thirty real questions from the team, with the answers you would want.
What it must not do. Answer from documents a user may not read. Train on your documents or your employees' questions. Run outside the EU. Be the only place a document lives: it reads your files, it does not replace them.
3. Lead handling that never sleeps
What it does. Receives every lead the moment it arrives (form, email, WhatsApp, a campaign site), asks the two or three questions that qualify it, scores it against your rules, writes it into the CRM with a summary and tells the right person. At three in the morning, in the right language, with a reply the lead finds useful.
What it needs. Your qualification rules in plain words, access to the CRM and the channels, the reply templates you already use, and a clear line for what it may promise (nothing about price or dates, usually) and what goes to a person.
What it must not do. Close a deal, quote a price or commit to a date unless you have allowed it. Contact anyone who has not consented to be contacted. Keep personal data outside your CRM. Hide that it is automated: the first message says so.
4. The generator of personalised materials
What it does. Produces a document per person or account from your templates and your data: a proposal for this client, a plan for this patient, a report for this shop. In your format and tone, in English or Spanish, ready for a human to check and send.
What it needs. The templates and three to five good finished examples, the data source (CRM, spreadsheet, form answers), the rules for what varies and what never changes (legal text, prices, disclaimers), and who reviews before anything goes out.
What it must not do. Send anything without the review step you defined. Change locked parts. Use one client's data in another client's document. Train on any of it.
Scope, and the 2 to 4 week plan
The scope: one pattern, the integrations named in the quote, English and Spanish, EU hosting in your name, a test set of real questions or cases, handover with a manual, two rounds of revisions. Anything else is a quoted extra, before the work, not after.
Week 1: data and a first prototype. The scoping call and its worksheet (below) come first: the scope in writing and the fixed fee reach you within 24 hours of that call. Then we map content and data: what exists, what is current, who owns it. We agree the test set: thirty real questions or cases and the answer you want for each. Hosting and accounts go up in your name. By Friday you have the data map and a first prototype on a sample of your content, so the conversation about tone starts from something real.
Week 2: build. The full content goes in. Integrations are wired: CRM, email, site, WhatsApp, calendar, whatever is in scope. Guardrails are written: what it may say, what it may not, when it hands over, what it does when it does not know. It answers the test set and we go through the misses together.
Week 3: real use. A small group of real users (your team, a few customers, a partner) uses it for a week. We read every conversation, fix the misses, tune the tone, add the questions nobody expected. For lead handling and the generator, the review flow is rehearsed end to end. Round one of revisions.
Week 4: handover. Round two. Monitoring and alerts. The one-page manual: how to update content, how to read the logs, what to do when it says something odd. Repository, accounts, access and data map handed over. Live.
A two-week build is the same plan with weeks 2, 3 and 4 merged into one: clean content, one integration, a smaller test group, revisions and handover in the same week. We tell you on the call which one you are.
The stack
Built on the same stack that runs our own products, including a marketplace app live on the App Store and Google Play that has to survive real users every day. Components we already run in production, not a demo. The model is one component among several, chosen for the job and replaceable when a better one appears, so your tool is not married to a provider. EU hosting throughout, and the data map says where each kind of data lives and for how long.
What the scoping call covers
Forty-five minutes. You leave with a written scope and a fixed fee within 24 hours. The worksheet:
- The job. One sentence: who uses it, for what, and what happens when it works. If we cannot write that sentence, we do not build yet.
- The pattern. Which of the four, and whether a second one is needed or just nice.
- The users. Customers, team, leads, partners. Their languages.
- The content. What it answers from, where it lives, who keeps it current.
- The data. What personal data it sees, where it comes from, where it goes, how long it is kept, who can access it. This becomes the data map.
- The systems. CRM, email, site, messaging, calendar, ticketing. Who gives access, and by when.
- The rules. What it may promise, what it must never say, when it hands over.
- The review. For anything it produces or sends: who checks it, before or after.
- Success in numbers. What you will look at in month two: leads qualified, questions answered without escalation, hours saved. Your numbers, in your accounts.
- Life without the tool. The honest alternative, because sometimes it is better.
When a tool is the wrong answer
We say this on the call, and we would rather lose the project than build something you switch off in three months.
- The same ten questions: a FAQ page is enough. Cheaper, indexed by search engines, built with the structure that lets assistants cite it, and it never says anything odd. Regular ChatGPT use in Spain went from 4% to 28% between 2023 and 2025 (Funcas); a clear page is what those assistants read, a chat widget is not.
- Collecting information: a form is enough. Good questions and lead routing into your CRM is a Launch Page job, 72 hours.
- A choice between three options: a quiz is enough. Fixed, predictable, fast to build: the Campaign Microsite, 1 to 2 weeks.
- The content does not exist, or nobody keeps it current. The tool will answer from what is there. Fix the content first; we can help, but it is a different project.
- The volume is low. A person is cheaper and better. A tool that handles four leads a week is a hobby.
The tool is right when the questions are open-ended, the content is real, the volume justifies it, and the job happens at hours or in languages your team cannot cover.
Who it is for
Brands and companies in Spain and Europe with real content, real volume and a job that runs when the office is closed: a catalogue people need help with, a team that keeps asking the same experts, leads arriving at all hours, documents personalised by hand today.
Agencies and consultants who want to offer an AI tool without building a development team. The tool is built in the client's name; you keep the relationship. Referral, co-branded or white-label, with a free AI Visibility Audit for one client to start. Details on the partners page.
Where to go next
- How visible your brand is in AI answers before you build: the AI Visibility Audit, 5 days.
- A fixed-question quiz or a calculator rather than an open assistant: the Campaign Microsite, 1 to 2 weeks.
- A page first: the Launch Page, 72 hours.
- All packages on the home page. This page in Spanish: Herramienta de IA para tu marca.
Tell me the job in one sentence. If a tool is the right answer, you will have a scope and a fixed fee within 24 hours of the call. If it is not, you will hear that too.