Harvey seats vs an operating package

August 11, 2026 · Mike Rodgers

Harvey is very good at the thing it sells: executing legal work. It does not sell the thing that keeps a 40-attorney firm from running through the managing partner's head. Those are two different purchases, and only one of them has a public price.

Naming. Harvey is a product of an independent third party and is named here only to identify it. Rodgers Intelligence Group is not affiliated with, endorsed by, or a partner of that company. No price for its product is stated in this article, because none is published; see the sources below for the retrieval that establishes that.

Ask a managing partner at a 40-attorney firm what actually ate her week and you will not hear "legal research."

You will hear: the comp conversation she has postponed twice. The lateral who has now asked three times what the origination credit split would be. The matter that is forty percent over budget and nobody flagged it until the pre-bill. The conflicts question on the intake from Tuesday that required someone to remember a 2023 engagement. The associate who asked how the firm handles a specific fee arrangement and got three different answers from three partners.

None of that is legal work. All of it is the firm running through one person's head.

This is the distinction that gets lost when a legal-AI evaluation turns into a bake-off, and it is worth being precise about, because the two products are not competing. They are solving different problems, and a firm this size will eventually want both.

What Harvey actually sells

Read their site rather than the discourse about it. As of 2026-08-11, Harvey's platform navigation lists: Agents, Vault, Knowledge, Shared Spaces, Command Center, Contract Intelligence, Ecosystem, and Harvey Mobile. Their own Mid-Sized Firms page names the use cases as research, drafting, document review, due diligence, client communications, and knowledge management, under the headline "Win the Work You Deserve."

That is a serious, well-built product for executing legal work. Surface the case law. Draft the brief. Read ten thousand documents and pull the change-of-control clauses. Turn your best work into reusable precedent. The security posture is enterprise-grade — SOC 2 Type II, ISO 27001, ISO 27701, ISO 42001, GDPR, CCPA are all listed on that page.

If your bottleneck is associate leverage on document-heavy matters, this is the right category of purchase and you should evaluate it seriously.

Now look at the list again and notice what is not on it. Compensation. Origination policy. Matter profitability by partner. Staffing model. Intake and conflicts policy. Lateral integration. "What did we decide about client X's fee arrangement, when, and who approved it." Harvey never claimed those. It is not a gap in their product; it is the edge of their category.

The pricing section, done honestly

Here is where most competitive content cheats, so we are going to do the opposite.

We are not going to tell you what Harvey costs. On 2026-08-11 we requested harvey.ai/pricing and got a 404. There is no list price published on their site. The primary call to action across the site is "Request a Demo." Third-party figures circulate in legal-tech commentary, but we did not retrieve one from a source we would stake a claim on, and a number you cannot source is a number you should not print.

You may have seen the shorthand "less than two Harvey seats" used to position packages like ours. We are retiring it. It asserts a competitor's price that we cannot verify, and an unverifiable comparison is worth less than no comparison — the first prospect who has an actual quote in hand will catch it, and then everything else you said is suspect too.

What we will do instead: our packages (Professional, Executive Pro, and Team for five or more seats, each on a 12-month term) are priced after a short conversation, as one whole price with no per-seat meter, and the price is confirmed in writing in the agreement.

Run the comparison yourself, in four lines:

1. Get a written Harvey quote for the seat count you would actually deploy — not firm headcount, the number of people who will log in weekly.
2. Multiply by the contracted term. Add implementation, training, and any minimum commitment.
3. Put our whole price next to it.
4. Then ask the question that actually matters: are these two numbers buying the same job?

They are not. That is the point. One is priced per lawyer because it scales with legal work executed. The other is priced per install because it scales with the firm, not the headcount. Comparing them on dollars alone is how firms end up owning two things that do the same job and none of the thing they needed.

The mechanism: a firm operating record

An operating package does one thing. It holds the firm's own decisions, in the firm's own documents, and answers questions about them with a citation.

