AI for law firms: client intake, drafting, contract review and deadlines for small firms, without risking client confidentiality

It’s 9 p.m. and you still have three enquiries from the website form to answer, today’s court notifications to go through, a claim to finish by Thursday and a week of time entries to reconstruct for billing. In a small or mid-sized firm, lawyers are also the receptionist, the paralegal, the librarian and the credit controller, and the legal work gets what’s left of the day. This page goes through it process by process: what AI can do today, what your legal research platform already does, and what still has to be signed off by a lawyer. It’s written from Europe, for firms that can’t send client files to just any tool.

The short answer

• What you can automate: answering and screening new enquiries (website, WhatsApp, phone) with a conflict check; first drafts of pleadings, letters and contracts from your own templates and the case file; reviewing contracts against your checklist; summarising case files, disclosure and hearing transcripts; reading court notifications and proposing deadlines for review; time capture and billing; keeping clients updated; and an internal assistant that searches the firm’s own precedents.
• What it saves: hours of admin and first-pass reading every week. On one of our document-heavy projects, analysing the specifications for large public tenders went from up to two weeks to a few days, with the same team.
• What it doesn’t do: it doesn’t sign, file or advise. AI prepares, summarises, checks and flags; a lawyer reviews every citation, every deadline and every document before it leaves the firm.
• What it costs: an assistant that answers from the firm’s own information, from around €1,000; an agent connected to 1–3 tools (your practice management system, email, WhatsApp), €2,000–5,000; a multi-system or phone project, from €5,000. Plus usage, which is cents per conversation or document. Details in our guide to how much an AI agent costs.
• How to start: with client intake and court notifications. They don’t touch the merits of any matter, and you can measure the result within weeks. Then drafting, contract review and the assistant over your precedents.

What you can automate in a law firm

Process	What AI or automation does	Who benefits
Client intake	Handles the enquiry by web, WhatsApp or phone, collects the facts, runs a conflict check, books a consultation or politely declines	Partners, front office
Engagement and onboarding	Prepares the engagement letter and payment on account, requests documents and opens the matter	Front office, accounts
First drafts	Drafts a first version from your templates and the facts in the file	Fee earners
Contract review	Compares the contract with your checklist and flags missing, extra or off-market clauses	Commercial and property lawyers
Legal research	Searches legislation and case law in verifiable sources, citing where each answer comes from	Lawyers, trainees
Summaries	Summarises case files, the other side’s documents and hearing transcripts	Fee earners
Court notices and deadlines	Reads incoming notifications, identifies matter, type and deadline, and adds it to the calendar for review	The whole firm
Time and billing	Suggests time entries from emails, calls and documents and prepares the invoice	Fee earners, accounts
Client updates	Notifies each milestone and answers “how’s my case going?” with the real status	Clients, front office
Firm knowledge	An assistant that searches your own pleadings, opinions and templates, with per-person permissions	Everyone

1. Client intake: screening, conflict checks and booking

Before: enquiries arrive through the website form, WhatsApp, the phone and referrals. Many are outside your practice areas, some just want free advice, and a few come from the other side of a current client’s matter. A partner or the receptionist reads them all, returns calls, asks the same questions five times and searches the practice management system by hand to see whether a name rings a bell.

With an AI agent: the agent answers on whatever channel the client uses, at any hour, and asks what your best receptionist would ask: what happened, when, who the other party is, whether any deadline is running (a judgment served, a dismissal, a letter before action). It checks party names against your clients and matters to flag possible conflicts of interest, classifies the enquiry by practice area and urgency, and books a slot in the right lawyer’s diary with a two-paragraph summary waiting for them. If the matter isn’t for you, it says so politely and, if you want, refers it on.

Two clear limits: the agent doesn’t give legal advice (it explains the firm, any fee guidance you publish and the next step), and a lawyer makes the final call on any conflict. To decide what it can answer on its own and what a person must see, we use decision models such as Jev, from TypeSafe, which return calibrated probabilities instead of text; we run it in production to classify the messages from our own contact form. “What are your opening hours?” goes out automatically; “I’ve just been served with a judgment” goes to a lawyer the same day.

We’ve built this in another sector: a WhatsApp AI agent for a domestic staffing agency that asks every candidate the screening questions and leaves the team only the ones who meet the requirements. More on channels in our AI customer service service.

