AI Tools for Law Firms: What Practices Are Actually Buying

The legal profession, historically the slowest large industry to adopt new technology, has just produced one of the fastest technology shifts in its history. The 8am 2026 Legal Industry Report, surveying more than 1,300 legal professionals, found that 69 percent now personally use general purpose AI tools for work, more than double the 31 percent of a year earlier, with legal specific tool usage doubling to 42 percent over the same period. Clio's 2026 survey shows solo practitioners at 71 percent adoption and small firms at 75 percent. Wolters Kluwer's Future Ready Lawyer research puts overall use of at least one AI tool at 92 percent of legal professionals, with a majority reporting weekly time savings in the 6 to 20 percent range.

The institution has not kept up with its members. The same 8am research found only about a third of firms have formally adopted AI, 43 percent have no AI policy and no plans to create one, and 54 percent provide no training on responsible use. Thomson Reuters reported legal technology spending grew 9.7 percent in 2025, likely the fastest real growth the industry has recorded, while most firms collect no return on investment data on that spending at all.

The Six Categories Firms Are Buying

1. General purpose assistants

The most used legal AI tools are not legal tools at all. ChatGPT, Claude, and Microsoft Copilot dominate actual usage because they are cheap, familiar, and genuinely capable at drafting, summarization, and first pass analysis. The business tiers of these platforms, compared in this publication's analysis of ChatGPT Business and Claude Enterprise, run roughly $20 to $30 per user per month with no training on customer data by default, which for a firm is the minimum acceptable configuration: consumer accounts with default settings have no place in privileged work. The limitation is equally plain: general assistants have no citation grounding in legal databases, so anything resembling legal research requires verification against primary sources, a lesson the profession has now learned repeatedly through sanctions for fabricated citations.

2. Research platforms with AI: CoCounsel and Protégé

The research incumbents attached AI to the databases lawyers already trust. Thomson Reuters CoCounsel, built from the Casetext acquisition, sells standalone in the low hundreds per user per month, with third party pricing analysis placing the AI line near $225 to $400, and increasingly ships bundled with Westlaw, where the all in research plus AI seat lands in the mid hundreds monthly. LexisNexis's answer, Lexis+ with Protégé, is sold as an add-on atop a Lexis research subscription. For most established firms this category is the lowest friction serious purchase: the AI arrives inside a research contract the firm already holds, grounded in citable primary law, which directly addresses the hallucination risk that makes general assistants unsuitable for research.

3. Enterprise platforms: Harvey and Legora

At the top of the market sit the enterprise legal AI platforms, led by Harvey, which reports more than 235 legal customers including firms such as A&O Shearman and DLA Piper, and offers an agent builder with which large organizations encode their own workflows. Harvey publishes no pricing. Third party estimates through 2026 place seats at roughly $1,200 to $2,000 or more per month, with reported seat minimums in the 20 to 50 range, putting realistic annual commitments between the high five figures and several hundred thousand dollars before implementation. Competitor Legora is estimated in the $300 to $800 per seat range. These platforms are built for large firm economics: due diligence, large scale contract review, and litigation workflows where associate hours are the cost being displaced. Below perhaps 50 lawyers, the seat minimums alone usually settle the question.

4. Word native drafting and contract tools

A distinct category lives inside Microsoft Word, where transactional lawyers actually work: tools such as Spellbook for drafting and inline redlining, priced in the $99 to $400 per user per month band typical of the small firm friendly segment. For contract heavy practices, these tools deliver the most direct value per dollar in the market, because they automate the precise motion, mark up this agreement against our positions, that consumes transactional time.

5. Practice management embedded AI

For solo and small firms, the most sensible entry point is frequently the AI already arriving inside the practice management platform the firm runs on: Clio's and MyCase's embedded assistants and their competitors. The economics and governance are both favorable: modest add-on pricing, and the AI operates inside a system that already holds the firm's matters under existing access controls, rather than in a separate tool where documents must travel.

6. Litigation and eDiscovery AI

eDiscovery platforms such as Relativity and Everlaw have shipped generative AI review features into a category that has used machine learning for a decade under judicial supervision. For litigation practices, this is the most mature and defensible AI category in law, with review cost per document the measurable outcome, and it now has federal support: a 2026 discovery order treated generative AI responsiveness review as ordinary Technology Assisted Review, examined in this publication’s analysis of the LinkedIn decision. Pricing is volume based and negotiated, as contract lifecycle platforms such as Ironclad also are, with deal sizes reported from tens of thousands of dollars annually upward.

The Pricing Picture in One Table

Category	Representative Products	Indicative Pricing
General assistants (business tiers)	ChatGPT Business, Claude Team, Copilot	$20 to $30 per user per month
Research with AI	CoCounsel, Lexis+ Protégé	Roughly $225 to $400 for the AI line; mid hundreds all in with the research bundle
Enterprise platforms	Harvey, Legora	Estimated $1,200 to $2,000+ per seat (Harvey), $300 to $800 (Legora); substantial seat minimums; no published pricing
Word native drafting	Spellbook and peers	Roughly $99 to $400 per user per month
Practice management AI	Clio Duo, MyCase AI	Add-on pricing atop existing subscriptions
eDiscovery and CLM	Relativity, Everlaw, Ironclad	Volume based, negotiated; annual contracts from tens of thousands

Enterprise figures above are third party estimates, not vendor rates, because the top of this market does not publish pricing. Treat every number as orientation and obtain quotes in writing.

The Governance Gap Is the Real Risk

The ethics framework is settled enough that no firm can plead novelty. ABA Formal Opinion 512 on generative AI applies the existing duties: competence requires understanding the tools well enough to use them responsibly, confidentiality constrains what client information may enter which systems under which terms, supervision covers AI output the way it covers junior work product, and fees must fairly reflect AI assisted efficiency. The now extensive public record of sanctions over fabricated citations makes the operational point vividly: unverified AI research filed with a court is a professional responsibility failure, not a technology failure.

Against that framework, the survey data describes a profession running ahead of its own controls: two thirds of lawyers using AI personally, while 43 percent of firms have no policy and only 9 percent have a written policy that is actually enforced. For firm management, the implication is direct. The first AI investment is not a tool. It is a one page policy stating which platforms are approved on which terms, what client data may and may not enter them, and what verification is required before AI output reaches a client or a court, followed by training that makes the policy real. That costs almost nothing and addresses the largest current exposure, which is the unsanctioned consumer grade usage already happening.

What This Means for Management

The evaluation path tracks firm size. A solo or small firm should start with the AI inside its practice management platform plus an approved general assistant on a business tier under a written policy, and add a Word native drafting tool if the practice is transactional. A mid size firm should evaluate the AI attached to whichever research incumbent it already pays, CoCounsel for Westlaw firms and Protégé for Lexis firms, since grounded research is where legal specific value is clearest. Large firms are the audience for Harvey class platforms and should evaluate them the way they evaluate any six figure system: on their own matters, against measured associate hour baselines, with procurement discipline about seat minimums and terms. Every firm, at every size, should have the policy and training in place before, not after, the tools.