Service

Automation and AI for law firms

I build custom systems for law firms: the system reads court notices from the DJEN by bar number, proposes deadlines for the lawyer to approve, alerts on WhatsApp and answers clients through a virtual assistant. The lawyer keeps deciding and reviewing everything.

Typical timeline: A first pilot in production 2 to 4 weeks after the diagnosis, on an agreed slice (for example, collecting the court notices and the deadline alerts)

Status of this service

I do not have a finished case in this sector to show yet. This page describes what is built and how it works. When a system is live and the firm authorizes it in writing, it will be published here as a case.

Who it is for

Solo lawyers and small or mid-size firms that handle dozens or hundreds of cases, receive many court notices, answer clients on WhatsApp all day and want to spend less time on triage and deadline control. It also fits firms that already use deadline software and want WhatsApp and the client record in the same place.

Problems it solves

  • A missed deadline because the notice was buried among dozens of publications a week.
  • Hours spent reading the gazette and copying dates to the calendar.
  • A client who writes "what about my case?" and waits hours for an answer that is already in the record.
  • Fear of putting AI in the firm and having it invent a ruling or tell the client what only the lawyer can say.
  • A deadline spreadsheet with no history, where nobody knows who changed what.

What I deliver

  • Daily collection of court notices from the DJEN, the national electronic gazette of the courts, filtered by the lawyer's bar number. When the firm allows it, it also reads emails from the courts.
  • AI reading of the text: case number, type of act, hearing date and the deadline mentioned.
  • A single record of clients, cases, deadlines and hearings, with a history of who changed what.
  • A deadline engine written in tested code: it proposes the due date and shows the full calculation. The lawyer approves with one tap.
  • Deadline alerts on the lawyer's WhatsApp 48 and 24 hours before, and an event on the Google calendar.
  • WhatsApp support in three levels. Green: answers with data from the record. Yellow: the AI writes a draft and the lawyer approves it before it goes out. Red: does not answer about the case, says the lawyer will reply and notifies the lawyer.
  • Draft pleadings built from the firm's own templates. A citation only goes in with the number and the link to the official source, and everything is sent for review.
  • A lean dashboard with an approval queue for deadlines and replies.

How it works

  1. Free diagnosis

    Where notices come from today, how many cases there are, who checks the deadlines and which systems the firm already uses.

  2. A written proposal, in phases, each with deliverables and acceptance criteria agreed before we start. The split I usually propose is in the next three steps.

  3. Phase 1, deadlines and support

    Collection from the DJEN, record keeping, proposed deadlines with approval, alerts, calendar and case-status queries on WhatsApp.

  4. Phase 2, replies and clients

    Drafts the lawyer approves, handover to a human, record and folder for a new client.

  5. Phase 3, pleadings

    Drafts in the firm's templates, based on the case file and with case law linked to the official source.

  6. Operation

    WhatsApp Business and Google stay in the firm's name, the server is dedicated to the firm, and at the end of the contract it receives the code, the data and the migrated server.

Typical timeline

A first pilot in production 2 to 4 weeks after the diagnosis, on an agreed slice (for example, collecting the court notices and the deadline alerts). The full system is delivered in phases, and the total timeline depends on the scope and on the firm releasing the accesses.

Related case

Frequently asked questions

Can the AI cite made-up case law?

Left alone, it can get things wrong. That is why the system only cites rulings that were retrieved from an official source, and the citation goes into the draft with the link. A citation without a link that can be checked does not go in. Every draft goes through the lawyer's review before it leaves. The legal press has already reported fines on lawyers for citing precedents that did not exist, with notice to the bar association. The lawyer is the one who answers for it, and the system is designed to help them check.

Does the Brazilian bar association allow AI customer support?

The Brazilian bar association's national council recommendation on generative AI, from 2024, does not treat AI customer support as forbidden. It asks three things of the chatbot: it must not perform activities reserved to lawyers, it must say transparently that it is a machine, and the client must be able to talk to a person on request. It is a recommendation, with no sanction of its own, and what applies to your firm is a matter for the bar association and your own lawyer. I build the system around those points: the first message of every conversation says it is the firm's automated service, the assistant never denies being an AI when asked, the lawyer approves anything about the case and the client can ask for a person at any time.

Does the AI calculate the deadline by itself?

No. The system proposes the deadline and shows the full calculation: availability date, publication date, start date, the days counted and the holidays considered. The lawyer approves with one tap. The count has points that only a person checks: local holidays, court suspension orders, double deadlines and the difference between small-claims courts and regular courts. That is why a deadline is never automatic.

Where do the court notices come from?

From the DJEN, the national electronic court gazette, which is the official channel for notifying lawyers, created by the national council of justice (CNJ). The public query asks for no password and accepts a filter by bar number. I prefer the bar number over the name, because a name can be stored with broken accents and the search then misses notices. The DJEN gathers state, labor and federal courts. Before starting, I check the coverage of each court against the firm's real cases.

Can I pull case law from Jusbrasil?

Not through a robot. Jusbrasil's terms of use forbid bots, content scraping and unauthorized API access. The case-law search uses the courts' own websites and the firm's own archive, and a draft cites only what has a number and a link for the lawyer to check. Each source's terms of use are read before it is added.

Is my clients' data safe?

The WhatsApp Business account and Google stay in the firm's name, and the server is dedicated to the firm, with an encrypted daily backup. Under Brazil's data protection law (LGPD), the firm is the data controller and I am the processor. I reduce what goes to the AI model (for example, swapping names and tax IDs for codes) and choose a provider that lets the firm's content stay out of model training. At the end of the contract, the firm receives the code, the data and the migrated server.

Let's talk?

The diagnosis is a conversation, free and with no commitment. You leave with the list of processes that eat the most hours of your team.

Talk on WhatsApp

Other services

Want this running in your operation?

Tell me in one sentence. I reply with what can be done, how long it takes and what it costs.

Tell me the problem