Legion AI Associate: Product, team and how it was built
Legion AI Associate (AI drafting of California litigation documents for lawyers, United States): 100+ paying California litigators, October 2026. Evidence grade E: no revenue figure was checked. Team: 2 co-founders (a practicing California litigator as CEO and a CTO) named on the company's page; headcount not published. Launched 23 April 2024; still running.Added 5 November 2026 · updated 6 November 2026
- Outcome
- 100+ paying California litigators, October 2026
- Evidence
- E no revenue figure was checked
- Launch
- 23 April 2024
- Status
- Still running
- Team
- 2 co-founders (a practicing California litigator as CEO and a CTO) named on the company's page; headcount not published
- Outside funding
- None reported (company's LawNext entry, January 2025)
- Group
- Legal and law firms
- Main channel
- Not stated
What is not verified
No revenue figure was found. I looked for a payment-processor profile, a revenue figure in the LawNext entry, the company's newsroom and its case study, and a funding announcement. The 100+ paying litigators and 28,600+ pages are the company's own counters on its site in October and November 2026, not checked by anyone outside. The US$286,000 on this page is my multiplication of pages by the standard price, not revenue: first documents are free, about four in ten drafts are not bought, and some page types cost more or less. The January 2025 figure counts firms and the 2026 figure counts litigators, so the two cannot be compared as growth. A reader could look for a funding round, an interview with a revenue figure or a customer count from an independent source. This story is grade E: the business is real and its product, price and customers are described from public pages, but no revenue figure was checked. Aiolum shows it as an idea worth studying, not as proof that it makes money.
Legion is an AI drafting platform, software in which a large language model writes a first draft of a document from the files that the user uploads, for California litigators, lawyers who handle lawsuits in court. It drafts pleadings (the complaint and the answer that start a lawsuit), written discovery (the formal questions and document requests that each side must answer before trial) and motion papers. The pricing page says the first document is free and most pages then cost US$10, billed to the client's matter.
Legion LegalTech was founded on 23 April 2024 in San Jose by Arthur Rothrock, a California litigator who says he still works his own cases, and Jeremy Kulchyk, the chief technology officer. In January 2025 the company told LawNext that it had 12 paying firms and no outside funding. In October 2026 its own site says it has drafted 28,600+ pages for 100+ paying California litigators.
No revenue figure was found, so this is a grade E story: the headline is the company's own count of paying users. Four other legal AI businesses on this site also came from LawNext Startup Alley ballots (Litmas AI, LegalBridge, Drafty AI and CounselPro); Legion is the one sold per page of output instead of per seat or per matter.
How much revenue does Legion AI Associate make?
Each row shows the evidence grade and the source numbers from the list at the end of this page. The grades are explained in How Aiolum verifies.
| Fact | Value | Grade | How we know | Sources |
|---|---|---|---|---|
| Paying users, October 2026 | 100+ paying California litigators | C | Company's home page, read 5 November 2026, and its comparison page: "As of October 2026 it has drafted 1,400+ documents and 28,600+ pages for 100+ paying California litigators." | [1, 5] |
| Output, October 2026 | 1,400+ documents and 28,600+ pages drafted | C | Company's comparison page. The same page says "about six in ten drafts started on Legion are purchased after the attorney reads them." | [5] |
| Pages drafted times the standard page price | about US$286,000 | calc | Calculated: 28,600 pages times US$10. This is not a revenue figure. The first document of each user is free, about four in ten drafts are not bought, some pages cost US$25, US$5 or US$3, and the company does not say whether "pages drafted" counts only paid pages. | [1, 4, 5] |
| Paying customers, January 2025 | 12 firms | C | Company's entry on the LawNext ABA TECHSHOW 2025 Startup Alley ballot, 9 January 2025: "We have 12 firms that are users/paying customers." LawNext says the entries are based on information provided by the startups. Most customers were 2- to 20-person firms in civil and probate litigation. | [2] |
| Founded | 23 April 2024, San Jose, California | C | Company's LawNext entry. The site footer names Legion LegalTech, Corp., San Jose. | [2, 1] |
| Founders | Arthur Rothrock (founder and CEO) and Jeremy Kulchyk (co-founder and CTO) | C | Company's About page. Rothrock writes that he still litigates. Kulchyk built production AI systems at Intact Financial, an insurer. The LawNext entry speaks of a "founding team" and "founders". | [3, 2] |
| Team size | not published | C | The About page names the two founders and a board director; the terms of service mention a "Legion QA Team". No source gives a headcount, and none suggests more than 10 people. | [3, 7] |
| Outside funding | none, January 2025 | C | Company's LawNext entry: "No outside funding." No later round was found on the company's newsroom page. | [2, 6] |
| Price | US$10 a page for most documents, first document free | C | Pricing page, November 2026. The same US$10 a page was stated in the LawNext entry of January 2025. | [4, 2] |
| AI models | third-party models such as OpenAI's API and Anthropic's Claude API | C | Terms of service, sections 1.1, 9.1 and 9.3, which also state zero-data-retention agreements with OpenAI and Anthropic. The About page says each workflow is routed to the frontier model strongest on that task. No source names a model version. | [7, 3] |
| Workflows shipped | fourteen since March 2025 | C | About page: "We have shipped support for fourteen production workflows for California litigators since March 2025." The newsroom lists each release from the Complaint Builder of 2 March 2025. | [3, 6] |
| Time from founding to the October 2026 figure | 30 months | calc | Calculated: April 2024 to October 2026. | [2, 5] |
How to read this table: a grade C value is a statement by the company on its own site or in its LawNext entry. A "calc" value was computed on this page. No value is grade A or B, and no revenue figure exists.
