Loud Few Days
These past few days were pretty loud. Legora launched their aOS, and their CEO opened the announcement with “legal AI is dead”. A marketing story not doing anyone any favours.
In their telling, the next chapter is agentic operating systems where the agents do the work and the lawyers review the results. Some say this is an answer to Harvey, who, just a few days before that, came out with their long-horizon agents.
Harvey also publicly confirmed they’re running on Anthropic’s Managed Agents, with their long-horizon completion rates going up around six times once they wired into Anthropic’s harness.
The more observant folks quietly saw this coming. The vendor layer has been moving in this direction for months. But now that two of the biggest names said it out loud in the same week, the trade press is going to be carrying it for a while.
But what does that mean to boutique law firms?
From Tool to Environment
When a vendor calls their product a tool, you trial it. You see if it helps with redlining or research, you decide it does or doesn’t, you renew or you don’t. Low commitment, easy to reverse.
When a vendor calls their product an operating system, what they’re really saying is they’d like your firm to live inside it. Intake comes in there, matters get tracked there, drafts get written there, the long running agents sit there. The integration goes deeper, the switching cost goes up, and the data the vendor is learning from you keeps growing.
That’s a much bigger commitment, and the calculation a partner has to do before signing on is a bit different from the one you did when you signed on for a Word add-in last year.
I’m not telling you to skip these tools by default. Some of them do really good work, and at the right firm with the right matter, they earn their seat. The point is that at boutique size, the cost of getting this decision wrong got bigger this week, and the market isn’t slowing down to let you think it through.
While Legal AI Died…
With every new announcement, bigger promises get thrown around. Is it just me or are the CEOs today competing in overpromising and underdelivering? In the past year alone we have killed billable hours, lawyers, the legal profession itself and now, to finish it off, legal AI as well. Those are, without a doubt, genuinely impressive achievements.
A few more grounded thoughts though. The first one is who these platforms are actually built and priced for. I’m working with partners who couldn’t get a sales call back over the past weeks from Legora - too few seats on their end to be worth the rep’s quarter. If the vendor isn’t picking up the phone for a 20-lawyer firm, maybe legal AI isn’t quite as dead as the headlines suggest.
The second one is that a lot of the boutiques I work with already have a working stack underneath them - some mix of Claude, ChatGPT, Microsoft 365, sometimes a vertical tool somewhere, and a real chunk of the work is moving through it. So when a new legal operating system feature drops or 500 legal agents get deployed on top of Claude’s harness, the question I hear in the rooms with partners is whether this significantly adds to what they’ve already got working, or are they paying to add another layer between them and the work? And a 30 min demo is not going to answer that question for them.
The third one is about the brand. So what I heard over the past year and a half is that a chunk of what gets paid for when a firm signs on with Harvey or Legora is the brand itself. A partner signs off on the well known thing, a prestige name on the market, and nobody can really argue with that purchase later if something goes sideways on the vendor side. Easy to defend later, even if it isn’t quite the right fit. At boutique size that pressure usually isn’t there the same way, which leaves the partner with a more honest question to answer - what does the firm actually need.
Why is AI the Partner’s Homework?
Every other non-legal function at the firm is staffed by external specialists. Marketing runs through a consultant or an agency. Accounting and bookkeeping run through a firm. IT, payroll, communications - same shape. The partners didn’t have to become marketing experts when the firm started doing marketing, didn’t have to learn bookkeeping when it moved from a yellow pad to a real system. They brought the discipline in on a rhythm that worked, and kept being lawyers.
So why is the AI question being treated differently? On most calls I’m hearing a variation on the same thing - the partner putting in evening and weekend hours comparing Spellbook to Co-Counsel, watching demos on YouTube, trying to keep up with what shipped this quarter. Decisions have to be made and so the whole thing lands on the partner’s desk.
The thing is that most managing partners I talk to want to remain lawyers. They want their law firm to remain a law firm. And those evening hours people are putting into AI homework are the hours that should be going into the parts of being a lawyer that AI can’t replicate - the judgment, the relationships, knowing what to do when.
The shape that’s been working at boutique size, from what I’ve seen, is the same one that works for marketing or accounting. A senior person on the technology side, on a rhythm that matches how the technology is moving. Holds the picture across the firm’s stack, vets tools as the market moves, sequences the next move, stays current on a market that shifts every quarter so the partner doesn’t have to.
Legal AI in Action
📰 The Fractional Boutique AI Partner
I wrote a longer piece on the shape I described above - a senior technology partner on a rhythm, the same way a boutique already brings in accounting or marketing.
The article goes deeper into the four pressures I’m seeing partners I work with grapple with right now, the multidisciplinary teams pattern from another industry the role borrows from, how to measure if this kind of help is actually working, and a couple of things you have to think about before you bring anyone in.
🎬 Claude Teams For A Law Firm, End To End
The three reasons the Teams plan is the minimal starting point for a firm handling client data, every setting in the workspace worth slowing down on before rolling Claude out to your team, and how to wirte Claude into your Microsoft 365 stack with a firm specific skill.
🎙 The Levels Between Using AI And AI-Native
Helen Fan is a California based attorney, Chief AI Officer at a Silicon Valley boutique, author of the Legal AI Value Stack, creator of AI-Native law firm experiment gone viral called OpenClaw LLP, and founder of the Legal Tech Frontier community spanning the US and Asia-Pacific. We talked through her five step framework of defensibility on the market, why she puts the firm’s own data at step three rather than step one and why the data cleanup project isn’t where boutiques should start, where Cowork actually fits as the right starting agent for a partner who hasn’t built before, and the talent-pipeline question forming quietly underneath all of it.
Coming up!
🎙 Next Tuesday at 2pm CET!
Next week’s guest on Rok’s Legal AI Conversations is Horace Wu, founder of Syntheia. After practicing law at Shearman & Sterling, DLA Piper, and King & Wood Mallesons, he’s spent the last 8 years building point solutions around the data layer of legal AI.
We discuss why he made that pivot and what it looked like, why the data layer is becoming the next investment frontier after the UX and intelligence waves, and why ROI in the AI era is so hard to measure when the unit of work itself has changed.
Each edition of Legal AI Brief brings practical lessons from firms using AI safely.