Legal AI is useful for research, contract review, discovery, billing, and intake — if a lawyer still checks the output. Start with research you already trust: LexisNexis or Thomson Reuters CoCounsel / Casetext. Add a practice system such as Clio, e-discovery such as Everlaw or DISCO, and contract tools such as Kira (Litera), LinkSquares, or ContractSafe. Do not paste client files into a consumer chatbot.
This is not legal advice. Bar rules on competence, confidentiality, and supervision still apply. Cite-check every case a model names.
Quick picks by job
| Job | First pick | Also use | Skip if |
|---|---|---|---|
| Research and memo assist | Lexis+ AI or CoCounsel / Casetext | Harvey if the firm already standardized on it | You will file without opening the opinion |
| Practice management | Clio | Clio + a receptionist like Smith.ai | The firm will not log time anyway |
| Contract review / CLM | Kira, LinkSquares, or ContractSafe | Harvey for first-pass markup | You have twelve contracts a year |
| Discovery review | Everlaw or DISCO | The platform your ESI vendor already runs | There is no collection protocol |
| PI demand packages | EvenUp | Your existing demand workflow | Records are incomplete |
| Intake calls | Smith.ai | Clio Grow + a human ring-down | The bot gives legal conclusions |
Research, drafting, and firm ops
LexisNexis and Casetext / CoCounsel (Thomson Reuters) are the research layer most firms already pay for. AI here should retrieve authorities you can open. Hallucinated citations are malpractice bait.
Harvey is the general legal assistant some large firms deployed for research, contracts, and drafts. It still needs a supervised workspace and a human signature.
Clio is practice management: matters, time, billing, some document automation. Smith.ai answers phones and books consults. Neither is a substitute for a conflicts check.
ClickUp is generic project software. Fine for a litigation to-do list. Not a DMS.
Contracts and litigation
Kira Systems (Litera) is due-diligence extraction. LinkSquares and ContractSafe are CLM: store, extract, alert. First-pass markup tools (Harvey, AmTo-class drafters, Detangle-style summarizers) are for triage. Playbooks beat vibes.
Everlaw and DISCO run review, clustering, and privilege workflows. Predictive ranking is only as good as the seed set and QC.
EvenUp turns medical records into demand-oriented drafts for PI shops. A lawyer still owns damages theory and ethics of the ask.
Darrow surfaces potential case themes from data. That is business development plus diligence, not a filed complaint.
Paxton and similar compliance monitors watch rule changes. Confirm the jurisdiction feed.
Ferret is relationship / diligence intelligence, not a court database.
Generic domains on long listicles (legalchatbot.com, legalbillingsoftware.com, and the like) are often content or lead pages. Buy from a named product with a BAA/DPA and a bar-aware security story.
How to use this without a bar complaint
- Put client data only in a contracted legal product.
- Cite-check every authority.
- Supervise staff and the model the same way you supervise a junior.
- Tell clients if your jurisdiction or engagement letter requires disclosure of AI use.
- Keep the human on settlement numbers, advice, and court filings.
What these tools get wrong
- Fake cases
- Consumer ChatGPT with a client lease
- Outcome “predictions” sold as certainty
- A receptionist bot giving legal advice
- Discovery AI with no privilege protocol
Suggested stack
- Research: Lexis or CoCounsel
- Matters: Clio
- Contracts: one CLM or Kira if you do diligence volume
- Discovery: Everlaw or DISCO when the matter size justifies it
How we compiled this page
This page reorganizes AI Tool Rack’s legal-tool list by job and flags generic domains and competence duties. Last verified: September 3, 2026.
FAQ
What AI tools help lawyers most?
Research assistants tied to real databases, CLM/extraction for volume contracts, e-discovery review, and a PMS for time and documents.
Can I use ChatGPT for a brief?
Not as the research source, and not with confidential facts on a consumer plan. If you draft there, you still verify every cite in Lexis or Westlaw.
Is Casetext still a product?
It sits in the Thomson Reuters / CoCounsel family. Check what your firm actually licenses today.
Do outcome predictors replace judgment?
No. They are statistics over past cases, not your facts and your judge.
Is this legal advice?
No.