When lawyers imagine an AI-first law firm — and this goes for the media writing about AI-first law firms as well — they tend to spin a narrative in their head about a clever lawyer automating the boring parts, shedding associates, and pocketing the spread. For better or worse (depending on the imaginer’s feelings about AI and technology generally) the machines do all the work in these AI-first firms and the founder just counts money without ever having to manage a junior through six drafts of a stock purchase agreement.
That story doesn’t neatly fit the reality of setting up an AI-first law firm. So I talked to Logan Brown, founder of Soxton, an AI-first firm that the former Cooley lawyer launched last year with $2.5 million in pre-seed funding, about the day-to-day experience of practicing in a firm based around maximally leveraging AI. Brown herself describes the firm as the space between Biglaw and a Google search, the niche where many of her core clients — startup founders who need a Delaware C-corp but can’t cover Cooley prices — might be tempted to let GPTsus take the wheel when they really, really shouldn’t.
And keeping those founders involves a lot more than “just pushing buttons” — to the dismay of tech fans dreaming of a future low-effort practice and the chagrin of critics who want to believe lawyers leveraging AI are not putting in hard legal work.
I asked Brown, in the currency lawyers actually understand, how she allocates her time every week. “I think the majority of my week is spent on the engineering side of things,” she explained. “We’re hyper focused on creating not only a cheaper option for our clients but also a better option.” Some lawyers may read this as though Brown is “not doing law” with most of her week, but to my mind, this sounds like the soft time that senior lawyers currently spend training and managing human associates and staff. No one questions that a Biglaw lawyer sitting down with a junior and going over a markup counts as lawyer work. How is that different than researching and refining the tools used to generate these drafts? To the extent it’s different at all, it’s that the time Soxton lawyers spend working with engineers on the technical side returns dividends across a number of matters as opposed to just those the hapless junior ends up working on.
Not that human training is completely off the table. “We still spend time managing and training lawyers — the job just now includes teaching them how to manage and develop tech.” But that development work is where the biggest advantages live. “We have to rigorously test and improve our tools to ensure that they are working as they should but also in conjunction with our team.” This workflow doesn’t have to be motivated by a specific matter — the lawyers and engineers devote time to building a better mousetrap even when there isn’t a matter demanding that answer at the moment.
Which is what makes Soxton’s work different from a firm just “throwing on some AI.”
We’re rebuilding a law firm from the ground up. We think through every single task about the best way for it to be done with all technology available to us and with a business model that makes sense. We’re not adding AI to existing workflows, we’re recreating them from first principles.
This transforms the Soxton model into something closer to a constant iterative thought experiment. It’s going to become an important distinction because within the next five years every firm will claim to be an AI firm, but the nature of the deliverable is very different when a team of lawyers and engineers have designed products to handle specific legal problems as opposed to a law firm doing the same work as always, but with CoCounsel or Protégé adding efficiencies. “I think that lawyers that don’t try to understand the new tooling (that will continue to change) will be providing a fundamentally different service to consumers in the future,” Brown said. “I think that we need to ensure lawyers are given every opportunity to try the new technology again and again to see how it can improve their own workflows.”
Brown thinks that firms just adding AI as opposed to rethinking the role of the lawyer as a participant in developing tools could soon run out of runway. She figures the list of top revenue firms may look radically different in 10 years — a lot of familiar names reduced to “skeletons of themselves and the field will flatten.”
The billable hour is notoriously entrenched, but AI is the most robust challenge it’s ever faced. As AI evaporates time that used to be billed to clients — either through efficiencies in existing processes or, in the Soxton approach, because lawyers are spending more time on general skill refinement than direct client work — will put firms in a bind. They can either take less money (LOL), raise hourly rates to obscene levels (double LOL), or develop some kind of alternative, output-based billing model. Soxton operates on flat fees already, and it’s going to become harder for the rest of the industry to justify hourly models as AI spreads.
Which is all to say that when clients transition to expecting to pay by deliverable, the gap between firms that have taken the time to rethink workflows from first principles and the rest of the pack could get very wide. Soxton aims to get ahead of that moment.
Earlier: The Grace To Dabble: Two Biglaw Firms Look To An AI-First Future
Billable Hour Dying So Slowly, You’d Think It’s Billing By The Hour
Law Firms Prepare To Automate Themselves Out Of Their Own Business Model
Joe Patrice is a senior editor at Above the Law and co-host of Thinking Like A Lawyer. Feel free to email any tips, questions, or comments. Follow him on Twitter or Bluesky if you’re interested in law, politics, and a healthy dose of college sports news.