2. Engagement letter, money on account and opening the file

Before: client accepted, so someone writes the engagement letter, requests money on account, collects ID and documents, creates the folder and opens the matter in the practice management and accounts systems.

With automation: using what the agent already collected, the engagement letter is generated from your template (matter type, fees, payment on account), sent for e-signature, the documents are requested with a specific list, and the matter is opened in every system at once. It’s the same thing we did for employee onboarding in a group of companies: from 6 systems by hand to a single entry.

3. First drafts of pleadings, letters and contracts

Before: find a similar document from another matter, copy it, change the names and facts, and hope the previous client’s name isn’t still in paragraph three.

With AI: from your templates and the case file (facts, documents, dates, amounts), AI prepares a first draft of a claim, defence, letter before action or contract in your structure and your style. The lawyer starts from an organised draft with the matter’s details in place, rather than a blank page, and spends the time on argument and strategy.

What matters is that the draft is built on your precedents, not on whatever the model thinks a pleading looks like. And anything that cites a statute or a case is checked, as we explain in point 5.

4. Contract review against your checklist

Before: a 40-page contract from the other side gets read end to end to find the usual suspects: limitation of liability, penalties, assignment, confidentiality, data protection, governing law, automatic renewal.

With AI: you define a checklist per contract type (lease, distribution, share purchase, data processing agreement…) and the AI reviews the document clause by clause: what’s missing, what’s outside what you normally accept and why, with the exact clause reference. It returns a table for the lawyer to decide on and can draft alternative wording. For bulk review (a due diligence with hundreds of contracts), the same logic classifies each contract and extracts its key terms into a spreadsheet.

This is the kind of heavy document work we already automate: for a group of companies in the Balearic Islands, analysing large tender specifications went from up to two weeks to a few days. The full story is in the multi-company group case study, and the technical side of reading documents in our document and OCR automation service.

5. Legal research: with sources, or not at all

This is the best-known risk. General-purpose AI models invent cases, with a plausible citation, date and judge. Courts have noticed:

• In Mata v. Avianca (New York, 2023), two lawyers and their firm were sanctioned $5,000 for filing a brief with cases ChatGPT had made up. It was the first widely reported case; trackers now count well over a thousand court decisions worldwide dealing with AI-hallucinated material.
• In Spain, the criminal chamber of the High Court of Justice of the Canary Islands fined a lawyer €420 in February 2026 for citing up to 48 non-existent Supreme Court judgments, and referred him to his Bar. In July 2026 the same court fined another lawyer €840 for invented Supreme and Constitutional Court rulings, noting he never acknowledged the error.
• In April 2026 the Spanish General Council of the Bar (CGAE) approved an interpretive circular on documents prepared with generative AI: AI is an auxiliary tool, the lawyer must review everything, check it against reliable legal sources and document how AI was used, and errors can lead to disciplinary liability. In the US, the ABA’s Formal Opinion 512 (2024) says much the same about competence, confidentiality and supervision.

How to do it properly: for case law, use the AI built into your legal research platform (Westlaw, Lexis, vLex and, in Spain, Aranzadi LA LEY, Lefebvre or Tirant), which answers from a real database and links every authority. In anything we build for your firm, every answer links to the document it comes from, and an automatic check flags any citation that can’t be found in the source. No citation leaves the firm until someone has opened it.

6. Summarising case files, disclosure and hearings

Before: a 600-page file or a two-hour hearing recording arrives and you need to know what was said about one specific point.

With AI: the recording is transcribed and summarised by speaker and topic, and you can ask questions that come back with the timestamp or page: “what did the expert say about when the damage occurred?”. For a case file, it builds a chronology of facts with a reference to each document. It doesn’t replace reading what matters; it tells you where what matters is.

7. Court notifications and deadlines

Before: someone logs into the court e-filing portal every day (LexNET in Spain) or receives notices forwarded by the court agent, opens each one, works out which matter it belongs to, calculates the deadline excluding non-working days and court holidays, and enters it in the diary. In Spain, if a correctly sent electronic notification isn’t opened within three days, it’s deemed served anyway (article 162 of the Civil Procedure Act): the clock runs whether anyone read it or not.

