Reading Time: 7 minutes
I was running through the syllabus highlights recently as we started a new semester of lawyering skills. It’s primarily a writing and research class, with professionalism and practice technology incorporated when possible. As we touched on artificial intelligence, there were no questions and I didn’t belabor it. If law students use it, they risk making the upcoming hurdles of exams, in school and for the bar, much harder. The use will catch up to them if it is replacing their skill development. I think most law students understand this risk balancing, perhaps better than some law firms. If anything, the law firms allowing AI to lead their strategy, rather than support it, seem to have lost this perspective.
It is hard to get excited about generative artificial intelligence in the grand scheme of things. It is a technology that the designers clearly don’t know how to make work with any regularity. There is some patchwork effort being made with Claude Skills and whatever this new OpenAI Astra for Law plugins is expected to deliver. We’re starting to see some of this become visible in legal research products. Lexis Protege has Skills now (although with a usage cap to avoid token maxxing) and my assumption is that, like the OpenAI plugins, is a one-size-fits-all approach developed by people who may be in the legal profession or adjacent to it. But “by lawyers, for lawyers” isn’t always what it’s cracked up to be.
The [OpenAI Astra for Law] launch includes 9 community plugins from lawyers and legal engineers at LegalQuants, LECG, and Skills.law, with 47 custom skills that practitioners can adapt, extend, or draw inspiration from to do their work in ChatGPT. Community plugins give the people closest to the work a way to keep setting the standard for what these skills can do.
Introducing Astra for Law, OpenAI
What is the likelihood that the average lawyer is going to tinker with Skills? Even back in the 1990s, when lawyers were selling each other their WordPerfect macros, there was never an obvious market for “software that lawyers customize”. Not that most lawyers will have access to Astra since it is being aimed at the largest law firms. And not that it matters.
“Unexplained” attacks by artificial intelligence applications on one another’s company don’t suggest to me that the technology is autonomous so much as it is poorly designed software. Any tool’s manufacturer that can’t explain why there are disclosures of wiped systems—or perhaps how to stop the tool from ignoring guard rails and protective requirements (erm, like Skills)—isn’t a good technology for law practice. I can’t even imagine how a lawyer would explain to (a) their regulator, (b) their insurer, (c) their clients, and (d) courts and other external organizations that “I failed to properly supervise an AI agent who wiped my law firm’s computer systems.” That’s a heck of a disclosure letter.
In the time since I started writing this post. a number of AI developers have come out with warnings that AI growth needs to slow down. These companies may be as worried about their out-of-control burn rate— AI companies will need to increase their own productivity by a factor of 2.7 to break even by 2030, accounting for the cost of capital and a 15% return, and depreciation of the assets—as much as anything else. People who have placed big bets on these app builders, investors and law firms alike, are right to be concerned.
The Point of Practice Technology
I have long held the belief that lawyers should adopt technology only when it enhances their ability to provide legal services. That could be because they become more efficient or because clients receive better service, whatever, when they use the technology. Anyone with a long perspective on law office technology will have seen fads come and go and seen lawyers adopt the ones that make sense. There may be a law firm practicing in Second Life but I doubt it’s widespread. I would be a bit curious to know about how Arent Fox’s move to the Metaverse is going. Are their VR clients just pervert glass wearers or are there legitimate, real clients using the Metaverse office instead of their brick and mortar locations?
We don’t know what the metaverse will be in five years, but we’re not waiting five years to find out. We’re going to help iterate that future
Anthony Lupo, Arent Fox chair, February 2022
It’s been five years so perhaps the passage of time has proved their pudding. It’s funny to me that the virtual universe really looks just the same as any old office tower or Zoom call. At least you could be hybrid while not physically being in a building (it would be bananas if the law firm requires staff to be in the office to be present in the Decentraland location).
Decentraland says they held over 300 community events that averaged 127 unique participants each; that seems small for a community that might support even one lawyer. They had 41,783 web site visitors last month. That’s a bit more than 2x the number of visitors we had to our law library’s LibGuides site. Although this also-2022 piece from Bloomberg hypes the metaverse, even the companies it names are starting to drop their own investments in the concept.
All of which is to say: we didn’t see a lot of lawyers run to the metaverse and so not many will have to figure out how to unwind those operations.
But Our Clients Expect …
The mistake that some lawyers make, I think, is that they feel they need to use the same technology as their potential clients (bitcoin and NFT enthusiasts, for example). Or perhaps its that the largest law firms drive media discussion in the legal profession and their windmill tilting becomes overlaid on the entire profession. But there is lots of technology that lawyers do not need to use to do their jobs because the software is for doing other things.
