AI-native firms may not be taking Big Law’s market share, but they are making incursions into a valuable asset: talent.
“The AI-forward attorneys are chafing at the slow pace of firm adoption and archaic thinking,” said Sam Shaddox, 38, a co-founder of Seattle’s Talairis Law Group. “They’re migrating to the firms that are leading the way on AI, or leaving Big Law entirely to chart their own path.”
WATCH: Can AI argue a Supreme Court case better than a human? This lawyer tried to find out.
Shaddox and Matt Souza met at the University of Washington School of Law, cut their teeth at Perkins Coie, and spent years inside legal departments of Seattle-area tech companies. In May, they launched Talairis Law Group to advise startups with the help of AI agents.
The number of law firms branded as “AI-native” or “AI-powered” is growing quickly, backed by millions in venture capital. Many of their leaders left major law firms early in their careers to launch businesses aimed at young companies and entrepreneurs.
Logan Brown, a 30-year-old Harvard graduate and former Cooley LLP associate, officially launched Soxton AI in New York in December. JP Mohler, 36, was an associate at WilmerHale and Cooley after Harvard Law School and tinkered with AI tools at Casetext and Reuters before forming General Legal through Y Combinator with two co-founders.
Matt Souza and Sam Shaddox
Talairis
Some Big Law veterans are also heeding the call. Norm Law appointed Mike Schmidtberger, the former Sidley Austin executive committee chair in January. Moritz, a San Francisco firm whose CEO and co-founder served as OpenAI counsel, has hired attorneys from Cooley, Goodwin Procter, and Fenwick & West.
General Legal’s 14 full-time lawyers are almost entirely Big Law alumni. They were recruited in part because mid-level associates are “frustrated” by a partnership track that offers little control and years of deferred reward, Mohler said. All full-time lawyers receive equity in the company, which provides flat-fee contract and employment law services.
General Legal has raised $11.5 million and reached $2 million in annualized revenue, Mohler said. That’s a speck of what large firms see in revenue each year and less than one-fifth of what the average Kirkland & Ellis partner earns in annual profits. Still, Mohler has lofty goals.
“Ten years, it will be the biggest law firm in the world,” he said. Change will unfold “a lot faster than any previous kind of disruptive cycle.” He says he is aiming for “venture capital scale”—meaning a company ultimately valued between $10 billion to $100 billion.
The Models
AI-native law firms like Soxton and Talairis are built around artificial intelligence from the ground up—every workflow, pricing model, and staffing decision is designed assuming AI does the first pass of the work.
At Soxton, Brown said AI is not a tool layered on top of existing workflows. It is the workflow with attorneys operating it. Soxton has about 40 contract attorneys, two engineers, and two full-time lawyers, including Brown. Every lawyer hired, she said, has spent at least four years at a Big Law firm focused on startups.
Shaddox takes a different approach, billing itself as a firm “built on the traditional model with AI sprinkled on top.”
Mohler is less interested in the labels. “Call it AI-native, call it AI-powered, call it AI-pilled,” he said, using a Silicon Valley term. What matters is whether a firm employs engineering talent working hand-in-hand with lawyers. “If you’re a firm of mostly lawyers and you don’t have engineering talent building you the most powerful AI tools, you only have half the equation,” he said.
Some large firms are taking AI matters into their own hands, including by building their own AI tools. Kirkland, the country’s largest law firm by revenue, is committing $500 million to an AI investment that includes a partnership with Palantir Technologies Inc., and references on‑premise GPU environments.
For Shaddox, it’s validation. He and Souza started Talairis with $15,000 of their own money and turned down outside investment, he said.
“What Kirkland is doing is a different scale of what we’re doing today, but it is working towards the same solution,” Shaddox said.
Read more: The Hottest Job at Big Law Firms Is Becoming Difficult to Fill
The firm runs on a three-layer architecture sitting atop large language models from Anthropic, OpenAI, and Google, which the founders treat as interchangeable. The first layer is more than 100 proprietary AI agents encoding their combined 25 years of legal experience across startups, investment rounds, and M&A. A second layer maintains what Shaddox calls a “client genome”—a detailed profile of each client.
The founders say they built the platform without outside engineers or product teams. That distinction matters, Shaddox said, because most legal AI tools on the market reflect an engineer’s interpretation of what attorneys do.
General Legal built a similar architecture, which Mohler calls Ansible and Sentinel. Ansible works by generating hypothetical legal scenarios and asking attorneys to answer questions about them—building a proprietary knowledge base from the firm’s own lawyers. Sentinel is the delivery layer attorneys use to execute the work.
“Our lawyers have all downloaded their brains into the system,” Mohler said. “Almost all” of the firm’s lawyers use Claude Code, Anthropic’s coding-focused AI tool, as part of their day-to-day workflow, he said.
‘Broken’ Economics
The firms’ pricing models are in stark contrast with Big Law, where even some associates bill out at $1,000 per hour.
Soxton charges $100 to $200 per contract and $50 to $100 per attorney conversation, usage-based, with no billable hour. Talairis’ actual work product typically runs 10%-15% of what a comparable big firm would charge, according to Shaddox. The firm offers flat fees for “repeatable” work, subscription-based “fractional” general counsel arrangements, and an “effective billable hour” that drops as AI accelerates the work, he said.
“The whole economic model of law firms is broken,” said Souza, 43, the Talairis co-founder. “Law firms are built on this billing model—it’s built on the backs of junior lawyers and their work, and that is exactly the work that AI is going to either eliminate or make way more efficient. ”
AI-native firms are “simply a continuation” of the “ongoing structural shift in the delivering of legal services” over the last two decades, according to Daniel W. Linna Jr., director of law and tech initiatives at Northwestern University School of Law.
“Tech is not the hardest part,” Linna said. AI-native firms must still hire lawyers who “embrace a technology-centric business model” and win clients.
‘Do or Die’
Peter Salib is an assistant professor at the University of Houston Law Center and advisor for AI safety groups. He said AI is fueling a “durable structural shift” in how legal services are delivered. “They are not perfect, but neither are first-year associates,” he said. “And the models are only going to get better.”
Salib is less sure on the future of AI-native firms. “If AI native means buying proprietary models, fine-tuning on internal documents, etc., then I’m skeptical,” he said. “But if AI native means ‘aggressively adopting AI into practice,’ then yes, I think that’s a durable shift. In the long run, every firm will do that or die.”
Norm Law, valued at $1.2 billion, is the most direct challenger targeting Big Law’s corporate clientele. The firm, which has more than 20 lawyers and nearly 50 legal engineers, is focused on advising finance-sector clients. It has raised more than $260 million from investors like Blackstone, Bain Capital, Vanguard, Citi, and Khosla Ventures.
Some of the new firms are drawing on modern financing tools to get outside investment within tight restrictions for US legal service providers. Soxton, General Legal, Moritz, and others are structured as management services organizations, a model that separates the tech and business operations from the law firm itself.
Those arrangements have also attracted some interest from large firms, in part as a potential way to pay for AI-related tech costs.
Shaddox and Souza rejected the model for their firm.
“When tech AI and the data are inextricable parts of the legal work itself, we don’t think that back-office structure works,” said Souza. “That’s why we kept the AI, the data, and the firm entirely lawyer-owned.”
