Harvey AI is enterprise legal AI, built for large law firms and in-house legal departments, sold by annual contract at enterprise pricing. There is no self-serve plan. If you run a solo or small firm, Harvey is not your tool, and skipping it costs you nothing.
That is an odd way to open a review. I write these for lawyers who want to know what to buy, and the honest answer here is that most readers of this site are not the customer Harvey is trying to win.
What Harvey actually is
Harvey is a legal-specific AI platform: research, drafting, document analysis, and workflow automation, configured on a firm’s own materials. The pitch is that the model works against your precedent bank, your deal files, and your matter history instead of the open internet. Many of the largest global firms have adopted it. It raised very large rounds at multi-billion-dollar valuations and became the name people say when they say “legal AI.”
The brand recognition matters more than lawyers admit. A managing partner who has heard exactly one legal AI company name has heard this one, and that shapes a lot of purchasing.
What it does well
By reputation and by what the company sells, four things:
Firm-wide knowledge integration. Answers that draw on the firm’s own work product rather than generic training data.
Matter-specific workflows. Configured sequences for the tasks a practice group repeats, rather than a blank chat box.
A security and compliance posture that enterprise clients demand. Large clients audit their firms’ vendors. Harvey is built to survive that audit.
Dedicated implementation and support. You get people, not a help center.
None of that is free, and none of it is instant.
The pricing question
Harvey does not publish pricing. It is quoted per seat, typically annual, and implementation is a project with a timeline rather than a credit card form. Anyone giving you an exact public dollar figure is guessing, and I am not going to add to the guessing.
What I can tell you is the shape of the cost: seat count times an annual per-seat number, plus onboarding, plus internal time from whoever at your firm owns the rollout. That third line item is the one firms underestimate.
Who it is for
Firms with hundreds of lawyers. Organizations with a knowledge management function, an IT and security review process, and an actual budget line for legal tech. In-house departments at large companies. If your firm already has someone whose job title includes “innovation” or “legal operations,” Harvey is worth a call.
Who it is not for
Solos. Small firms. Mid-size firms without a dedicated legal ops person. If nobody at your firm owns tool adoption as part of their job, an enterprise platform will sit unused while the invoice renews.
For those firms the honest recommendation is boring and cheap: Claude or ChatGPT at $20/month plus a prompt library built for your practice area. If you already pay for Westlaw, CoCounsel is the natural add because it sits inside research you are already buying.
| Tool | Who it’s for | Pricing model | Grounded in | Realistic setup time |
|---|---|---|---|---|
| Harvey | Large firms, in-house departments | Enterprise, per seat, annual contract, not published | Your firm’s own documents and configured workflows | Weeks to months, implementation project |
| CoCounsel | Solo through mid-size, especially Westlaw subscribers | Per seat subscription, published tiers | Thomson Reuters research content plus documents you upload | Same day |
| Lexis+ AI | Firms already on LexisNexis | Bundled with a Lexis contract | LexisNexis research content with linked citations | Days, once your Lexis account is enabled |
| Claude plus a prompt library | Solos and small firms | About $20/month per person | Whatever you paste or connect, plus your prompts | An afternoon |
My take
I own 10 law firms and I do not use Harvey.
My flagship handles lemon law in all 50 states. That practice is narrow and statute-driven. The same fact patterns, the same manufacturers, the same handful of statutes. What my firms need is not a generalist enterprise platform. It is a tested prompt library, plus custom tools built for our exact workflow, plus the research tools we already pay for.
I built a document review system for lemon law files, an intake scoring tool, and a dashboard across all 10 firms. Owned, not rented. For my economics that beats an enterprise contract, and it would beat one at any price, because the fit is wrong regardless of the number on the quote.
That is the whole lesson. The most expensive tool is not the best tool. The best-fitting one is. A firm doing bet-the-company M&A across four continents has a genuine problem that Harvey is designed to solve. A 12-lawyer plaintiff shop does not have that problem, and buying the solution to a problem you do not have is how firms end up with a $60,000 line item nobody logs into.
How to evaluate any enterprise legal AI
If you are taking the sales call, take it with these five questions written down:
- How does the system handle hallucination? Not “is it accurate,” which every vendor answers the same way. Ask what happens when the source material does not contain the answer, and ask to see it happen live.
- What is the output grounded in? Your documents, a licensed research database, open model knowledge, or a mix? Make them be specific per feature.
- Will you do a paid 90-day pilot with a real practice group instead of a year commitment? A vendor confident in adoption says yes.
- How is our data handled? Is it used for training, is it segregated, where does it live, who at the vendor can see it?
- What happens to our data and our configured workflows if we cancel? Get the export path in writing before you sign, not after.
The answers to 3 and 5 tell you more about the vendor than the product demo does.
Do this today
Skip the demo request. Open a blank document and write down the five tasks your firm repeats most, with rough hours per week next to each. If the top task is drafting and the total is under 20 hours a week across the firm, you have a prompt library problem, not a platform problem. Fix that first for $20 and revisit enterprise tooling when your headcount actually justifies it.