By Braeden DeWan, Director, Marks Baughan

The legal industry has been hearing “disruption is coming” for roughly two decades. It first came from B2C platforms like LegalZoom, then from Alternative Legal Service Providers, then from a wave of legal tech SaaS tools. None of them fundamentally changed the practice of law. A new category of AI-native law firm threatens to do exactly that, and the distinction matters.

What’s an AI-Native Law Firm?

The term “AI-native law firm” is everywhere right now. Depending on who you ask, it’s also inaccurate, contested, or already obsolete. When 50 attorneys, founders, and engineers gathered at Stanford in May 2026 to define it, no one agreed, although three words kept surfacing: “client-facing AI,” “self-learning data loop,” and “flat fee.”¹

The ambiguity is real, and our own reporting confirms it. The firms we spoke with for this article span a wide range of structures, business models, and self-descriptions:

• Ryan Wenger’s Inhouse AI is a self-service platform backed by a law firm, designed primarily for small businesses that have never had a lawyer and aren’t looking to hire one. 
• Richard Perris’s Covenant is, by contrast, a full-service law firm for private markets clients that has built its entire delivery infrastructure around AI. 
• Rick Merrill’s self-named law firm is an AI-first litigation defense practice targeting Fortune 1000 companies. 
• Michael Drapkin’s Landfall IP is a specialized intellectual property firm using AI to supercharge elite patent practitioners. 
• And Michiel Van Lerbeirghe’s company, Duely, is a technology-enabled M&A advisory service that embeds AI engineers alongside legal experts to process legal due diligence data at levels of quality, speed, and transparency no traditional firm can match. 

Today, many might use the term “AI-native law firm” to describe all five of these businesses. Tomorrow, the term itself might be irrelevant. As Perris said, “we’re exploring the future here, so there’s always a degree of uncertainty attached to that.”

Terminology aside, how is the market categorizing this work? An Altis research report defines an AI-native law firm as one where “AI sits in the production line vs. a part of the toolkit.”² That’s a useful start, but it doesn’t fully capture the range of what’s being built, nor does it reflect the philosophical core shared by these companies, which is that the old model of billing for time cannot survive in a world where AI can perform the majority (or at least some) of that work without human cognition. Everything else — structure, pricing, target client/user, complexity tier — is a variation on that shared conviction.

The Original Sin of Billing for Time

The billable hour’s fatal flaw in an AI world is that getting faster at something earns you less money for doing it. AI-native firms have inverted this concept completely. Their margins grow as human minutes per matter fall.

As the axis on which the traditional law firm turns, the billable hour is also a structural problem. Partners vote on firm strategy, and those partners  make money under the existing system. Proposing to restructure around efficiency is asking them to voluntarily undermine the economics that made them and their firms successful. As Van Lerbeirghe put it, the problem is not that lawyers lack access to the technology. “The problem is the business model or incentive of legal services. Even where traditional firms have access to the same tools, they have no structural reason to push automation as far as it can go. AI-native firms have every reason to,” he said.

Perris put it plainly: “The thesis is that the old law firm model just cannot exist anymore. Relying on the fact that work takes time by people — time that we can charge for by the hour — as a proxy for value, just isn’t going to work in a universe where LLMs exist.” 

That universe is here, and the firms that are building on this thesis are boldly taking on a frontier that traditional law firms can’t or won’t embrace.

The New Workflow and a Flywheel to Match

At a traditional firm, a contract request triggers a cascade of human hand-offs. A partner confirms receipt and assigns the matter to an associate who drafts from scratch or hunts through a document management system for useful precedents, with or without the assistance of technology. Redlines go back and forth over days, each round billed separately. When the associate who led that deal leaves, the institutional knowledge —  drafting and negotiation preferences, etc. — goes with them.

