Horizontal AI models are not built for legal drafting.
They invent authority
They hallucinate, misstate, summarize and truncate legal authority, then produce overconfident content that appears legitimate.
The errors are hard to find
Even excellent attorneys have trouble understanding why the text is incorrect and how to correct it.
An RFE response is one shot
Legal practice demands accuracy. A single unsupported citation can cost the petition and the beneficiary's status.
We start with one of the highest-stakes filings in immigration law: the response to a Request for Evidence (RFE), which U.S. Citizenship and Immigration Services (USCIS) issues when it needs more proof before deciding a visa or green card petition.
AI models don't use a legal workflow.
Overconfident draft that attorney has to figure out what's incorrect in it. No collaboration, no choices, and no judgment is left to the lawyer.
Better than horizontal legal AI: a collaborative workflow that works with lawyers to help them build the best tested, authoritative, and verified arguments, starting with immigration.
Every claim shown beside its exact source authority.
In head-to-head testing our verification layer outperformed Westlaw Quick Check.
Legal AI can draft. It cannot yet reason
the way a veteran attorney does.
| Capability | RFE Responder | Harvey · Visalaw · LEGORA · others |
|---|---|---|
| Strategy playbook | 30+ years of a practicing firm's winning filings | Relies on synthetic data |
| Verification | Compares generated text next to source text with verification package | Citation checking, no proof artifact |
| Agents | Multiple agents collaborating with you | Have to build your own agent |
| Collaborative workflow | Steps attorneys through issue, evidence, authority, and application for best results | AI generates drafts without an attorney-in-the-loop workflow |
We compete on data no amount of capital can purchase: outcome labels from a working practice.
Filed on real client matters, with approvals to show.
Baseline: 85.97% of H-1B petitions receiving an RFE were approved in FY2025 (USCIS data).
USCIS is scrutinizing more petitions,
and strategy decides the outcome.
USCIS issued 61,505 RFEs on Form I-129 in FY2025 across 572,582 completions. Standards vary by service center, by officer and by administration. No policy manual tells you what actually wins.
Source: USCIS Form I-129 Request for Evidence workbook, FY2025 (October 2024 to September 2025). DHS/USCIS Office of Performance and Quality.
Win immigration first, then expand.
TAM: total addressable market, the total annual spend in each segment. RFE Responder: ~300k RFEs a year × $179. Visas.ai: 18k immigration attorneys × $5,000 a year. LegalEngine: 1M U.S. attorneys × $5,000 a year.
We deploy alongside the firm, then productize.
Configured on your own filings
We build it on the firm's own cases so it reasons the way that firm's best attorneys do. Each engagement makes the next one faster.
Standards move, so we keep pace
Adjudication standards change with each administration. A static playbook goes stale, so firms subscribe to stay current.
Attorneys pool what wins
Attorneys contribute winning strategies and get access to the pool in return. That library only grows through use, so it cannot simply be bought.
Deployment work gets us in the door.
Target customers: solo, small and mid-size immigration firms, plus in-house teams.
24 months: who we win, in order.
Velie Law's network, AILA chapters and direct outbound. Self-serve Starter and Pro plans.
Pro and Enterprise plans, configured on each firm's own filings and winning responses.
Employers that file their own petitions, on Enterprise contracts with Visas.ai.
A first practice area outside immigration, with design-partner firms on LegalEngine.ai.
Customer and ARR figures come from our financial model, built on current list pricing ($349, $999 and $2,999 a month plus per-response credits).
30+ years of winning strategies encoded into an AI.
"We aren't just engineers building a wrapper on an API. We are domain experts with 30+ years of legal expertise solving a problem engineers alone have failed to solve."
Jon Velie, Founder & CEO · Immigration attorney since 1993 · Member, U.S. Supreme Court Bar
30+ years of winning cases
We know what strategies worked and how to win cases
Deep AI expertise
Founding team includes experienced AI product builders and researchers
We are the users
Our law firm uses our tool to file petitions and respond to RFEs
Leadership and domain mastery.
- Immigration Attorney Since 1993
- Member U.S. Supreme Court Bar
- ABA & Stevie Award Winner
- Bestselling Author on Immigration Law
- CS Lecturer @ UC Berkeley
- Built CourseAssist AI (SIGCSE 2024)
- Founding Engineer @ Radiance Commerce (acquired by Bloomreach)
- UT Austin Economics Graduate
- Product manager and growth operator across Insurtech, AdTech, and AI software
- 20+ Years in Growth, Operations and FinanceFortune 250 and Startup Leadership
- CS @ Columbia
- Experience with user-facing agentic AI systems and core product infrastructure
What the $2M buys, quarter by quarter.
Cumulative operating spend from our financial model. Revenue offsets part of it, so peak cash need is about $0.6M.
$2,000,000 seed.
24 months to $1M+ ARR and operating profit.
Use of funds, 24 months
What it buys
Team costs include 15% payroll taxes and benefits. Shares of the $2M: 51% engineering and design, 16% sales and customer ops, 4% AI and cloud, 29% growth and reserve.