With automation and AI: the notices you download (from the portal, from the court agent’s email or from administrative e-notification sites) are read automatically. AI identifies the matter, the court, the type of order and what’s required, proposes the deadline and due date, creates a task for the responsible lawyer with a summary and escalates anything urgent. The deadline is proposed, not decided: a person always confirms it. The value is that no notice goes unowned because it landed in the inbox of someone on holiday.

One important detail: portal access usually depends on the lawyer’s digital certificate. Automating it badly means certificates copied onto servers or into scripts. More on that under the risks.

8. Time capture and billing

Before: time is recorded from memory on Friday, or not at all, so hourly bills come out short and nobody knows whether fixed-fee clients are profitable.

With AI: from sent emails, calls, meetings and documents edited in each matter, the system proposes daily time entries per matter, which the lawyer accepts or corrects in a minute. At month end it prepares the invoices or activity reports, and flags fixed-fee retainers that are eating more hours than agreed.

9. Keeping clients updated without the phone calls

The question that interrupts a firm most is “how’s my case going?”. With automation, clients get a plain-language update at each milestone (claim filed, hearing date, judgment), and can ask on WhatsApp or a portal: the agent answers with the actual status of their file, never with a legal assessment, and passes anything substantive to the lawyer.

10. An assistant over the firm’s own precedents

Before: “how did we run that claim against the same insurer two years ago?” lives in one partner’s memory and in a network folder with thousands of documents.

With AI: an internal assistant that searches your pleadings, opinions, templates and past outcomes and answers citing the exact document. This is called RAG (the AI answers from your documents, not from what it learned on the internet). In a law firm it must respect permissions: someone without access to a matter shouldn’t see its content through the assistant. We explain the options (cloud, private or fully on your own server) in our guide to local AI and RAG for business.

For more open desk work (outlining an argument, rewriting a paragraph, analysing a specific document), the team can use tools like Claude Cowork, always on a business plan and with a clear policy on which client data can be used.

What a typical rollout looks like

1. Discovery (one or two sessions). We look at how a client comes in, how a notice arrives, how a document is drafted and how time is billed in your firm, and measure where the hours go. We check what your practice management system and research platform already do, so we don’t build what you already have.
2. First process: client intake. The agent that answers and screens enquiries, with conflict checks and booking. Easy to measure: enquiries handled, consultations booked, partner time recovered.
3. Second: notifications and deadlines. Reading and routing notices with a proposed deadline, running alongside your current control until you trust it.
4. Third: documents. Templates for first drafts, the contract checklist, case file summaries.
5. Fourth: the assistant over your precedents, once the earlier steps have organised the data, plus billing.
6. Training the team at each step: what the system does, what to check and when not to trust it.

What still needs a lawyer: strategy, advice, signatures, every citation that goes into a document, the final calculation of every deadline, decisions on conflicts, negotiation and the courtroom. Automation removes the mechanical work so the firm’s time goes there.

The risks of doing it yourself

A law firm holds its clients’ secrets and is legally bound to keep them. Before you connect AI to your files, think about what can go wrong:

• Breaching confidentiality without noticing. If someone pastes a client’s pleading or contract into the free or personal version of an AI chat, that data may be used to train models and there’s no data processing agreement. Professional secrecy is the lawyer’s duty, and “a colleague put it into ChatGPT” is not a defence.
• Special category data. Files contain health data (injuries, incapacity), criminal offence data and sensitive family and financial information. GDPR gives these extra protection and the firm, as data controller, has to know where they’re processed. We explain how in EU data residency for AI models and GDPR.
• Invented citations in a signed document. Courts on both sides of the Atlantic are already fining lawyers and referring them to their Bar. A system with no sources and no citation check is a disciplinary risk for whoever signs.
• Miscalculated deadlines or notices lost in silence. An automation that stops downloading because the portal changed, and nobody notices until a deadline has passed. Every automation needs alerts when it fails, and a person always confirms the deadline.
• Exposed digital certificates. A lawyer’s certificate copied to a random server lets anyone act in their name.
• An assistant that shows what it shouldn’t. An internal search with no permissions can show a trainee a matter behind an ethical wall, or a partner’s file to someone who doesn’t work on it.
• The EU AI Act. It treats as high-risk the AI used by judicial authorities (or on their behalf) to research and interpret facts and law, and AI used in a similar way in alternative dispute resolution (Annex III, point 8). A firm’s internal tools for its own work don’t, in principle, fall under that point, but other obligations do apply: AI literacy for your team (Article 4), telling clients when they’re talking to an AI chatbot, and high-risk rules if you use AI to screen job applicants. If you arbitrate or mediate with AI support, look at it carefully. We summarise it in our EU AI Act guide for businesses.
• A half-finished project. A tool that doesn’t integrate with your practice management system just moves the work around, and the team goes back to doing it by hand.

