Most law firms that evaluate Harvey AI are not Harvey AI's target customer. That's not a knock on those firms — it's a structural reality about how Harvey is built, priced, and sold. Harvey's go-to-market runs through BigLaw managing partners and enterprise procurement, not practice-management software marketplaces. Its deployment model assumes dedicated IT and LegalOps support. Its pricing reflects that reality.
The problem is that vendor demos don't make these filters obvious. You can schedule a Harvey demo from a 25-attorney firm and get a polished walkthrough. What you won't get is a straight answer on whether your firm is actually in the addressable market. This framework closes that gap.
The Firm Profile Harvey AI Is Built For
Harvey's documented customer base — A&O Shearman, PwC Legal, Cleary Gottlieb, Davis Polk — defines the profile better than any marketing page. These are Am Law 100 or equivalent firms running high-volume transactional work across multiple jurisdictions. They have LegalOps teams. They have IT security functions capable of reviewing vendor data handling agreements. They have associates who can absorb a 6–12 hour training investment per workflow.
The common thread isn't firm size measured in headcount. It's operational maturity. Harvey requires a deployment partner inside the firm — someone who owns integrations, manages user provisioning, and tracks adoption. Without that internal owner, implementations stall at the 60-day mark regardless of how good the demo was.
If your firm doesn't have someone whose partial job is managing enterprise software deployments, Harvey is going to create a new job rather than automate an existing one.
The Four Questions to Answer Before Evaluating Harvey
Before entering any Harvey sales process, answer these four questions with specifics, not aspiration.
First: Do you have a dedicated LegalOps or IT resource to own the implementation? Not a partner who's interested in AI. Not an associate who's good with technology. A named person with time allocated to implementation work.
Second: Is your primary work high-volume transactional? M&A due diligence, contract review, and document-heavy regulatory matters are Harvey's native use cases. Litigation, family law, estate planning, and advisory practices are further from the product's core. The ROI calculation changes significantly depending on where your volume sits.
Third: Can your budget absorb the full cost — not just the license? See the implementation cost framework for the full breakdown. The overhead layer — IT security review, integration, training — typically runs 20–40% on top of the license. That's a real number that needs to be in the budget before the evaluation starts, not discovered after.
Fourth: Are your clients or management already asking about AI at the platform level? If the answer is yes, part of Harvey's value is signaling — being able to say you run Harvey in the same sentence as Cleary and A&O. That's a legitimate consideration, but it's a different justification than pure workflow efficiency, and it should be named as such.
Three or four yes answers means Harvey is worth a serious evaluation. Fewer than three means the evaluation cost — in time, distraction, and partner attention — likely exceeds the expected benefit.
How to Run a Bounded Harvey AI Pilot Without Overcommitting
If you clear the four-question filter, structure the pilot before you start it. A bounded pilot has three fixed constraints: use case, user set, and time window.
The use case should be specific enough to measure. "Due diligence checklist generation for M&A matters under $50M" is a use case. "Helping associates work more efficiently" is not. The narrower the use case, the cleaner the signal from the pilot data.
The user set should be 5–10 attorneys — large enough to get meaningful usage data, small enough that you're not running a firm-wide rollout before you know if it works. Avoid including partners who won't use it regularly; their low-engagement data will pollute the results.
The time window should be 60–90 days with a hard evaluation date. Define success metrics before day one: hours saved per matter, accuracy rate on issue spotting, or associate satisfaction on a defined scale. Pilots that don't have pre-committed success criteria almost always end in expansion regardless of actual performance — the sunk cost dynamic kicks in, and no one wants to be the person who killed the AI pilot.
Build in a genuine off-ramp. If the metrics don't hit, the firm should be positioned to walk. That means not letting the pilot expand scope mid-flight, not adding users before the evaluation date, and not letting Harvey's sales team reframe the metrics when results come in below target.
What Firms Outside the Target Profile Use Instead
The alternatives aren't consolation prizes. For firms outside Harvey's profile, several of them are better fits by design.
Spellbook is contract review built on GPT-4, integrated directly into Microsoft Word. The workflow is familiar, the onboarding is measured in hours rather than months, and the pricing is quote-only with a 7-day free trial. For firms where contract review is the primary AI use case, this is worth evaluating before anything with an enterprise sales cycle.
CoCounsel covers a broader range of legal workflows. Third-party pricing benchmarks report $75–$500 per user per month depending on tier. It competes with Harvey at the high end and positions itself as a more accessible enterprise option.
Claude API is the build-your-own path. For firms with developer resources — or access to a developer — pay-per-token pricing with no seat minimums makes this the most flexible option. The full cost analysis is in the Harvey vs. build-your-own comparison. The trade-off is build time and maintenance; the upside is no lock-in and complete control over workflows.
Clio Duo and MyCase Pro embed AI directly into practice management. MyCase Pro is publicly priced at $89/user/month. No separate implementation layer, no IT security review, no integration project — the AI is already inside the software the firm uses daily. For smaller firms where practice management integration matters more than purpose-built legal AI depth, this is often the fastest path to actual usage.
A Decision Framework for Mid-Market Firms Considering Harvey
If you're a mid-market firm — roughly 15–100 attorneys — and you're serious about AI, the decision isn't really "Harvey or nothing." It's "Harvey or what, and when."
The most common mistake is treating Harvey as the default serious option and everything else as the fallback for firms that can't afford it. That framing is wrong. Harvey is the right choice for a specific operational profile. Other tools are the right choice for different profiles. The question is which profile matches your firm.
Start with your highest-volume, most time-intensive workflow. Map out what a 20% time reduction on that workflow would be worth in billable hours or associate capacity. Then map out which tool can deliver that reduction with the least implementation friction given your current IT and LegalOps maturity.
For most mid-market firms, that analysis doesn't land on Harvey. It lands on Spellbook, CoCounsel, or a practice-management-embedded option. That's not a failure of ambition. That's correct prioritization. See the full cost structure analysis for the numbers that back this up.
Harvey makes sense if you're running BigLaw-scale transactional work and your IT and LegalOps teams have bandwidth for an enterprise rollout. If you're not at that scale or don't have that internal capacity, Harvey is a product you'll spend six months evaluating and then not deploy effectively. The alternatives — Spellbook, CoCounsel, Claude API, Clio Duo — aren't compromises. They're the right tools for a different operational context.