Short answer: Harvey, CoCounsel and Lexis+ Protégé all process privileged work on the vendor’s cloud and price per seat, with Harvey requiring a 20–50 seat minimum on annual terms. A private, self-hosted legal AI deployment runs the same research, drafting and review workflows inside the firm’s own tenant, so matter data never leaves the firm.
Three legal AI platform demos blur together fast. The separator is the part the demo skips: where your matter data is processed.
Legal AI platforms apply artificial intelligence (AI) across a firm’s work; research, drafting, review, and matter management in one integrated product. The questions that matter are which workflows they automate, how the platforms differ, and where privileged work is processed. This guide covers all three. Firms weighing the private route can start with our overview of private, self-hosted AI for law firms.
What workflows do Legal AI Platforms automate?
The leading platforms target the same core workflows — the appeal is having them in one place rather than four tools.
Legal research. Cited answers and memos from a single integrated product, so research happens where the rest of the work does. Outputs still need a lawyer’s verification.
Drafting. Clauses, correspondence, and filings generated from a prompt or template, ready for the lawyer to refine.
Document review. Contract, discovery, and due-diligence review built into the platform rather than a separate tool.
Knowledge and matter management. Search and Q&A over the firm’s own documents, turning the matter archive into an answerable resource.
Where today’s Legal AI Platforms fit?
The four options below run the same core workflows, so capability is rarely the deciding factor. What separates them is where privileged work is processed and what each one costs to run at your firm’s size. Harvey, CoCounsel, and Lexis+ Protégé all process matter data on the vendor’s cloud and charge per seat; the ranges differ, but the deployment model does not.
A self-hosted stack is the exception as it runs the same research, drafting, and review inside the firm’s own tenant.
Read the table with the last column first: for a confidentiality-bound firm, where the data goes decides more than any feature line.
Here’s a table showing how the four platforms stack up on the workflows that matter.
| Platform | Best for | Reported pricing (2026) | Case law source | Where data is processed |
|---|---|---|---|---|
| Harvey | Large-firm multi-step workflows | No public pricing; est. ~$1,000–$2,000/seat/mo, 20–50 seat minimum, annual contracts $50K–$300K+ | LexisNexis integration (add-on) | Vendor cloud |
| CoCounsel | Task skills: review, memos, depo prep | Core from ~$225/seat/mo; case law search needs Westlaw on top; no seat minimum | Westlaw | Vendor cloud |
| Lexis+ Protégé | Research with an assistant, 300+ workflows | Quote-only; bundled with the Lexis+ subscription | LexisNexis | Vendor cloud |
| Self-hosted stack | Private models over firm documents | Build/run cost, not per-seat; scales with infrastructure, not headcount | Firm’s own sources | Stays in the firm |
Pricing is quoted per seat unless noted. Harvey does not publish rates; figures are third-party estimates and vary by firm size, term, and add-ons. Confirm current pricing with each vendor.
On capability the cloud platforms are close; they diverge on deployment — all process privileged work on the vendor’s cloud and price per seat.
The same workflows run on managed platforms or on a private stack that keeps privileged work in the firm.
Legal AI Platforms — Two Ways to Run Them
The same workflows — research, drafting, review, knowledge — run on either model.
Managed platforms
Harvey · CoCounsel · Lexis+ Protégé
Vendor-run models on the provider’s cloud.
Privileged work leaves the firm
Private, self-hosted
Private models + firm documents
Runs inside the firm’s own tenant.
Privileged work stays in the firm
What the demo skips?
The things that rarely lead a platform pitch — and they decide the purchase.
Where your data is processed. The capability looks similar across platforms; the deployment model, vendor cloud versus your tenant is the real differentiator, and it rarely leads the demo.
Verification overhead. Every platform’s output needs lawyer review, so ask how much, not whether.
Lock-in and cost at scale. Per-seat pricing and proprietary formats compound as the firm grows so weigh switching cost before committing.
The Private, Self-Hosted Alternative
A self-hosted platform, private models and an orchestrator over the firm’s documents, inside its tenant runs the same research, drafting, and review while keeping privileged work in-house and consolidating point tools into one stack.
It meets the confidentiality duty ABA Opinion 512 places on the firm. For a tool-level survey, see our companion guide to the best legal AI tools for lawyers and law firms.
How to evaluate a Legal AI Platform?
Ask where data is processed. Make the deployment model the first question, not the last.
Test on your own work. Pilot with real, appropriately handled matters to see verification load and fit before committing.
Model the cost at scale. Compare per-seat cloud pricing against a consolidated private build as headcount grows.
Want a legal AI platform that keeps privileged work in your firm?
Which Legal AI Platform Should Your Firm Choose?
The four platforms converge on what they do and split on where they do it and for a confidentiality-bound firm, that split is the decision.
Harvey, CoCounsel, and Lexis+ Protégé deliver strong research, drafting, and review, but all three process privileged work on the vendor’s cloud and price per seat, from CoCounsel’s ~$225 entry point to Harvey’s six-figure annual floor.
A private, self-hosted stack runs the same workflows inside the firm’s own tenant, so matter data never leaves and cost scales with infrastructure rather than headcount.
So, which legal AI platform should your firm choose? The right answer depends on firm size, existing research stack, and how much of your work is privilege-sensitive. If keeping client data in-house is non-negotiable, the private route is built for that from the start; if you already live in Westlaw or Lexis, the managed option that matches your stack is the shorter path. A short scoping conversation will match the workflow load to the deployment model that fits. A short scoping conversation will match you to the right Legal AI platform.