AI for Lawyers: How to Use It Responsibly in Daily Practice
Where generative AI genuinely helps in legal work, where it does not, and the habits that keep its use within professional duties.
Generative AI can summarise a deposition, propose a first draft of a clause or suggest search terms in seconds. It can also invent a case citation with total confidence. For most firms the question is no longer whether to use AI for lawyers' daily work, but how to use it so that speed does not come at the cost of accuracy, confidentiality or professional duties.
This guide is for practising lawyers and legal operations staff at small and mid-sized firms. It sorts common tasks by risk, describes the supervision each needs and sets out habits that line up with the ethics guidance bar regulators have issued so far.
This article is general information, not legal advice. Professional conduct rules differ between jurisdictions; check the rules and opinions that apply to you.
What the ethics guidance says, in brief
In the United States, the American Bar Association addressed generative AI in Formal Opinion 512. It ties AI use to existing duties rather than creating new ones: competence, confidentiality, communication with clients, supervision, candour to the tribunal and reasonable fees. Many state bars have published their own guidance, and courts in some jurisdictions have issued standing orders on AI-assisted filings.
The practical takeaway is that AI does not change who is responsible. The lawyer who signs the work owns it.
Sort tasks by risk before you start
Not every task carries the same exposure. A simple three-tier view helps you decide how much checking is required.
| Risk tier | Example tasks | Minimum review |
|---|---|---|
| Lower | Rewording your own text for clarity, building a checklist from a statute you have read, brainstorming deposition topics | Read the output before using it |
| Medium | Summarising documents in a matter, first drafts of routine clauses, organising a chronology | Compare against the source documents line by line |
| Higher | Legal research conclusions, citations, advice to clients, anything filed with a court | Independently verify every authority and statement of law |
The tier depends on consequences, not on how simple the prompt felt.
Habits that keep AI use defensible
Verify every authority
Never cite a case, statute or regulation you have not pulled up and read in a trusted source. Language models can produce plausible citations that do not exist or misstate what a real case held. Courts have sanctioned lawyers for filing briefs containing fabricated citations, and the explanation that a tool produced them has not been accepted as an excuse.
Keep confidential information under control
Before entering client information into any tool, know where the data goes, who can access it, whether it is used to train models and how long it is retained. ABA Model Rule 1.6 requires reasonable efforts to prevent unauthorised disclosure. Consumer chat tools with unclear terms are a poor fit for client matters.
Record what you used AI for
A short note in the matter file, such as "first-draft summary of the lease generated with AI, reviewed against original on 8 October," makes supervision visible and helps if questions arise later about how work was produced.
Supervise staff use
Paralegals and junior lawyers may already be using tools on their own. Supervisory duties extend to how non-lawyer assistants and others in the firm use technology. A short written policy and a list of approved tools is more effective than an informal "be careful."
Bill honestly
If AI cuts a three-hour task to forty minutes, billing three hours raises obvious problems. Opinion 512 discusses fees, including that lawyers generally should not charge clients for time spent learning a tool they will use across their practice. Agree in advance how AI-assisted work will be billed.
Where AI tends to help most
Firms often see the clearest gains in tasks where a lawyer can quickly check the output against a known source:
- Document review triage: flagging which documents in a set mention a particular party, date or clause type.
- Summaries of long material: depositions, contracts and correspondence, checked against the original.
- First drafts of standard language: engagement letters, routine notices, simple clauses, which you then edit.
- Issue spotting: asking for a list of questions to consider, which you research yourself.
It tends to help least where accuracy depends on recent or jurisdiction-specific law that the tool may not have, or where the output cannot easily be checked.
Choosing tools
General-purpose chatbots, legal research platforms with AI features and dedicated legal AI workspaces differ in where they get information and how they handle data. When comparing them, ask whether answers come with citations you can open, whether firm data is isolated from other customers and whether you can control retention.
eLaw is one example of the dedicated approach: an AI platform for law firms and corporate legal teams covering document analysis, legal research and purpose-built agents, with cited answers and firm-isolated data. Whatever you choose, the verification habits above still apply.
A first week with AI, step by step
- Pick two lower-risk tasks you do every week and try AI on those only.
- Read the terms of service and data handling documentation for the tool.
- Write a one-page firm note on approved uses and required checks.
- Review a sample of AI-assisted work with a colleague after a month.
FAQ
Is it ethical for lawyers to use ChatGPT or similar tools?
Bar guidance generally allows lawyers to use generative AI, provided they meet existing duties of competence, confidentiality, supervision and candour. That means understanding the tool's limits, protecting client information and verifying output. The specific expectations vary by jurisdiction, so check your own bar's opinions.
Do I have to tell clients I am using AI?
It depends on the circumstances and your jurisdiction. ABA Formal Opinion 512 suggests disclosure may be needed in some situations, such as when client information is entered into a tool or when AI use affects fees or the basis of advice. Many firms address it proactively in engagement letters.
What happens if an AI tool produces a fake citation I file with a court?
Courts have imposed sanctions in such cases, and the lawyer remains responsible for every authority cited. Rules such as Federal Rule of Civil Procedure 11 require a reasonable inquiry into the legal contentions in a filing. Reading each cited authority in a reliable source is the basic safeguard.
Can AI replace junior associates?
It can speed up some tasks juniors traditionally do, such as first-pass review and summaries. It does not replace judgment, client contact or responsibility for the work. Many firms are rethinking training so juniors learn to supervise and check AI output rather than skipping the underlying skills.