| Plan | Per month, billed yearly (US$) | Per month, billed monthly (US$) | Limit |
|---|---|---|---|
| Pleadings: complaint, cross-complaint, answer | 10 per page | per page drafted | |
| Discovery you serve or answer, meet-and-confer letter | 10 per page | per page drafted | |
| Memorandum of points and authorities | 25 per page | per page drafted | |
| Separate statement | 5 per page | per page drafted | |
| Notice of motion, declaration, judicial notice, proposed order | 10 per page | per page drafted | |
| Demand letter, probate petition or objection | 10 per page | per page drafted | |
| Amend a document you already have | 3 per page changed | per page changed | |
| First document | 0 | 0 | one free document |
All prices are per page of output; there is no subscription, no seat fee and no minimum. The first document is free, nothing is charged until the lawyer downloads the draft, and revisions are free. A page is a California Rules of Court page (28 lines, 12-point type); partial pages round up, per the terms of service. Charges are billed to the matter in one statement a month, so the lawyer can pass them to the client as a case cost.
When did Legion AI Associate launch, and how did it grow?
From a San Jose company founded in April 2024 to fourteen drafting workflows and 100+ paying litigators in 30 months. Each entry carries its grade.
- 23 April 2024CLegion LegalTech is founded in San Jose, California. [2]
- 9 January 2025COn the LawNext Startup Alley ballot the company reports 12 paying firms, a price of US$10 a page and no outside funding. [2]
- 2 March 2025CThe Complaint Builder launches: civil complaints for California Superior Court from uploaded case files. Discovery, motion and answer builders follow through 2025 and early 2026. [6]
- March and April 2026CA partnership with Litigation Labs, a managed legal services provider whose paralegals and contract attorneys run Legion for firms, and the launch of Centurion, an AI analyst beside each draft. [6]
- June 2026CA Demand Letter Builder, and the written discovery suite extended to New York matters. [6]
- October 2026CCenturion searches and checks case law through CourtListener on every matter. The company states 28,600+ pages drafted for 100+ paying California litigators. [6, 5, 1]
Why did Legion AI Associate work?
Four causes, in the order they mattered. Each one is a reusable lesson.
One state, one kind of document
Legion drafts only California litigation papers, with the state's formatting and procedure rules built in, so the draft comes back on pleading paper with the parts that a California court expects. A general chatbot or a national platform leaves that work to the lawyer.
Pay per page, after reading
The pricing page states that nothing is charged until the lawyer downloads the draft, and the first document is free. The cost goes on the client's matter as a case cost, which the page says the State Bar's 2026 guidance allows when the fee agreement says so.
A builder who is also a user
The CEO writes that he still litigates and that real cases drive the roadmap. The newsroom shows a new workflow about every four weeks since March 2025, each one a document type that a litigator files.
How it was built
These details come from the company's own pages and terms of service. They are grade C. [7, 3, 1]
- Language models from third-party providers, named in the terms as OpenAI's API and Anthropic's Claude API, under zero-data-retention agreements.
- Each workflow is routed to the model that the company judges strongest for that task, with guardrails at the prompt and validation layers against invented facts and cases.
- For specialized workflows, a proprietary database of California case law; for general motions, the lawyer supplies the cases.
- Centurion, the built-in assistant, searches published opinions in CourtListener and checks that each cited case exists, but not its later history.
- Customer data is stored and processed only in the United States, encrypted with AES-256 at rest and TLS 1.2 or higher in transit.
- Exports to Word and to bookmarked PDF files with exhibit slip sheets.
Tools named in the sources
Each one opens its page in the Tools tab, with every story that uses it.
Who tried this before Legion AI Associate, and why did they fail?
Three companies in our database of failed startups tried to automate legal work with software and failed. The values below are from that database and are not independently verified.
Atrium
The promiseLegal work for startups made faster and cheaper by a law firm built around software.
Why it diedThe automation could not handle real legal complexity, so the company carried law-firm costs without the quality of a large firm.
ROSS Intelligence
The promiseAI legal research to replace searches in the large legal databases.
Why it diedLawyers stayed with the incumbent research databases, which had decades of data and added AI themselves.
jEugene
The promiseAn automatic checker for errors in contract drafts.
Why it diedLaw firms adopted it slowly, and better-funded AI competitors overtook it.
The failed startups come from our database of 1,749 dead startups, retold here in our own words. Its entries are AI-assisted summaries of public sources, so these cards carry no evidence grade.
