A disclosure before anything else: we build Caseagent, one of the tools on this list. We have put it last, labeled it clearly, and tried to describe every competitor the way its own happiest customer would. AI tools for lawyers are not interchangeable, and the most expensive mistake in this market is buying a good tool for the wrong job. So this roundup is organized around jobs: research, enterprise-wide drafting, practice management assistance, contract work, and working the case file itself.
Legal AI tools at a glance
| Tool | Best for | Pricing signal | Main limitation |
|---|---|---|---|
| CoCounsel | AI research and document review at scale | Publicly discussed at roughly $225–$500/user/mo | No case management; priced for larger firms |
| Harvey | Enterprise firms standardizing AI across practice groups | Enterprise-only, no public pricing, waitlist | Effectively inaccessible to small firms |
| Clio Duo | Firms already running Clio who want AI assistance | Paid add-on to Clio plans ($39–$139/user/mo base) | Chat-style assistant, not an agent working the file |
| Spellbook | Contract drafting and review inside Microsoft Word | Subscription; quote-based, aimed at transactional teams | Transactional focus; not built for litigation files |
| Lexis+ AI | Research grounded in the LexisNexis database | Quote-based, typically bundled with Lexis subscriptions | Research-centric; lives outside your matters |
| Caseagent early access · our product | An agent inside case management: reads, drafts, calendars | Planned: $149–$899/mo per firm, not per user | Early access; onboarding in batches, no track record yet |
A note on the pricing column: legal AI vendors change pricing often and several publish nothing at all. Where we state numbers, they are publicly known anchors (published price lists or widely reported figures), phrased as ranges. Where a vendor sells by quote, we say so rather than guess. Always confirm current pricing directly before budgeting; treat this table as a map of magnitudes, not a rate card.
The tools, one by one
CoCounsel (Thomson Reuters)
CoCounsel, originally built by Casetext and acquired by Thomson Reuters, is the reference point for AI legal research and document review. It runs defined skills: research memos, deposition prep, document review across large sets, contract analysis. Its research work is grounded in a real case law database, which is exactly what you want after the hallucinated-citation era.
Strengths: depth and reliability on research tasks, large-scale document review, the backing of the Westlaw ecosystem. Tradeoffs: pricing publicly discussed in the range of $225 to $500 per user per month puts it out of reach for many small firms, and it has no concept of a matter: no deadlines, no client communication, no case file. It answers questions; it does not run your caseload. If cost is the sticking point, we compared the field in our CoCounsel alternatives guide.
Harvey
Harvey positions itself as the AI platform for elite law firms and in-house teams, and by public reporting it has signed some of the largest firms in the world. It offers assistant-style drafting and research plus custom workflows built with clients.
Strengths: serious enterprise engineering, firm-wide deployments, custom workflow work. Tradeoffs: it is enterprise-only, does not publish pricing, and access historically ran through a waitlist and sales process. For the roughly three quarters of US firms with fewer than ten lawyers, Harvey is effectively not an option, which is precisely the gap smaller-firm tools compete for. If that describes your firm, our roundup of Harvey AI alternatives prices the options that will actually take your call.
Clio Duo
Clio Duo is Clio's AI add-on, layered onto the most widely used practice management platform in the small-firm market. It summarizes matters, answers questions about your Clio data, and helps with routine tasks like drafting emails and logging time.
Strengths: if your firm already lives in Clio, Duo adds AI where your data already is, with a vendor whose base plans publicly run $39–$139 per user per month. Tradeoffs: it is a chat-style assistant bolted onto software built for record keeping. It helps you find and summarize what is in the system; it does not autonomously work the file, draft court documents from the record, or maintain a deadline timeline the way agentic tools aim to.
Spellbook
Spellbook lives inside Microsoft Word and focuses on contracts: drafting clauses, redlining, flagging unusual terms, and benchmarking language against market practice. For transactional lawyers who spend their day in Word, the workflow fit is the whole point.
Strengths: excellent workflow fit for contract work, fast adoption, no new system to learn. Tradeoffs: it is a contracts tool. Litigation practices, deadline management, research memos, and client communication sit outside its lane, and pricing is quote-based rather than published.
Lexis+ AI
Lexis+ AI is LexisNexis's answer to CoCounsel: conversational research, drafting assistance, and summarization grounded in the Lexis database, with citations linked to the underlying authority for verification.
