// ai for professions — lawyers (us)

AI for Lawyers — Roadmap

24 practical topics for US practice: writing prompts, contracts, demand letters, briefs and records, and case law research and verification.
A learning path, not legal advice: it was not written by a lawyer.
Argentine version, in Spanish: IA para abogados →
Total
24
Done
0
Progress
0%
Left
24
Est. time
~11 h

Before you start

basics 0/3 ▾

Why AI invents citations

It predicts plausible text; it does not look anything up unless a tool makes it. Mata v. Avianca and a thousand cases since show what happens when nobody checks.

What not to paste: confidentiality

Rule 1.6 in practice: strip names and identifiers before you paste, and know whether the tool keeps or trains on your inputs.

Your signature is still on it: ABA Opinion 512

Competence, confidentiality, candor, supervision and fees with generative AI, in one read.

Writing prompts

prompts 0/4 ▾

How to write a good prompt

Role, matter context, task, output format, the jurisdiction and what not to invent. If the answer misses, fix the prompt instead of starting over.

Give it your text, don't ask it to invent one

Paste the contract, the letter or your template and ask about that text. It works far better than asking for one from scratch.

Don't know how to ask? Let the AI write the prompt

Tell it what you need, have it ask you questions and draft the prompt. Then use it and adjust.

Your prompt library

Save the prompts that worked (contract review, demand letter response, deposition summary) and reuse them.

Contracts

contracts 0/4 ▾

Contract review against a playbook

Missing clauses, inconsistencies and risk. You supply the playbook; the AI flags where the draft falls short.

Drafting clauses from your own models

Start from a clause you trust and adapt it to the deal, instead of asking for a generic one.

Comparing contract versions

What changed between two drafts, and what was quietly dropped. Check every flagged change against the text.

Explaining the contract to the client

Plain language without changing its legal meaning.

Demand letters

letters 0/3 ▾

Responding to a demand letter

Paste the letter and your facts, and have it propose a point-by-point response. You decide what to deny, what to concede and what to leave out.

Drafting a demand or cease-and-desist letter

From your templates and the facts of the matter. Always check the recipient, deadlines and amounts.

Calculations and deadlines

Let it build the spreadsheet or the calculation, then check it yourself. AI gets numbers and dates wrong.

Briefs & records

briefs 0/3 ▾

First drafts of briefs and memos

Structure, arguments and prose. You supply the facts, the verified authority and the judgment.

Summarizing depositions and long records

Timelines, parties and admissions with page and line cites. Check them against the transcript.

Templates and clause libraries

Your own good work as the base, instead of asking the model to invent one.

Tools

tools 0/2 ▾

Choosing a legal AI tool

Where the data goes, whether it shows sources and how often it is wrong. Test it on matters you already know.

Local models for confidential work

Open models on your own machine: the file never leaves the firm.

Case law research & verification

research 0/5 ▾

Free primary law: CourtListener and LII

Where to check a citation when you are away from Westlaw or Lexis.

Use AI to find what to search for

Terms, doctrines and angles to take into your research database. Never as the final citation.

Verify every citation before filing

Pull and read each case: the holding, the pin cite, whether it is still good law. Every time, whatever the tool claims.

Judges' standing orders on AI

Disclosure and certification requirements that vary by judge. Check the individual practices before you file.

General chat vs research tools with retrieval

Tools that search a database and show their sources save time. Even the paid ones get 17–33% of queries wrong, so read the source.