Legalweek is the legal industry's largest annual technology conference, with thousands of lawyers, legal ops leaders, and technologists descending on Manhattan’s Javits Center for four days of keynotes, panels, and hallway conversations that kept going long after the sessions ended.
By the end of the first day, a theme was already clear. Everyone agrees AI is going to reshape legal, but almost no one has figured out how to fully operationalize it. That gap between conviction and execution turned out to be the thread running through nearly every session and keynote at Legalweek. Here's what stood out.
"AI is making legal analysis abundant. Judgment—and accountability—remain scarce."
This sentence comes from the opening panel, Shaping Tomorrow: Visionaries Discuss the State of AI Across the Legal Market, and it set the tone for everything that followed.
During this sharply intellectual session, Danielle Benecke, Founder and Global Head of Baker McKenzie's Machine Learning Practice, laid out a framework that kept resurfacing throughout the conference. She mapped AI's impact in legal across four levels:
Most of the industry is operating at levels one and two. The firms genuinely differentiating themselves are building toward three and four. The gap between those two groups is widening, and Legalweek 2026 made clear that closing it requires deliberate choices most firms aren't making yet.
Oliver Roberts, Founder of Wickard and Editor-in-Chief of the National Law Review’s “AI & the Law” newsletter, pushed back on the popular refrain that "lawyers using AI will replace lawyers who don't"—calling it a framing that has already expired. He argues that it’s a strategic mistake to presume you know the upper limits of what AI can do given the pace of development. VCs and startups have real financial incentives to close the gap between AI capabilities and legal work. The end-to-end components of agentic legal work, while imperfect, are increasingly in place.
Ilona Logvinova, Chief AI Officer of Herbert Smith Freehills Kramer, brought it back to infrastructure, asking "What happens when agents need to use your infrastructure? Is it good enough?" Previously a back-office concern, enterprise architecture has become the foundation that determines whether thoughtful AI experimentation is even possible.
Zeynep Ersin, Chief Innovation and Strategic Design Officer of Seyfarth Shaw, framed the future plainly. The winners won't be the firms that adopt AI. They'll be the ones that operationalize it by defining accountability, owning their services narrative, and communicating that value clearly to clients.
Patrick Fuller and Heather Nevitt's closing keynote left the room genuinely—but usefully—unsettled. Fuller, the Chief Strategist, Legal at ALM, and Nevitt, Editor-in-Chief of Corporate Counsel and Global Leaders in Law, opened their presentation with an on-screen Penrose Triangle. The message behind this structurally convincing, visually coherent, and completely impossible shape landed before either of them said a word.
The data made the metaphor feel generous. Of the original Am Law 200 from 1999, 74 firms, 37% of that inaugural list, are no longer active or ranked as of 2026. They were disproportionately smaller firms, collectively representing only 27% of the original cohort's revenue. For over 25 years, scale proved to be the most reliable survival mechanism.
Fuller and Nevitt used the stories of Apple and Netflix not as motivational analogies, but as documented warnings. Apple built the iPhone knowing it would make the iPod, their most profitable product, obsolete. Netflix bet on streaming before DVD sales gave them any reason to. Both companies made an uncomfortable choice before external pressure forced their hand. The point was direct: the structural data undermines the legal industry's tendency to believe its situation is different.
Real Revenue Per Lawyer growth is essentially flat across most firm segments. The figure that warrants attention is the Non-Equity Partner CAGR, running at 9% for Am Law 51–100 firms. This is a sign, the presenters argued, that firms are managing Profits Per Equity Partner optics by expanding non-equity headcount rather than genuinely growing productivity: "PPL is going down for anyone outside the Am Law 50. Follow the geometry all around and it doesn't add up."
The recent Law.com Fellows and Wicker Park Group Survey on AI in Law told a similar story. Firms rated AI's significance to competitive advantage at 6 out of 7. They rated their lawyers' actual preparedness to use it at 3.9 out of 7. 75% are offering AI training, but 66% are not hiring the specialists to back it up. Only 19% are modifying fee arrangements, and 72% have no plans, or don't know, whether to change attorney compensation structures.
As Fuller succinctly put it, "The firms that win won't be the ones who adopted AI first. They'll be the ones who redesigned everything else to make it work."
Jae Um, Co-Founder, Chief Growth Officer & Head of Knowledge at Lumio, delivered some sharp observations during the Law Firm Deconstruction panel. She noted that while the headline for 2023 was AI risk and safety, the headline for 2026 is AI ROI. This shift matters because execution is now the differentiator—not access, intent, or awareness.
The change management data she shared was sobering but clarifying. Only one in six people are naturally drawn to new things. Another one in six will embrace change when they can see it will meaningfully impact their lives. That leaves roughly two-thirds of legal professionals who are averse to change by default. Her prescription: "If necessity is the mother of invention, pressure is the mother of adoption." Prioritize the clients and partners who are receptive. Build from there.
Economics adds useful context to the story. Skadden was the first firm to reach $1 billion in revenue in 1999. By 2025, 58 firms had crossed that threshold. 2026 will likely see the first firm exceed $10 billion. Law firms are already among the most profitable businesses in the world, which is precisely why the structural incentive to self-disrupt is so low. The firms that choose to disrupt themselves anyway will hold a meaningful and durable advantage.
As Um said, "Position is not destiny in law firm rankings. There is no magic in the model. The difference is execution."
Steno moderated the AI ethics panel framed around a single provocative question: if you're not using AI, are you committing malpractice? Joe Stephens, our Director of Legal Solutions, led the conversation with Allison Harbin, AI Portfolio Manager at Jenner & Block, Matt Krengel, Director of Information Retention Counseling at Cooley LLP, and Shannon Boettjer, Partner - Commercial Litigation & Governance at Jaspan Schlesinger Narendran LLP.