A team that has done this before starts with the data architecture (what leaves the firm and what doesn’t), puts a lawyer at every point that carries liability, builds citation and deadline checks in from day one, and monitors that everything keeps working.

Want to use AI in your firm without putting client confidentiality at risk? Tell us how you work in a free call.

How we do it at Aitomatik

• We don’t compete with your research platform. For case law, your publisher’s tool is the right one. We build what sits around it: client intake, notifications, drafts from your templates, contract review against your criteria, billing and the assistant over your documents, integrated with what you already use.
• Your clients’ data stays in Europe. For EU firms we use models processed in Europe, with zero data retention and no training on your data, plus a backup model from a different provider, also in the EU. International firms aren’t limited to EU models. When AI needs to query your practice management system or your folders, we install a read-only gateway inside your network: no open ports, per-person permissions (through your Active Directory, if you have one), credentials that never leave the gateway, and everything logged. More in what is MCP.
• Experience with heavy document work. For a group of companies in the Balearic Islands we mapped 75 processes across 8 departments, freed more than 1,600 hours a year, and cut tender analysis from up to two weeks to a few days. See the multi-company group case study.
• Custom software when you need it. If no off-the-shelf tool fits how your firm works, we build it: custom software development with AI, using AI-assisted development to move faster. The code is usually yours, agreed per project.
• Human review where there’s liability. Deadlines, citations, conflicts and any answer with legal content go through a lawyer. With decision models like Jev we separate what can go out automatically from what needs review.
• We save and evaluate every conversation our agents have, and test every change against real conversations before it goes live.
• We train your team on their own cases, which also covers the AI literacy duty in Article 4 of the AI Act.

If your firm also offers tax and payroll services, see our page on AI for accounting firms. All our guides are in guides and our services in services.

Frequently asked questions

Which AI is best for lawyers? It depends on the task. For case law and legislation, the AI in your legal research platform, because it cites real authorities. For drafting, summarising and analysing documents, a business assistant (Claude, ChatGPT or Gemini on a business plan, with a data processing agreement). For automating the firm itself (intake, notifications, billing), a solution connected to your own systems. We compare the assistants in ChatGPT vs Claude vs Gemini for business.

Is there a ChatGPT for lawyers? The legal research platforms and legal AI vendors offer chat assistants trained to work over legal databases and cite sources. General assistants like ChatGPT or Claude can also be used by lawyers on a business plan, for drafting and analysis, but not as a source of case law without checking every citation.

Can I use ChatGPT with client documents? Only on a business plan with a data processing agreement and no training on your data, plus a clear internal policy. Even then, consider where the data is processed and whether the client should know. Never on a personal account.

What’s the best AI tool for a small law firm? Usually not one tool. Most small firms get the most from three things: the AI in the research platform they already pay for, a business AI assistant for desk work, and automation of their own intake, notices and billing. The third is where most of the admin hours are, and it’s what off-the-shelf tools cover least.

Will AI replace lawyers? Not the work clients pay for: judgement, strategy, signature and responsibility. It does replace a lot of the mechanical work (answering enquiries you won’t take, copying documents, searching folders, logging deadlines), and that changes how many matters the same team can handle well.

How much does it cost to bring AI into a law firm? As a guide, an assistant that answers from the firm’s information starts at around €1,000, an agent connected to 1–3 tools costs €2,000–5,000, and a multi-system or phone project starts at €5,000, plus usage.

Shall we talk about your firm?

Tell us how many you are, which practice management system and research platform you use, how clients reach you and what eats the most hours each week. We’ll tell you what we’d automate first, how we’d do it without client files leaving where they should be, and what it would cost. See our AI and automation consulting service or book a free call.

Cases and rules cited are as of October 2026. This page isn’t legal advice. Last reviewed: 15 October 2026.