“Legaltech” is the umbrella category for any technology used in the practice or business of law — CLM, eDiscovery, matter management, legal research platforms, e-billing, practice management, and more. “Legal AI” is the subset of legaltech built on machine learning and, since 2023, generative and agentic AI applied to legal tasks.
Legal AI is not a purchasing category. That is the distinction most buyers get wrong. Legaltech is a set of systems of record you buy one per function; legal AI is a capability layer that now runs through all of them, and increasingly through the general-purpose AI your company already licenses. It is also not a new-vendors-versus-old-vendors split: by 2026 the incumbents ship agents and the AI natives ship repositories. If you are drawing up a budget with a “legaltech” line and a separate “legal AI” line, you will double-buy.
The legaltech landscape
Legaltech historically organized into seven categories, each with its own vendor set, sales cycle, and integration pattern:
| Category | Examples | Primary user |
|---|---|---|
| Contract management | Ironclad, Agiloft, Sirion, Concord | In-house Legal Ops |
| eDiscovery | Relativity, Everlaw, DISCO, Logikcull | Litigation teams, in-house and firm |
| Practice management | Clio, MyCase, Filevine | Solo and small firms |
| Legal research | Westlaw, Lexis+ with Protégé, Bloomberg Law, vLex | All practicing attorneys |
| Matter and spend management | Onit, Mitratech, BusyLamp, Brightflag | In-house Legal Ops |
| Document production | Litera, iManage, NetDocuments | Mid-to-large firms |
| Court filing | One Legal, File & ServeXpress, ECF systems | Litigation teams |
Legal AI as a layer
Legal AI organizes by capability rather than by category:
- Drafting AI. Harvey, Spellbook, Legora — drafting contracts, briefs, and memos.
- Review AI. Ivo, Luminance, BlackBoiler, LegalOn — reviewing inbound contracts against a playbook. Luminance will negotiate a standard NDA end to end; BlackBoiler returns a finished redline rather than a report about one.
- Research AI. Thomson Reuters CoCounsel, Lexis+ with Protégé — research with citation validation against primary law.
- eDiscovery AI. Relativity aiR, Everlaw AI, DISCO Cecilia, Reveal ASK — privilege review, document classification, case analysis.
- Knowledge management AI. Litera Foundation and its Lito agent, or direct Claude Skills — knowledge retrieval against the firm corpus.
- General-purpose AI. Claude, ChatGPT under enterprise terms — used across every category above.
How the categories converge
Four patterns, the fourth of which is new since 2025:
- Legaltech adds AI. Ironclad ships Ironclad AI; Relativity ships aiR; DISCO folded Cecilia and its agentic reviewer into an all-inclusive platform at no extra list charge. The legaltech vendor becomes a legal-AI vendor by extension.
- Legal AI broadens scope. Harvey started as drafting AI and now spans research, review, and document analysis. Its June 2025 alliance with LexisNexis put Shepard’s Citations and primary law inside the product, so the drafting tool became a research tool without building a research corpus.
- General-purpose AI enters legal. Claude and ChatGPT with custom Skills displace specialized tools for generalizable work.
- The AI layer gets absorbed into the system of record. Litera relaunched in July 2026 around a single agent, Lito, running on one dataset across its whole portfolio rather than per-product AI features. LexisNexis retired the Lexis+ AI name in February 2026 and shipped Lexis+ with Protégé, adding an orchestration layer, drafting agents, and shared workrooms in May. The standalone AI feature stops being a SKU and becomes the interface.
What the 2025 shakeout proved
Two vendors that this page named as live options in earlier revisions no longer exist as things you can buy, and both died the same way — absorbed, not outcompeted.
- Casetext was acquired by Thomson Reuters in 2023 for $650M and retired as a standalone product on April 1, 2025. CoCounsel Legal relaunched that August inside the Westlaw ecosystem, sold per attorney. Former self-serve subscribers on small-firm rates were migrated onto enterprise-shaped contracts.
- LawGeex had its enterprise contract-review business dismantled in 2023: technology assets went to Robin AI, the client base to LegalSifter, the founders to Superlegal. Robin AI itself then collapsed in late 2025.
The lesson for a buyer is not “avoid startups.” It is that a standalone AI capability with no system of record underneath it is the most acquirable asset in this market, and acquisition usually means repricing. Weight that when you sign a multi-year deal with a point AI tool.
Specialists vs general-purpose AI
| Use case | Legal-AI specialist | General-purpose AI + Skills |
|---|---|---|
| Highest-bar drafting (M&A, complex commercial) | Harvey, Spellbook | Borderline; needs a fine-tuned playbook |
| Routine NDA review | Ivo, BlackBoiler | Claude + contract redline Skill |
| Legal research with citations | Thomson Reuters CoCounsel, Lexis+ with Protégé | Not viable — you need verified primary sources |
| Knowledge retrieval from firm corpus | Litera Foundation | Claude + custom Skills against the DMS |
| First-pass eDiscovery review | Relativity aiR, Everlaw AI | Not viable — needs defensible production-grade scale |
| Generic summarization, drafting, analysis | Specialist overkill | Claude is the right answer |
Specialists win when the data, the workflow, or the integration is legal-specific. General-purpose wins when the task generalizes and the data already flows in.
How to think about the budget
Most in-house legal AI budgets in 2026 have three line items:
- Enterprise general-purpose AI. Claude Enterprise or equivalent, at predictable per-seat pricing, covering the broad use cases.
- One or two legal-AI specialists. Typically Harvey or Spellbook for drafting, plus whatever AI is already bundled in your CLM.
- Specialty AI where volume justifies it. Research AI for research-heavy practices; eDiscovery AI inside the matter platform when discovery recurs.
The over-buying pattern is one specialist per category. The under-buying pattern is trying to run everything through general-purpose AI alone.
Watch-outs
- Paying twice for the same capability. Your CLM’s bundled AI and your standalone review tool overlap on playbook redlining. Guard: before renewing either, run the same ten contracts through both and keep the one with the better redline acceptance rate.
- “Included” AI that meters. Relativity bundles aiR into its platform but charges per document processed; DISCO’s all-inclusive bundle does not. Guard: ask for the meter, not the list price, and model it against last year’s document volume.
- Buying an AI feature as if it were a platform. The acquired-and-repriced pattern above is the failure mode. Guard: cap point-AI commitments at 12 months until the vendor has a system of record you would keep anyway.
- Assuming the research tool you licensed still has that name. Lexis+ AI, Casetext, and LawGeex all changed hands or names inside 24 months. Guard: re-verify your legal-AI vendor list at each renewal cycle rather than at contract end.
Related
- AI policy for legal teams — governs which tools are authorized
- Legal Ops maturity model — describes when AI investment compounds
- What is Legal Ops? — the function that owns legaltech vs legal-AI strategy
- Best AI tools for legal ops — head-to-head comparison