Krengel drew a useful early distinction. For tasks like eDiscovery document review, the case for using AI is becoming hard to argue against. But he pointed out that "when you're getting into quasi-lawyerly analysis, it becomes a much tougher question." How do you maintain the duty of competence when clients are directing you to use AI, but the analytical reasoning underpinning that work is increasingly being handled by AI systems?
Boettjer offered the killer analogy of the conference: think about how we evaluate a new drug when evaluating the use of AI. There are real benefits to generative and agentic AI. There are also known side effects—for example, AI hallucinations. But what are the rarer, more serious side effects that haven't yet revealed themselves? Confirmation bias and automation bias are two strong candidates Shannon identified that the legal industry is still wrestling with.
Harbin closed out the discussion with a critical observation. She noted that in an environment where the rules, tools, and norms are all changing simultaneously, "we all need to be open to dialogue." Calm, she said, is a professional asset right now. The lawyers who navigate this well are the ones who can sit with ambiguity, stay curious, and resist false certainty in either direction.
The AI Agents 101: What They Are and Why They Matter session was overflowing beyond capacity, which tells us a lot about what practitioners are actually thinking about right now.
Ethan Wong, Product Counsel at Anthropic, opened with a structural frame that breaks agentic AI into two categories: workflow-based agents that operate within defined guardrails, and autonomous agents that act based on the task at hand. For practical application, the sequential workflow model is the most useful lens—what are the individual tasks in a given legal workflow, in what order do they need to happen, and where can an agent take over?
Anthropic's own legal team uses Claude to handle thousands of internal employee requests—intake, review, follow-up, classification, and escalation to the legal team when needed. It's not theoretical. It's running.
At the Real-World Impact: How AI Agents Accelerate Legal Innovation panel, Kyle Poe, VP Legal Innovation & Strategy at Legora, argued that the onus for moving the needle on agents rests with partners, not innovation teams. It's their books of business that will be most directly affected, and they're the ones who need to have both the internal conversations about commoditization risk and the external conversations with clients about what comes next.
He suggests building AI accountability into partners’ annual evaluations, asking specifically what they've done with AI, how they're thinking about self-disrupting their practice, and how they're incorporating AI into business development. Underlying all of it: clean data is the prerequisite for any of this to work.
Farrah Pepper, Legal Innovation Partner at Harvey, moderated a panel of in-house and outside counsel on where the client relationship is actually heading. Winston Weinberg, Co-Founder and CEO of Harvey, traced the evolution clearly.
Initially, in-house counsel wanted to know how their law firms were using AI generally. Then, over the last 6–9 months, the question became, "How are you using AI on my matters?" Most recently it shifted to, "What are the net new things we can do together?"
The firms gaining ground are the ones treating AI not as an efficiency tool but as a foundation for building new kinds of client relationships. Winston's advice for navigating this moment: "Be honest and adapt. People are looking for who's going to be honest and work with me in the journey we're all on versus who's going to be defensive and hide things."
The question of what legal work is actually for also surfaced here. Is the goal of legal work to produce a document, or is it to achieve a client's objective? If it's the former, commoditization is the likely outcome. If it's the latter, the relationship, and the judgment inside it, becomes the product. The firms that understand that distinction are building differently.
Two-time Super Bowl-winning quarterback Eli Manning brought his keynote down to a single throughline: rely on discipline, trust, and mentorship to lead during uncertainty and change, regardless of whether your team is the New York Giants or a legal organization. His point on resilience was worth sitting with—every interception has its own story. Analyze it. Understand it. Then move on. A learner's mindset isn't optional, it's the job.
During her keynote, writer, producer and actor Mindy Kaling, best known as Kelly on The Office, brought a different energy. Her career was defined by calculated risk, building teams around a clear vision, and sustaining momentum in industries not always receptive to change. Kaling made an observation that landed differently in a room full of lawyers: talent is important, but so are curiosity and generosity.
She spoke candidly about AI, calling it the greatest existential threat to her business. But then she immediately reframed it. AI won't replace the best writers, she said, but it will replace mediocrity. Her answer to that? "I'm very, very thankful for my taste."
Both keynotes made the same argument in different ways: human skills matter most right now. The ability to lead through ambiguity, build trust across differences, stay curious, and keep showing up. AI makes a lot of things faster. It doesn't make those things easier.
Beyond the sessions, the Steno booth was rarely quiet, and dinners with clients and partners were great opportunities for face-to-face conversations. That's the part of a week like this that's hard to capture in a recap. But the ideas have a way of staying with you. Here's what stuck.
The window is real. The firms and legal teams actively building now–not just planning to build—are creating advantages that will be structurally difficult to close. The data on Am Law firm survival over 25 years is clear.
Honest conversations beat polished positioning. Over and over, the one thing in-house counsel and firm leaders said they wanted most was a partner willing to be honest about what's working, what isn't, and what comes next.
The human elements are not going away. Every speaker who touched on AI, even the most bullish ones, said the same thing: judgment, accountability, trust, and relationships remain scarce resources. AI raises the floor. It doesn't replace what's above it.
Calm is a competitive advantage. The lawyers, vendors, and legal ops leaders who will navigate this well are the ones who can sit with ambiguity without reaching for false certainty.
The junior lawyer question isn't going away. Nobody has a clean answer yet. But the firms asking it honestly, and building pathways for junior lawyers to develop genuine judgment rather than just processing volume, are the ones most likely to get it right.
***
Throughout Legalweek, we hosted David Cowen's Careers & the Business of Law podcast live at the Steno booth. All eight conversations with legal leaders are live now:
Legalweek confirmed what we already believed: the firms moving forward are the ones having honest conversations and making deliberate choices. If you want to see how Steno can support that work, we're ready when you are. Book a demo.