What changed
The three predecessors had to build their own language technology, sold to firms that were slow to change, and met incumbents with more data. Legion rents its language models from OpenAI and Anthropic, so a company with two named founders could ship fourteen document workflows between March 2025 and October 2026.
It also avoids the hardest part of Atrium's model: Legion is not a law firm and does not give legal advice. The terms of service state that the lawyer reviews and signs every document, so the legal judgment and the liability stay with the licensed attorney.
Can one person copy Legion AI Associate now?
Yes, for someone with real practice experience in one jurisdiction.
The method is a narrow drafting product for one state's courts, built by a lawyer who uses it, sold per output with a free first document. The large platforms sell seats to big firms; small firms with a few litigators are left with general chatbots that do not know local formatting.
Where it is still open: Other states' court documents, administrative filings such as workers' compensation or immigration forms, and the same per-page model for other professions with fixed document formats.
- 1Pick one jurisdiction and one practice area whose documents follow fixed formats.
- 2Build each document type as its own workflow, with the local rules encoded, and release them one at a time.
- 3Charge per page or per document only after the user reads the draft, so the first trial costs nothing.
- 4Ground the drafts in the user's own files and check every cited case against a public case-law database.
What could go wrong for Legion AI Associate?
The risks are the mirror image of the causes above.
What is not verified
No revenue figure was found. I looked for a payment-processor profile, a revenue figure in the LawNext entry, the company's newsroom and its case study, and a funding announcement. The 100+ paying litigators and 28,600+ pages are the company's own counters on its site in October and November 2026, not checked by anyone outside. The US$286,000 on this page is my multiplication of pages by the standard price, not revenue: first documents are free, about four in ten drafts are not bought, and some page types cost more or less. The January 2025 figure counts firms and the 2026 figure counts litigators, so the two cannot be compared as growth. A reader could look for a funding round, an interview with a revenue figure or a customer count from an independent source.
Invented citations in court filings
Lawyers who file AI-written briefs with invented cases risk court sanctions; a customer testimonial on Legion's home page speaks of "ChatGPT Lawyers" who "face sanctions". Legion's About page says "Nobody can honestly promise zero hallucinations or citations you can ship without checking", its Centurion check confirms that a case exists in CourtListener but "does not check subsequent history", and its terms (section 13.2) say hallucinations cannot be eliminated and the user remains solely responsible. I could not open an independent court record on AI sanctions: calmatters.org and www4.courts.ca.gov answered with bot challenges, and CourtListener's robots.txt disallows Claude agents.
Competition
Legion's own comparison page names Harvey, CoCounsel, Lexis+, Briefpoint, AI.Law, Anytime AI, Clio, Midpage, Paxton and several California drafting tools. It reports that Harvey raised US$550 million in September 2026 and that Harvey and LexisNexis launched, on 7 October 2026, a Motion to Dismiss workflow that recognizes a California demurrer, the first motion-level overlap. The same page states Briefpoint's price as US$150 per attorney a month, cheaper than Legion for a lawyer who drafts more than about 15 pages of discovery every month.
Dependence on model providers
The terms of service (sections 1.1, 9.1 and 9.3) say that the platform sends client content to third-party models such as OpenAI's API and Anthropic's Claude API, under zero-data-retention agreements. A change in those providers' prices, terms or retention agreements would reach the product directly; the About page says the company switches to whichever model is best for each workflow, which lowers the dependence on any single provider.
Sources
The grade next to each source is the best grade that the source can support. "Read in full" means that the text of the page was downloaded and the numbers were checked in it.
- C Legion home page, read 5 November 2026Read in full. Source for the 28,600+ pages, the 100+ paying litigators, the US$10 a page price and the customer testimonials.
- C LawNext (Bob Ambrogi), 9 January 2025: Voting Is Now Open: Pick the 15 Finalists to Compete at Startup Alley at ABA TECHSHOW 2025 in AprilRead the section on Legion AI Associate. The entry was written by the company. Source for the founding date and place, the 12 paying firms, the price, the growth plan and the absence of outside funding.
- C Legion, About: Why We Built LegionRead in full. Source for the founders, the release list, the model routing and the hallucination guardrails.
- C Legion pricing page, read 5 November 2026Read in full. Source for the pricing table.
- C Legion library: Legion vs Harvey, Briefpoint, AI.Law, and Other Legal AIRead in full. A company page. Source for the October 2026 counts, the purchase rate and the competitor list with their prices as the company reports them.
- C Legion newsroomRead in full. Source for the release dates from March 2025 to October 2026 and the Litigation Labs partnership.
- C Legion terms and conditionsRead the sections on the platform description, fees, third-party AI models, attorney review and hallucinations.
- C Legion library, 14 January 2026: How a San Francisco Litigation Firm Reduced Staffing Costs 27% and Increased Profits Using AIRead in full. A case study published by the company about a six-attorney customer firm.
- C Database of failed startups, local exportLocal export. AI-written summaries, not independently verified. Source for the failed predecessors.
Run this company? Send corrections or a verified figure to aiolum@livyden.com