AI agents and automation for companies.

A chatbot follows a script: if the question is not on the menu, it gets stuck. An AI agent understands what people write, checks your systems and does things: it looks up stock, quotes with your rules, books appointments and logs the lead in your CRM. You give it a job, the tools to do it and the limits of what it can do on its own.

If you only need to answer frequent questions, an off-the-shelf tool is probably enough, and cheaper: try it first. What it does not do is work inside your systems: check your stock, quote with your rules, write to your CRM or ask the right person for approval. That is what an agent built on your operation is for.

Yes. On WhatsApp Business, through the official API, the agent answers, qualifies the lead, quotes with your pricing rules, books the appointment and follows up. By phone, a voice agent picks up, understands what the caller needs and logs the call in your system. In both cases, when it is needed, it hands the conversation to a person with the summary and the history.

The work that is copied and pasted today. It reads invoices, purchase orders, emails, scanned documents and product photos, and passes the data into your system. It fills the CRM, sorts the inbox, builds reports, researches prospects or suppliers and prepares drafts of replies, quotes and documents. A long job is split across a team of agents: one researches, another reviews and another writes.

We do not publish prices, because an agent that answers frequent questions and one that quotes, books and writes to your ERP do not cost the same. What moves the cost most: how many jobs it does, how many systems it connects to, whether it serves customers or your team, how many permissions and approvals it needs and how much volume it handles, which sets the model usage. That usage goes on your own account, with a spending limit you set. After a first call we give you a range, and the proposal comes in stages.

As long as the first stage takes, and we scope it to go live soon: a single job, such as answering WhatsApp or reading one type of document, tested on real cases while everything else stays as it is. Once that stage works, the rest follows. The schedule for each stage is in the proposal.

No model is perfect, which is why we fence it in. It answers from your documents (the manual, the price list, your policies) and cites where each answer came from. What it cannot find it does not guess: it says so and hands the conversation to a person. Prices come from your rules, not from the model's memory, and we test it on real conversations before letting it loose.

If you decide so, yes. You define what it does on its own, such as answering a question about opening hours, and what waits for approval: a quote, an email to a customer, a payment. The approval reaches the right person with the draft ready to review, and it is logged.

Yes, when the system allows a connection: an API, a webhook, an export or the database. We connect ERPs, CRMs, WhatsApp Business, Gmail and Outlook, Google Sheets and Drive, calendars, payments, marketplaces and your CFDI invoice data. Integration works without AI too: two systems that are copied by hand today can talk to each other on their own. We check it during discovery, before promising anything.

MCP (Model Context Protocol) is the standard an AI assistant such as Claude or ChatGPT uses to work with outside tools. We build an MCP server on top of your systems, so your team can ask their assistant for the open quotes or to log a visit, and it happens inside your system. Each person keeps their own permissions, and every action is logged.

Whichever does each job best: Claude, GPT or Gemini, and smaller models where they are enough. The agent is not tied to one: if a better or cheaper model comes out tomorrow, it is swapped without rebuilding the system.

To your systems and to accounts you control. The model accounts (Anthropic, OpenAI or Google) are in your company's name, with their own keys and billing, on API plans that, under their own policies, do not use your data for training. Each agent sees only what its job needs, and every action is logged.

Yes. On delivery, the code, the database and the service accounts are in your company's name. If you decide to work with someone else tomorrow, you take all of it with you.

We are based in Querétaro. We meet in person across the metro area and the Bajío: San Juan del Río, Celaya, León, Irapuato, San Luis Potosí and Aguascalientes. With companies elsewhere in Mexico, and with anyone who prefers to work in English, we work remotely.

Iñaki Zamores, the founder, designs and engineers every platform. The person who understands your business is the one who writes the code, with nobody in between.

We stay with you on a monthly engagement for as long as your company wants, and we recommend it: a system used every day needs someone who knows it. It covers hosting and the database, support for your team, improvements and updates to the site or the system, and the adjustments that legal changes (privacy, invoicing) or security call for. Larger new features are proposed as one more stage, with their own scope. Everything stays in your company's name: you keep working with us because it suits you, not because you depend on us.