Legal AI: the model layer
and everyone building on it

Firm-wide assistant for research, drafting, diligence and workflow automation, sold to large firms and enterprise legal departments.

Last round$200M co-led by GIC and Sequoia, Mar 2026

Model useOriginally OpenAI-exclusive (an OpenAI Startup Fund portfolio company from Dec 2022). Opened to Anthropic and Google models in May 2025; users can pick a model or let Harvey auto-route. Publishing its own model research (Tenet) and moving toward a proprietary legal model.

TractionReported 40+ Am Law 100 firms; publishes BigLaw Bench

Collaborative legal AI platform. Best known for Tabular Review — turning a folder of contracts into an interactive grid for clause comparison and extraction at scale — plus agentic workflows spanning drafting, review, research and translation, and a client portal for firm-client work.

Last round$550M Series D led by Accel at $5.55B (Mar 10 2026), extended +$50M in Apr 2026 with Atlassian and NVentures (NVIDIA)

HistoryFounded 2023 in Stockholm as Leya, rebranded to Legora Feb 2025. Y Combinator W24; $80M Series B (May 2025, General Catalyst and Iconiq); $150M Series C (Oct 2025) at $1.8B; $550M Series D (Mar 2026, Accel) at $5.55B; +$50M extension (Apr 2026) at $5.6B. Founders Max Junestrand (CEO), Sigge Labor, August Erséus.

Model useAnthropic publishes Legora as a Claude customer, using Claude across assistant, document review and workflows. Also an OpenAI API customer with Astra for Law access, and now inside Google's Gemini Enterprise for Legal. Legora is in all three model ecosystems — the most complete hedge in the market.

TractionReported $100M ARR by April 2026, roughly 18 months in. 1,000+ law firms and in-house teams across 50+ countries. US buildout: NY (11 Madison Ave, ~100k sq ft), Denver, Houston, Chicago; 300+ US staff targeted by end of 2026.

Agentic legal assistant across research, analysis, drafting and review, preserving matter context between steps. Expanded agentic capabilities announced 2026-08-24, four days after CoCounsel's next-gen GA.

Model useHistorically multi-model across Anthropic, OpenAI and Mistral; routing not publicly disclosed.

TractionThe other half of the research duopoly; competing head-on with CoCounsel for the layer where lawyers finish work

Caveatmodel attribution: low

Agentic legal platform grounded in Westlaw primary law and Practical Law, taking a matter from research and issue analysis to work product in one workflow. Next generation reached GA 2026-08-20.

Model useNext-gen CoCounsel Legal is built on Anthropic's Claude Agent SDK — it plans, selects tools, retrieves authoritative content and adapts mid-workflow.

TractionDistribution is the moat: bundled with Westlaw and Practical Law, already the default research stack in most large US firms

Drafting-assistant layer inside Word — definition and cross-reference navigation, now extended with AI review.

E-signature plus Intelligent Agreement Management — the execution and obligation layer.

Enterprise CLM for the Fortune 500 — obligation management at procurement scale.

Caveatvaluation: dated

CLM for in-house teams — workflow, repository, AI review and playbook enforcement.

Last round$333M total raised

Why it mattersCo-founder Jason Boehmig now leads OpenAI's legal effort — a direct pipeline from the CLM incumbent into the model layer

AI contract review tuned to an in-house team's own positions and playbook.

Contract review grounded in attorney-authored playbooks and standards; strong in Japan, expanding in the US.

CLM and contract analytics for in-house legal.

Contract analysis and negotiation across the contract lifecycle; deep roots in diligence-scale document review.

Last round~$115M total raised

Model useRuns its own legal-domain models alongside third-party LLMs.

High-volume routine contract negotiation for private fund managers and investment banks — NDAs above all. Sold two ways: Accord as in-house software, and Contract Automation as a fully outsourced service pairing AI with Ontra's global network of lawyers.

Why it mattersThe clearest example of the hybrid that isn't a law firm: it does the negotiation, not just the tooling, and reports an early adopter saving 67% of time per contract.

Contract review plus lightweight CLM for mid-market teams with recurring contract types.

Model useLong-standing public Claude user; among the earliest Anthropic legal customers.