In a boutique that means, concretely:

• Matter playbooks. Not templates — the record of how this firm runs a matter type. Intake criteria, staffing pattern, budget shape, the escalation trigger, the three things that went wrong last time.
• Engagement and fee terms by client. What was agreed, when, in which engagement letter, and what the exception history looks like.
• Intake and conflicts policy as an answerable thing. Not a PDF in a shared drive that three partners interpret differently.
• The decision trail. What the partnership decided about comp, origination, or lateral terms, when, and on what reasoning. This is the single most expensive thing a firm loses when a partner leaves.
• Weekly operating cadence. Pipeline, WIP, realization, and the two or three questions that need a partner's yes — drafted before Monday, cited, ready for a decision instead of a meeting.

The governing rule is the same one we apply everywhere: no source, no number. If the system cannot cite the document an answer came from, it does not give you the answer. In a profession where the entire product is defensible judgment, an uncitable answer is worse than no answer.

Why the leverage math favours this

Clio's 2025 Legal Trends Report, in its online edition, reports a growth cohort whose revenue rose roughly four times faster than lawyer headcount — about a 25% increase in lawyers alongside roughly a 50% increase in clients and matters. Read that as an operations finding, not a technology one: the firms that pulled ahead did more work per lawyer and got more out of each matter and client — which is a statement about how the firm is run, not about how fast anyone drafts.

Seat-based legal AI moves the first term. It makes each lawyer's output larger. It does nothing about the second — the operating decisions that determine which matters you take, how you staff them, and whether they make money. That is the term an operating package moves, and it is the one that does not scale by hiring.

The seat-sprawl problem nobody budgets for

There is a second cost that shows up eighteen months in, and it is not on anyone's spreadsheet.

A firm this size does not buy one AI product. It buys a research tool because litigation asked, a contract tool because the transactional group asked, a note-taker because someone saw a demo, a document automation add-on because the practice management vendor bundled it, and a general-purpose chat subscription because the associates were already pasting client matter into a free one and IT wanted that stopped. Each purchase is defensible. The aggregate is five subscriptions, five renewal dates, five security reviews, five sets of admin, and — the part that actually hurts — five places where firm knowledge is now stored and none of them talk to each other.

The consolidation argument is not that one tool does everything. It does not. The argument is that you should be deliberate about which layer you are willing to fragment. Fragmenting execution tools is survivable; different practice groups genuinely need different instruments. Fragmenting the firm's operating record is not, because the whole value of that record is that there is one of it.

What you get, and what to read first

• Professional — priced after a short conversation A named seat with persistent memory, installed on hardware you control, with a curated vault where every answer is cited. Role skills installed: matter and meeting prep from your calendar, KPI review, decision memos, competitor watch, month-end pack, inbox and follow-up triage. Two support hours a month included.
• Executive Pro — priced after a short conversation. Adds the compliant Microsoft 365 lane running inside your firm tenant, ingestion of your own engagement letters, playbooks and reports with citations, ten custom skills built for your workflows, monthly knowledge refresh, and a quarterly strategy session.
• Team — priced after a short conversation. Five or more seats across the management committee with one shared knowledge plane and written governance on what each seat may see and do.

The packages page also describes a 90-day performance out and a written Intent Contract. Treat those as contract terms to be read, negotiated, and reviewed by your own counsel — not as a headline. You are lawyers. You will anyway.

Start here: the free 30-minute assessment. Bring one firm question you have answered inconsistently in the last month. We will show you what a cited answer to it looks like, on your real constraints.

→ Professional AI Packages · Law vertical · AI for law firms · How governed agents work

Four questions before you sign anything

1. Per-seat or per-firm? If the price scales with lawyers, you are buying capacity. If it scales with the install, you are buying an operating layer. Know which one you are short of.
2. Whose documents does it read? A model trained on the corpus of American law does not know your engagement letters.
3. Can it cite? Ask for a sourced answer to a question about your own file, in the demo, on your data.
4. What is the whole twelve-month number? Setup plus monthly plus implementation plus training. Then put both vendors' whole numbers side by side, and ask again whether they are buying the same job.

Mike Rodgers is the founder of Rodgers Intelligence Group. He builds the systems, agents, and operating procedures that let a one-person company run like a fleet. Learn more →