At an AI-native firm, the same request moves through some combination of AI-driven processing and human triage. The exact mix varies by firm, client, and the nature of the work. AI may handle initial classification, document review, issue-spotting, or first-pass drafting. A lawyer may step in at predefined triggers or assess from the outset how much human judgment the matter requires. In some workflows, the attorney is barely visible until the final review. In others, they’re involved throughout, with AI accelerating each stage rather than replacing it. What’s consistent is that when the matter closes, all intermediate steps update the firm’s knowledge base. The next deal on that client account starts from a smarter position than the last one.

The operational numbers are striking. General Legal turns an 8–10 hour MSA markup into ~ 2 hours at a $500 flat fee. Crosby delivers a redlined contract in 1–4 hours versus the 1–2 days a traditional firm requires.² Landfall IP can draft a very high quality patent application in less than 10 hours, which used to take 40+.

Beyond speed, there is a decentralization of knowledge that directly benefits clients. As one former product leader at an AI-native firm described it, “The biggest difference is that Crosby is trying to decentralize the knowledge. In big law firms, there’s only a small number of lawyers who take on a certain client because they might have prior knowledge about that client. But the way that Crosby is trying to operate is: how can we decentralize all that so that any lawyer can step in mid-negotiation and jump in without having that prior knowledge stored in their brain.”²

Perris describes Covenant’s flywheel in similar terms: “If you build the platform right, you can capture much more than used to be capturable in a human’s OneDrive or Google Drive. And the AI can access it readily and never forget it. The model’s ability to pull from experience the next time it’s answering a question is the next level of where we’re going.”

The model also applies to litigation. Merrill said, “For every case we do, the system learns and gets better. We track absolutely every detail of every document, every action of opposing counsel, every action of the judge, every action of the parties, jurisdictions, everything.” 

The vision is a proprietary data repository that compounds over thousands of cases — something that traditional law firms have the historical case volume for but haven’t built their systems around because they don’t think analytically that way, Merrill argues. “We’ll be able to tell our existing clients and prospective clients that we’ve done this now 5,000 times. We understand the plaintiffs, we understand the judge, we understand what happens in a detailed way that most law firms just don’t,” he said. Or law firms do, but that intelligence resides in a few partners’ heads or in inaccessible, unstructured documents. If that intelligence is retained and used to benefit the client, it certainly doesn’t come with a cheaper bill.

Each firm we spoke with describes a version of this same principle. 

At Inhouse AI, the flywheel runs on volume — millions of interactions with non-lawyers that reveal exactly where AI fails and how to fix it. The company is building a level of reliability for users who don’t know what they don’t know, something competitors without that user base simply can’t replicate. 

At Landfall IP, it runs on practitioner depth where every complex patent matter is handled by elite attorneys working alongside AI agents to produce better invention strategy, faster processing of technical material, and a stronger model of what winning patent strategy looks like. 

At Duely, it runs on data density. From the moment a client grants data room access, AI engineers begin processing and synthesizing the contents of the deal data so that by the time the legal expert arrives, the institutional knowledge on the transaction is easily digestible. 

In each case, the promise is the same: the system that serves the next client is better than the one that served the last.

The Automation Threshold

While the flywheel component is compelling, the harder question is whether the economics hold while firms are still building toward it. These firms are chasing software-like unit economics. At ~80% gross margins, they have it. At ~50% gross margins, this looks more like a traditional law firm, and scaling is more difficult. But the real answer seems more nuanced. 

The founders building these firms tend to frame that threshold more as a moving target because margins may improve continuously as the system learns. If every matter makes the next engagement cheaper to deliver, then the automation rate isn’t a line to cross. It’s a number that keeps moving in the right direction as long as the data flywheel is working.

Where the Automation Stops, for Now

While some research suggests AI-native firms are not coming for all legal work, the firms we spoke with are operating across low, medium, and high complexity work. They aren’t replacing lawyers on complex matters, instead they’re using AI to handle what it can and routing the rest to attorneys who are now freed up to focus on judgment rather than process.

