Ben Smith wants small law firms to stop drowning in paperwork. His solution: AI agents that churn through intake forms, draft engagement letters, and populate decades-old practice-management software while attorneys sleep.
Async, the New York startup Smith co-founded with Connor Park, emerged from Y Combinator's Summer 2026 cohort with a pitch tailored to solo practitioners and small firms in law, healthcare, and real estate. The company calls it "digital labor, deployed for you." In August, Async claimed it had automated thousands of hours across its customer base, though it declined to specify how many clients that represents.
Small businesses have watched corporate legal departments and Big Law experiment with generative AI for document review and research. Now a wave of startups wants to bring similar automation to firms that can't afford dedicated IT staff or multi-year software rollouts.
Async's agents don't require businesses to abandon their existing systems. Instead, the software layers onto what's already there: Microsoft Word for contract drafting, email for client communications, Anthropic's Claude interface for research. Customers describe their workflows to Async's team during configuration calls, then watch as agents take over tasks like scheduling therapy appointments or generating real-estate closing documents.
Joshua Lowenthal, who runs a solo practice, described the product as "far and away a better resource" than baseline AI models in an April 17 testimonial on Async's website. His firm deployed agents for drafting and legal research. Output arrives "a quick note away," he wrote.
Four law firms published similar case studies on the company's site between April and August. Adriena Clifton, founder of Washington-based Clifton Law PLLC, reported feeling the impact within hours of turning on agents for intake and conflict checks in late April. "They have me for life!" she wrote.
Jonathan Clemente, managing partner at New Jersey bankruptcy firm Clemente Mueller, P.A., called it "magic" in an early August testimonial. Carlos D. Martinez, whose firm Martinez Legal also practices bankruptcy law, said Async automated standard procedures "at a fraction of the cost of an additional hire."
The enthusiasm from early customers sits alongside a less polished operational reality. Y Combinator's directory lists Async's team as just two people as of late September, though the company's LinkedIn profile shows 11–50 employees. Smith dropped out of college at 20; Park studied computer science at the University of Michigan. The company operates formally as Skope Technologies, Inc. d/b/a Async, per its terms of service.
Privacy remains a central pitch. Async's policies, updated September 7, state that customer data doesn't train its models. The startup offers zero-data retention options and holds SOC 2 Type I certification, a security audit standard for service organizations.
The company sells outcomes instead of per-seat software licenses, a model that mirrors how consulting firms bill but applied to AI. Async's website references backing from Y Combinator and other investors, though it hasn't disclosed funding details beyond YC's standard investment.
Competition has intensified quickly. Anthropic launched "Claude for Small Business" with pre-built agentic workflows in May, according to Axios. Every.io announced back-office agents for startups tackling banking and payroll in March. Make.com and AnyMind Group both rolled out agent platforms this year, the latter arriving in August.
Whether a two-person operation can scale fast enough to hold ground against better-capitalized rivals remains an open question. Smith framed the mission in broad terms on LinkedIn in late August: "We're on a mission to bring frontier, specialized AI to American small businesses."
For now, Async is betting that law firms and healthcare practices will choose targeted agents over general-purpose AI tools. The early testimonials suggest that for some attorneys, at least, the relief of offloading intake forms and conflict checks outweighs any skepticism about delegating judgment calls to software.