Contract drafting and review inside Microsoft Word, aimed at small and mid-size firms and lean in-house teams.

Last round$40M debt facility from RBCx (Mar 2026), earmarked for acquisitions

Model useHistorically GPT-based; now multi-model.

Why it mattersWord-native positioning is now directly contested by both ChatGPT for Word and Claude's native M365 integration

CLM for high-velocity in-house teams, with AI review and self-serve business workflows.

Contract workflow that meets the business where it works — Slack, Teams, Word — rather than in a legal portal.

Managed review and legal services at scale — the human layer under e-discovery, now automating itself.

End-to-end plaintiff-side case platform — intake, case evaluation, demand drafting.

Model useReported as built on Claude; named Claude for Legal partner.

Personal injury automation — medical record review, damages modelling and demand package generation for plaintiff firms.

Model useNamed by Anthropic as a Claude for Legal customer.

TractionPart of the plaintiff-side cohort that minted three unicorns in autumn 2025

Cloud e-discovery and litigation platform with generative review, chronologies and storybuilding.

Caveatvaluation: low

Case management for plaintiff and contingency firms, with AI document and demand tooling layered on.

Caveatvaluation: reported estimate

Dominant e-discovery platform (RelativityOne), with aiR for Review and related generative layers.

TractionIncumbent with the corpus — review data is the asset everyone else lacks

Free AI-summarised case law search, aimed at access to justice.

Open primary law and PACER archive. Quietly one of the most consequential entities in the market — OpenAI built the Astra for Law legal search index using CourtListener data.

Legal research and data retrieval connector.

Caveatthin public information

AI legal research built around verifiable citation to primary sources — litigator-first research workflow.

State trial court records and judge analytics — the layer neither Westlaw nor Lexis covers well.

AI assistant purpose-built for in-house generalist counsel — the one-lawyer-department use case.

Model usePublicly Claude-forward; publishes analysis of the Claude legal plugins.

Converts regulation into executable code — regulatory compliance agents that check actions against rules pre-emptively. Heaviest traction in financial services.

The 'legal front door' — AI agents that intercept and handle inbound business requests to in-house legal before they reach a lawyer.

Last round$70M Series B (Jun 2026); Series A led by Index Ventures

Patent drafting assistant embedded in the attorney's Word workflow.

AI copilot for patent drafting and prosecution.

Model useReported as built on Claude; named Claude for Legal partner with an MCP connector.

Cloud-native entity management: corporate records, minute books, cap tables, filings and real-time org charts. Used by corporate paralegals, law clerks, corporate secretaries and legal ops.

Why it mattersThe modern default for the corporate secretarial workstream — minute books, registers, filings and org charts as software rather than as billable paralegal hours.

Registered agent, entity management and filings at enterprise scale, tying minute book and resolution workflows to the entity record. Software plus a managed service layer.

Why it mattersThe incumbent that already does corp sec as a service globally. Competing on corp sec means competing with CSC's filing footprint, not just with software.

Cap table, equity and fund administration, now positioned as an agentic ERP for private capital — and, since acquiring Avantia, the owner of a law firm.

Why it mattersWent from adjacent SaaS to legal provider by acquisition. The pattern to watch: platforms that already hold the corporate record adding the legal layer on top.

Automates incorporation and ongoing corporate paperwork for venture-track Delaware C-corps: formation, SAFEs, board consents, equity grants and the long tail of routine corporate documents.

Why it mattersHas automated startup corporate paperwork end to end, from formation onward, for over a decade — as document automation, not advice, which is how it avoids practising law.

Enterprise entity management and governance — centralises the corporate record as a single source of truth, tied to board and compliance tooling.

Formation plus the operational wrap — registered address, mailroom, bookkeeping, tax and compliance from one vendor.

AI-embedded legal entity management for multinationals: formation, registers, ownership, officers, subsidiary governance, annual filings and approvals, with local know-how modules across 100+ jurisdictions.

Last roundAcquired by Quantios (2026)

Why it mattersProven across 100+ multinationals in 120+ countries — the closest existing product to automated corp sec at global scale, and it is software, not a firm.

Cap table and equity management for startups; integrates with Clerky so incorporation flows straight into a cap table.