Drapkin argues that what looks like a complexity ceiling is really a moving floor. At Landfall IP, his team uses AI agents to handle invention extraction interviews, patent drafting workflows, and technical document processing at a level of speed and depth that was impossible a few years ago. “A couple of years ago, just to get a patent attorney to understand a highly complex invention disclosure would have taken more than 5 hours. We now have a voice agent, and in an hour-long conversation with an inventor, we cover more and get a better understanding of the strategy than you would have been able to do in 20 hours two years ago,” he said. 

Where human judgment remains essential, the founders are consistent about the fact that it’s less about legal knowledge and more about context, client relationship, and accountability. “The human touch can never be replicated,” said Van Lerbeirghe. “If more complex work is introduced, then it might just require more of the human touch.”

Perris agrees there isn’t “much of a ceiling in terms of the theoretical capabilities of the models.” He said the models are already smart enough to do the “vast majority” of work “even the most experienced and highest-paid lawyers do.” The challenge is “getting [the AI models] the right context, asking them the right questions, and getting the output in a way that’s useful, which means building workflows around them,” he said.

So, if the complexity ceiling isn’t a ceiling at all, but rather a floor that adjusts as humans get better at giving the technology the right context and stepping in when necessary, then it certainly seems as though AI-native law firms could come for all legal work…one day.

“Why Not Just Hire One Good In-House Lawyer?”

The intuitive objection to the AI-native model sounds like this: If the underlying technologies are widely available and improving rapidly, why wouldn’t a company just hire one strong, tech-forward, in-house lawyer, give them access to Claude or GPT, and get equivalent results for all their legal work for the cost of a salary?

The answer has several parts, and the most obvious is cost. Wenger put it directly: “Hiring a full-time attorney is going to cost $250,000+.” For the SMB market his company targets, that’s a non-starter. 

Perris approaches this from the enterprise angle, where the in-house option is more realistic. He said, “Just because you’ve got AI doesn’t mean you’re suddenly an expert in every single area of law and can apply the right levels of judgment and market knowledge and nuance. Lawyers will be better with the help of Claude, but you still need to get specialist advice.” There is a reason GCs spend so much money currently on outside counsel.

Furthermore, an in-house lawyer would have to systematize a data flywheel to get the self-improving aspect these firms tout. While not impossible, that’s a tall order. At Inhouse AI, every user interaction feeds back into the platform, making the system smarter across the entire user base. Wenger calls this the “glass box” advantage: Since users interact directly with the platform, his firm can see precisely where the AI is failing and correct it at scale. “We have, to my knowledge, the most legal AI chats for non-lawyers. We’ve been able to build these massive eval sets that allow us to build really high quality AI,” he said. 

And there is the speed argument that directly impacts the bottom line. The Altis report found that customers aren’t primarily buying on cost — they’re buying because legal review was the bottleneck slowing down their sales close cycles.² An in-house hire with access to Claude or GPT still has the same calendar and the same approval queues. The AI-native firm’s promise is that the bottleneck disappears entirely.

The Four Horsemen of the AI-Native Apocalypse: Speed, Price, Quality, and Accountability

The number of AI-native law firms nearly doubled from 2025-2026.³ Three — General Legal, Arcline, and Vector Legal — appeared in Y Combinator’s Winter 2026 cohort alone. The AI Firm Index tracks the emergence of these firms, and 15 appeared between April-June 2026.⁴ As more enter the scene, who will come out on top and why?

It might seem fruitless to predict anything now as the market is in its nascence, but a few factors will be essential to any competitive moat. After speed and price, a major factor is ownership of the workflow and the data that compounds it. The firms building proprietary clause libraries, negotiation pattern datasets, and client-specific playbooks are building assets that compound with every matter, and those assets belong to the client, not to a partner who might leave. 

Wenger underscored the importance of this differentiator when he said, “We’ve built such a large database of these chats that it would be very hard for somebody to catch up. You can use off-the-shelf AI, but it’s not going to be dummy-proof. A restaurant owner can go to Claude and print out a menu because they know what menus look like. But that restaurant owner can’t produce an MSA or vendor agreement because they don’t know what a good one looks like.” 

And workflow ownership is tied to the data flywheel. Clients know that AI-native firms’ systems improve the more matters they handle, and they want that accumulation of institutional knowledge to be visible and actionable. 

