AI for lawyers is already in firms and in corporate legal teams, and the profession has set out its position: errors produced by AI systems can create professional liability for a failure to verify and to supervise (Consejo General de la Abogacía Española, interpretive circular on generative AI 2026). For the Council, AI must not operate as a substitute for the professional, «but as an auxiliary function, subject moreover to human supervision». This guide sets out where it fits, how to verify it, and how to train the legal team.
What is AI for lawyers used for?
AI for lawyers is used mostly to review and to prepare: locating clauses in contracts, comparing versions, searching the firm's or the company's internal documentation, summarising case files and writing first drafts. In all of these the AI speeds up the mechanical part, and the lawyer reviews, decides and signs.
| Task | What the AI does | What the lawyer does |
|---|---|---|
| Contract review | Locates clauses, compares versions and flags differences | Weighs the risk and decides what to negotiate |
| Internal documentation | Finds the document and the page holding the information | Interprets and applies it to the case |
| Summaries | Summarises case files, minutes or long reports | Checks that nothing relevant is missing |
| First drafts | Prepares a first version of a filing or a letter | Reviews, verifies every reference and signs |
Contracts have a difficulty of their own: terms are defined in one clause and used throughout the document, with cross-references between sections. An AI reading loose fragments can lose that relationship, so on long documents it is worth always asking it to say which clause each statement comes from.
What is the best AI for lawyers?
The best AI for lawyers depends on the task: for researching legislation and case law, specialised legal assistants working over legal databases fit; for drafting and summarising your own documents, general models such as Claude, Microsoft Copilot or ChatGPT will do, always with verification. Two different families live under the label of legal AI, and they do not solve the same problems.
Before choosing, check three things:
- on what terms the tool handles the data you put into it, and whether that fits your organisation's confidentiality obligations,
- whether it integrates with the systems the team already works in, such as the document manager or email,
- and whether the team knows it thoroughly, something the Consejo General de la Abogacía itself recommends.
In our AI consulting we start with an audit of the tools already circulating in the organisation, unauthorised ones included. In a legal team, shadow AI is a confidentiality risk as well.
We compare the two most used general-purpose assistants in Claude vs ChatGPT.
How do you stop AI inventing laws or rulings?
To stop AI inventing laws or rulings, every citation it generates is checked against the original source before the filing is signed. That is what the Consejo General de la Abogacía recommends: do not use results without a full review, and always cross-check against reliable legal sources (Consejo General de la Abogacía Española, interpretive circular on generative AI 2026).
The hardest risk to catch is not the obvious error but the answer that looks very correct and is not: a reference in the right format, an article that sounds plausible, a ruling with a credible date and number. Which is why the review cannot be a quick read of the output.
Two design decisions cut that risk:
- Point at the source instead of answering. The AI says which document and which page holds the information, and the lawyer reads it in the original.
- Admit when it does not know. The instructions ask it to say so when the information is not in the documents it holds, rather than filling the gap.
Legal · Verification
Five steps, and the last is always a person.
Draft
The AI prepares the first version of the filing.
Citation check
Every reference is looked up in the original source. None is taken on trust.
Full review
A reading of the whole text, not of the AI's summary.
Record of use
When and what for the AI was used is written down.
The lawyer signs
The responsibility is not delegated.
Source: Ironhack for Business, based on the recommendations of the Consejo General de la Abogacía Española.
If the tool also acts on its own, apply the rules for AI agents.
Confidentiality, traceability and transparency: what the profession requires
The profession's institutions have already published criteria for using AI. This guide is not legal advice, but these are the published principles worth keeping in mind:
- Human control and responsibility. AI can never replace human control, and the professional keeps full responsibility for their work (Consejo General de la Abogacía Española, ten principles for the ethical use of AI 2026).
- Transparency with the client. The client must know when automated tools are involved, according to the same document.
- Traceability. Documenting when and what for AI has been used makes decisions and responsibilities defensible (Consejo General de la Abogacía Española, interpretive circular on generative AI 2026).
- Internal protocols. Define which tools may be used, for what and under what conditions, and name who supervises their use, as the ICAM guide sets out (ICAM, practical guide to the responsible use of AI 2025).
The ICAM guide puts it in one line: «technological delegation cannot mean delegating responsibility». In practice, protecting professional privilege starts with deciding what information may go into each approved tool and what may not.
To review what training and use evidence your organisation has, you can take the Article 4 test of the AI Act.
Human resources and finance teams face the same demands on confidentiality and traceability.
Is there AI training for lawyers?
Yes, AI training for lawyers exists, and the profession itself asks for it: the Consejo General de la Abogacía's ten principles insist that a lawyer must understand how AI works in order to explain its limits and results, and keep up continuous training (Consejo General de la Abogacía Española, ten principles for the ethical use of AI 2026). The ICAM guide recommends investing in technical training, with functional testing and hands-on workshops.
Our hands-on workshops can be run per department, legal teams included, and work on the tools your organisation already has (Claude, Copilot or ChatGPT) with the team's real use cases. Every team leaves with prompt templates and a repeatable demo agent built on its own case, and the session ends with commitments to apply what was learned, signed in the room.
We measure tasks, not attendance: the hours a task takes before the training against the hours it takes 60 to 90 days later. We set the method out in the guide on the ROI of training.
Where to start with AI for lawyers: 4 steps
- Pick a low-risk task. Summarising internal documentation or locating clauses in your own contracts are good candidates. Leave filings that go to third parties for later.
- Define the use policy. Which tools are approved, what data may be entered and who supervises.
- Train the team on real cases. Include how to verify citations and how to record use.
- Record and measure. Document when and what for AI is used, and compare the hours the task took before and 60 to 90 days later.
When a procedure works, save it as a Claude skill so the whole team repeats it the same way.
Next step
If you want to know which task in your legal team is the best candidate to start with, we will come back with a starting point in under 1 working day.