Strengths: grounding in one of the two major research databases, linked citations, a natural upgrade for existing Lexis subscribers. Tradeoffs: pricing is quote-based and typically tied to broader Lexis contracts, and like CoCounsel it is a research destination, not a system that knows your matters. How to verify what any research AI hands you is a topic big enough that we wrote a separate guide: AI legal research and how to verify it.
Caseagent (early access, our product)
Caseagent is the tool we are building, and it exists because of the gap the table above makes visible: research tools have no case file, and case management tools have no real agent. Caseagent is legal case management software with the agent inside the matter: it reads the file, builds the chronology, drafts documents from the record, researches with citations flagged for verification, and keeps every deadline on a watched timeline. The full loop is on our how it works page.
Strengths: the agent works from the actual file rather than a chat box, and planned pricing ($149–$899 per month) is per firm, not per user. Tradeoffs, stated plainly: we are in early access with zero public customers. Onboarding opens in batches, and the only proof we can offer today is the live demo on our homepage. If you need a tool with years of production history, pick one of the tools above; if you want the agent-in-the-file category early, that is what the early-access list is for.
What about ChatGPT and general chatbots?
They belong in this conversation because, realistically, they are the AI tools most lawyers have already tried. General chatbots are genuinely useful for tasks with no client data and no citation risk: rewording your own paragraph, explaining an unfamiliar concept, brainstorming a deposition outline you will rebuild anyway. They are the wrong tool for anything that touches a real matter, for two reasons this list keeps returning to. First, consumer tiers may retain prompts and use them in ways incompatible with your duty of confidentiality. Second, a general chatbot is not grounded in a legal database, which is how fabricated citations end up in filings and lawyers end up in sanctions hearings. Every purpose-built tool above exists, in one way or another, to fix those two problems.
A related note on tools this roundup leaves out: there is a long tail of legal AI point solutions for immigration forms, e-discovery, patent drafting, and court analytics. They can be excellent in their niches. We limited this list to the tools a general small-to-mid-size practice is most likely to shortlist in 2026.
How to choose: four questions that decide it
- Can you verify its output? Every claim, citation, and computed date should trace to a source you can open. Tools that ground answers in real databases and flag authority for verification are usable; tools that produce unsourced prose are a sanctions risk wearing a subscription fee.
- What happens to client data? Get written answers on retention, training use (must be no), processing location, and deletion. If a vendor is vague here, the evaluation is over.
- Does it fit where you already work? Spellbook wins contract work because it lives in Word. The same logic applies everywhere: AI that sits inside your matter workflow gets used daily; AI in another tab becomes a monthly curiosity. Count the copy-paste steps in the vendor demo.
- What does the price actually buy? Per-user pricing stacks fast in a growing firm; add-on pricing assumes you keep the base subscription; enterprise pricing assumes a procurement team. Model the real annual cost for your headcount before comparing anything else.
- Who is accountable when it is wrong? Every tool on this list will occasionally produce a bad draft or an incomplete answer; the professional responsibility stays with the reviewing attorney regardless of vendor. What differs between products is how easy they make the review: linked sources you can open in one click, flags on unverified authority, and clear markers on what is AI-generated. Prefer the tool that assumes it will be checked.
Red flags in any legal AI sales process
Whichever direction you lean, a few patterns should slow you down regardless of the logo on the deck:
- No hands-on trial. If you cannot put your own sanitized fact pattern into the product before signing, the vendor is telling you the demo script matters more than your workflow.
- Accuracy claims without methodology. "99% accurate" means nothing without knowing accurate at what, measured how, on which tasks. Ask for the evaluation setup; watch what happens to the conversation.
- Vague answers on data handling. Retention, training use, and deletion should be answered in one email, in writing. A vendor that needs to check with someone has already answered.
- Pressure pricing. Expiring discounts and this-week-only tiers are a signal about churn, not about you.
- Promises that remove the lawyer. Any pitch implying you can skip review of AI output is inviting you into an ethics problem. That includes our pitch: if you ever hear it from us, hold us to this paragraph.
One practical closing suggestion: shortlist two tools from different rows of the table, not two from the same row. Comparing CoCounsel against Lexis+ AI tells you which research assistant you prefer; comparing a research assistant against a tool that lives in your workflow tells you which job you actually needed done. Run both on the same real task for two weeks, count the edits and the copy-paste steps, and the decision usually makes itself.
The honest summary: buy CoCounsel or Lexis+ AI for research depth, Harvey if you are an enterprise firm and can get through the door, Clio Duo if you are committed to Clio, and Spellbook if your practice lives in contracts. If what you actually want is the case file worked for you, that category is just arriving, and we would rather show you than tell you.