This quality of work is just as essential to the competitive moat as the speed and price value propositions. “The only reason for law firms like ours to exist is if we’re better or cheaper,” said Merrill. “If you’re not better in your legal work product or if you’re not cheaper in your costs, then why do you exist?”

The fourth factor is accountability and the human relationship, which might seem like an ironic competitive advantage in a market built around automation. Some firms we spoke with were adamant that clients are not buying software, they are buying a service with a person behind it. Van Lerbeirghe put it directly when he told us that the EQ dimensions of legal services — the relationship, trust, and accountability — will always require a human in the loop. Perris affirmed, “These are deals that are done between human beings. You need to know that there’s a human there. That’s always very important to people.” 

The firms that treat the human layer as a part of the value proposition are the ones best positioned to retain clients long after speed and price advantages narrow.

Our interviewees optimistically agreed that this is not a winner-take-all market, and some like Wenger proposed that the market will fragment into a collection of segment-specific leaders rather than one dominant player. The question is not who wins the category, it’s which firms within each segment can best deliver on those four value propositions of speed, price, quality, and accountability. The firms that get all four right may end up defining what legal services look like in the future.

“The Only Direction of Travel”

“AI-native law firm” is a useful label for now, but the founders in this space believe it is a transitional category and a new operating model that will soon become the only model.

“I think it has to become the standard,” said Perris. “There’s only one direction to travel. We think that all law firms will have to look a lot different in five years, maybe sooner. You won’t need to use the term ‘AI-native’ or ‘AI-first’ because eventually it might just be how any law firm looks.” 

Merrill emphasized that there will indeed be a reckoning, and it will largely be about pricing. “Lawyers are not going away, but there will be an enormous repricing of legal services,” he said. “There is going to be huge downward pressure on legal fees, especially for routine matters. For the really complicated bet-the-company stuff, that’s different. But for almost everything else, there’s going to be a significant downward price pressure that very, very few firms will be able to manage properly, at least at first.”

The traditional firms that recognize this sea change are already moving. In late May 2026, Kirkland & Ellis, the world’s highest-grossing law firm, announced it would invest $500 million over three to four years to build its own proprietary AI platform rather than license third-party tools.⁵

“The telling detail is not the size of the investment,” said Perris. “It’s that only the world’s most profitable law firm can afford to make it. For everyone else, the choice is either to restructure around AI economics or to defend the billable hour while the market reprices around them.”

The sea change is underway. As Legaltech Hub’s Nicola Shaver observed, “AI-native firms are important not because they will replace traditional firms, but because they demonstrate that alternative legal production systems are now viable. They make visible what was previously theoretical.”⁶

The firms building in this space today are not disrupting the legal market. They’re defining what legal services look like next.

Sources: 

1. Lab, H. L. A. (2026, June 1). How Silicon Valley is defining “AI-Native law firm” — 3 words that keep coming up. Helen’s Legal AI Lab. https://helenfan1.substack.com/p/how-silicon-valley-is-defining-ai 
2. Altis. (2026, April 15). Crosby. https://www.altis.vc/reports/crosby 
3. Legal IT Insider. (2026, April 24). AI-native law firm index hits 40 listings. https://legaltechnology.com/ai-native-law-firm-index-hits-40-listings/ 
4. The AI Firm Index. (n.d.). The AI Firm Index: Directory of AI-native law firms. https://aifirmindex.com/ 
5. Killelea, E., & Strom, R. (2026, May 28). Kirkland & Ellis investing $500 million to build AI platform. Bloomberg Law. https://news.bloomberglaw.com/business-and-practice/kirkland-ellis-investing-500-million-to-build-ai-platform 
6. Shaver, N. (2026, February 12). AI-native law firms and the innovator’s dilemma: A fabric-change signal for the legal industry. Legaltech Hub. https://www.legaltechnologyhub.com/contents/ai-native-law-firms-and-the-innovators-dilemma-a-fabric-change-signal-for-the-legal-industry/