AI where it saves hours. Not where it gets in the way.

You write the intent in a line — "ask the notary for the draft deed and remind them of the deadline" — and Mandato writes the email with the matter's context already inside it: parties, references, dates and amounts.

The same goes for the pleadings and contracts a firm produces every week. Your templates carry fields filled from the client and matter records; the AI writes what the template leaves open, in the tone you have taught it.

In Spanish or in English. The same document, in the language its reader actually reads.

Upload a contract, a deed, a set of standard terms. Mandato reads it, extracts the parties, the dates and the amounts, and flags the clauses that deserve a second look: uncapped penalties, automatic renewals, submission to a foreign forum, lopsided warranties.

The output is not a verdict. It is a map: what each flagged clause says, where it sits in the document, and why it was flagged. The judgement is the lawyer's.

Record the meeting — with the consent of those present — or upload the recording. Mandato returns a structured summary, the list of commitments with an owner and a date, and the tasks already created on the matter.

What disappears in practice is not the meeting. It is the half hour afterwards spent typing it up, which is almost never billed and almost always done late.

It is the right question, and the honest answer starts by admitting that one of the models Mandato uses for routine work comes from OpenAI — the same company as ChatGPT. The difference is not the model: it is where what you do with it ends up living. There are three.

The first is context. A general assistant does not know who your client is, what stage the matter is at, or what was agreed on the call in March. You have to tell it every time, pasting fragments. Mandato's AI already knows, because it lives inside the matter.

The second is where what you type ends up. Pasting a deed or a client's passport into a consumer chat means taking personal data — often special categories of it — outside the firm's perimeter, with no processing agreement and no idea where it lands. That is a professional-secrecy problem before it is a data-protection one.

The third is what survives afterwards. A chat thread is not a matter file. An email drafted in Mandato is sent from Mandato and recorded there; the analysis is stored beside the document it analysed; the summary hangs off the meeting. None of that lives in a browser tab somebody will close on Friday.

Your firm's matters — the database and the documents — are in the European Union. AI processing is a different thing, and it deserves to be stated separately rather than hidden under the same sentence: two of the three model providers, Anthropic and OpenAI, process in the United States under Standard Contractual Clauses and the EU-U.S. Data Privacy Framework. The third, Mistral Small 4 served through Melious, runs in the European Union.

We put it that way because it is the first question a firm asks about AI, and because the honest answer is still a good one: the processing happens under a processing agreement, with a contract behind it, inside the firm's perimeter, and with every query logged. That is exactly what does not happen when someone pastes a deed into a consumer chat.

What does not change with the provider: your data is not used to train models — neither ours nor the provider's — isolation between firms is absolute, and Mandato acts as processor and signs the corresponding agreement. Every AI interaction goes into the audit log: who ran it, on which document and when.