Corporate services at global scale — entity formation, corporate secretarial, accounting and compliance across jurisdictions, with Blueprint as the workflow layer.

Why it mattersWith TMF and CSC, one of the incumbents that has offered end-to-end corporate secretarial work across jurisdictions at fixed prices for decades — without AI and without being a law firm.

Content cloud used as a deal room and matter repository.

Virtual data rooms for M&A — where diligence documents actually live.

Knowledge search across a firm's own work product — semantic retrieval over the DMS, permissions intact.

Why it mattersSits exactly on Helen Fan's Level 3: the firm's proprietary data layer

Matrix — document-set analysis at scale. Horizontal, but heavily used for diligence and legal/finance document work.

Caveatvaluation: dated

Firm operations — intake, conflicts, ethical walls, pricing, time. The compliance plumbing AI has to respect.

Cloud DMS and the other half of the firm-knowledge gate, strong in mid-market firms.

The document management system most large firms run on — and therefore the gate to firm knowledge for every AI vendor.

Why it mattersConnected to all three model platforms. Whoever controls the DMS connector controls context.

Practice management for small and mid-size firms — intake, matters, billing, payments — now with AI layered across. Acquired vLex (~$1B, 2025), buying its way into legal research.

TractionOwns the long tail of the profession — the segment BigLaw-focused vendors don't serve

Startup, venture capital and M&A firm serving founders, investors and tech companies on cross-border US–Asia matters.

Why it mattersOne of very few in this category built cross-border from the start rather than US-first with expansion later.

AI-native firm built around 'AI coworkers' for legal teams — founded by a former Paul Hastings corporate lawyer, with AI handling research, drafting and analysis as the first pass.

Why it mattersExplicitly testing whether AI can replace the associate pyramid rather than assist it

AI-native legal services for startups: AI generates first drafts, lawyers do final revisions, flat fee per document rather than hourly. Claims ~80% reduction in lawyer work and frequent same-day turnaround.

Last roundY Combinator W26

Traction50+ venture-backed startups across the US and Europe. Lawyers drawn from Cooley, Goodwin and Fenwick; founder was outside counsel to OpenAI.

AI-native legal and compliance for asset managers, now a division of Carta. Runs an AI workflow engine (Ava) folded into Carta's ERP so fund legal and compliance sit next to fund operations.

Last roundAcquired by Carta, announced 2026-05-13

Traction200+ asset managers including 30% of the world's largest funds, across ~$15T AUM. Carta's fourth acquisition since Oct 2025.

Why it mattersThe single most important structural event for this question: a SaaS platform bought a law firm rather than partnering with one, explicitly to remove the need to instruct outside counsel for routine work. The SaaS/law-firm boundary is not holding.

An actual law firm running on AI: high-volume contract review with licensed attorney sign-off in a median of 58 minutes, priced per document rather than per hour.

Last round$60M Series B, Apr 2026

Model useNamed Claude for Legal partner and named as using Claude on live matters.

TractionClients include Cursor and Clay

Augmented legal for enterprise departments — AI plus its own legal professionals delivering contracting, litigation support and compliance work as a service.

Last round$105M+ total raised

Why it mattersAcquired traditional ALSP capacity to pair with the model layer — the services-roll-up pattern

The original partner-only distributed firm — 'Law Firm 2.0' — founded two decades before the AI wave. No associates, no offices, partners keep most of their billings.

Why it mattersProof the senior-only economic model works without AI at all. AI makes it cheaper to run, it did not make it possible. Pierson Ferdinand's founders and 130+ of its launch partners came from here.

Small-claims and debt recovery run end-to-end by AI. The first fully AI-driven law firm authorised by the UK Solicitors Regulation Authority.

Why it mattersThe regulatory precedent matters more than the product — SRA authorisation is the template others will cite

AI-native firm for startups, staffed almost entirely by Big Law alumni with AI doing first-pass work.

Last round$11.5M raised; Y Combinator alum

Traction14 full-time lawyers; ~$2M annualised revenue

AI-driven personal injury and mass tort — data-led case identification and valuation.

Why it mattersApproved by the Arizona Supreme Court in July 2025 as the first AI-native PI/mass tort firm under an ABS licence — the US regulatory precedent that Garfield is to the UK

Consumer legal services delivered by AI plus a network of solicitors, at fixed prices.

Last roundGoogle-backed

Why it mattersAcquired Woodstock Legal Services (Sep 2025) — believed to be the first acquisition of a traditional law firm by an AI company

Operating system for AI-native law firms: outcomes-based fixed pricing instead of the billable hour. Firms running on it practise under the Manifest Law brand; the first incubated firm is in business immigration, structured under Arizona's ABS programme.

Last round$60M Series A, Apr 2026 — Menlo Ventures, Kleiner Perkins, First Round Capital, Quiet Capital. Reported as the largest Series A in legal tech history.

Traction100+ attorneys serving 3,000+ clients across the US

Why it mattersThe clearest expression of the 'services as software' thesis — capital is funding the firm, not the tool, and Arizona ABS is what makes the ownership structure legal

AI-native firm for in-house teams. AI does the first ~80% of the work and clients pay lawyers for the final 20%, at flat upfront prices with same-day turnaround.

Last round$9M seed/pre-seed — Y Combinator, 20VC, Urban Innovation

TractionSince launching in early 2026: 100+ companies supported, deals representing $2B+ in aggregate contract value across Europe, the US and Australia; reported ~4-hour average turnaround

Partner-only, fully distributed law firm across 90+ practice areas including corporate. No associates and no junior training by design — AI absorbs the work the junior layer would have done. Launched Jan 2024 with 130+ partners, one of the largest launches in US history; now near 300.

Model useRuns on Harvey, chosen for legal purpose-build and for SOC 2 Type II / ISO 27001 / GDPR posture. Harvey is multi-model, so this is indirect exposure to all three labs.

Traction~300 partners, 90+ practice areas, US and UK. Partners retain a far larger share of billings; the firm reports most joiners earning 2–3x their prior Big Law income.

Why it mattersThe largest working example of partner-only economics paired with AI: at scale, across two jurisdictions, with AI named explicitly as the replacement for the associate layer.

Startup contract work with AI doing the first pass. Usage pricing, no billable hour: roughly $100–200 per contract and $50–100 per attorney conversation.

Why it mattersThe most legible pricing in the category — per artefact and per conversation, which is what unbundling actually looks like

Launched May 2026 to advise startups using AI agents. Work product typically priced at 10–15% of a comparable large-firm bill.

Startup legal combining lawyers and paralegals averaging 10+ years of Big Law experience with custom-built AI, priced well below Big Law rates.

Caveatthin public information; details from directory listings

On-demand senior legal talent — secondments, fractional GC and specialised advice from lawyers with Am Law 200 and Fortune 500 backgrounds, now with an AI-enabled services layer.

Why it mattersHas sold senior-experience-without-the-pyramid for 25 years. The talent thesis is not new; pairing it with AI and fixed pricing is the new part.

On-demand legal talent platform for in-house teams — fractional and project-based senior lawyers.

Marketplace matching in-house teams to vetted outside counsel, including cross-border and specialist coverage, with managed pricing.

Named Claude for Legal user; also a Gemini Enterprise for Legal systems-integrator partner.

Gemini Enterprise for Legal launch firm, and also a named Legora customer — running both the model platform and the app layer on top of it.

Model useGoogle directly via Gemini Enterprise for Legal; Anthropic and OpenAI indirectly through Legora, which sits in all three ecosystems.

Buying across the platform layer — named by both Anthropic (Claude for Legal user) and Google (Gemini Enterprise for Legal launch firm).

Why it mattersThe clearest example of firms refusing to pick one model vendor

Named Claude for Legal user, running Claude on live matters.

Building rather than buying: $500M over three to four years on a proprietary AI platform, with $100M committed in 2026.

Traction~50 AI engineers engaged — roughly 35 via strategic partnerships and 15+ in-house, and growing. Still licensing some third-party AI.

Why it mattersThe reference case for build-vs-buy at the top of the market. One GC called it 'a few steps closer to killing the billable hour.'

Worked with OpenAI on information permissions and ethical walls for the Astra for Law Trusted Access program.

Named as using Claude on live matters.

Gemini Enterprise for Legal launch firm.

Gemini Enterprise for Legal launch firm.