I was struck, then, by the same assessment—technology adoption is not driven by fear of being left behind but by usefulness—being applied to ransomware exploit teams. As this analysis of the use of AI for security exploits discusses, technology has to have a point. Why would you iterate for the future if that is not where the money is?
However, such assessments start from a false premise about what drives the behaviour of cybercriminals. The history of modern cybercrime demonstrates two things. First, that cybercriminal behaviour has been driven more by innovation in business models than technical capabilities. Second, that cybercriminals tend to innovate when they have to, not simply because a new kind of technology becomes available.
Beyond the Hype: AI, Ransomware and Business Models, Will Lyne and Jamie MacColl
They “innovate when they have to, not simply because a new kind of technology becomes available.” I mean, really, why would you adopt each technology that comes out without knowing whether the investment will pay off. This may be particularly true where technology use is constrained by ethical rules and market saturation.
It was no surprise to me, then, when I saw some recent Google research that showed an analysis of occupational categories and adoption of generative AI tools, specifically Gemini. I realize that using this sort of research, considering Google’s market share in generative AI, may be like looking at Bing search users: unrepresentative due to a small market slice. When I have heard people in the legal profession talk about generative AI, Gemini isn’t a brand name I hear very often (okay, never).
Law is down squarely among the sorts of work, like truck driving or construction or security, where it is difficult to apply AI to the tasks at hand. It is lower even than libraries and education. I would assume that the difficulty is not one of acquiring the technology (high-earning workers have the highest AI adoption, but it is mostly adopted for tasks requiring low-to-middle expertise, p. 3).
There isn’t a lack of adoption, either. Lawyers are one of the highest adopters of Gemini. The difference appears to be the tasks that the top group, computer and mathematical, can use generative AI for and the tasks (and expertise) that lawyers can use it for.
The whole report is interesting to me. But it does seem to suggest that lawyers are not using it for many of the key things that lawyers do. There is some data about people using AI for legal services and government services but it isn’t clear whether lawyers are involved or not (my assumption is no). And since one of the charts shows that “nearly half of all medical, legal, financial, and government and civic services AI conversations occur outside standard business hours”, I’m guessing a lot of those are not involving people on both ends of the conversation.
The Strategy is Money
Perhaps my starting point is wrong. Maybe the legal profession doesn’t have a client-oriented strategy so much as a money-making one. That would explain the state of the justice system as it is impacted by the largest law firms. I literally laughed out loud when I saw that the personal injury firm Morgan & Morgan—the firm fighting to have celebrity endorsements but that can’t manage a pro hac vice admission due to “inattention and carelessness”—is going to make its own AI platform … for lawyers. While there are plenty of law firms that have spun off service agencies for clients, I can’t really imagine how this is going to be appealing to anyone.
As more large law firms become affiliated with private equity, it makes me wonder if we’ll see more distortion in the legal profession. The largest organizations, who can both afford the absurd costs of generative AI deployments and feel a competitive need to publicly promote their use of it, may find their practice strategy yaw in a way that isn’t replicated by regional and smaller law practices. As private equity and their “managed service organizations” end-run professional regulators in order to link up with large law firm cash flows, larger law firms may not have the independence of thought to go a different direction. There is a reason that companies like OpenAI and Thomson Reuters target their products to the large law firms. It’s the money, not the service model, that is the attraction.
When I see companies adopt the Next Big Thing, I often think of that American classic, The Music Man. When Harold Hill, f/k/a Gregory, a/k/a Professor Hill, is asked “How far are you going, friend?”, he answers, “Wherever the people are as green as the money, friend.” Or as someone put it on Mastodon recently, “I’ve never encountered an actually revolutionary technology that people had to be forced to use“
Again, not that it matters much: the firms at this end of the legal services spectrum account for 10-15% of lawyers. As large law firms become less and less like law firms, both in operations and independence, their strategy may need to diverge from that of professional lawyers. An attempt to convert a “laughing stock” into a financial security, so to speak.
A decade ago, I guessed that AI would mostly be a technology that impacted lawyers, as opposed to one they widely adopted. I still think that’s the case. What will be interesting is to see what the heavy AI strategy in the largest law firms does to their trajectory on things like financing (do they remain independent) and hiring (are they able to afford to hire new lawyers or just laterals as bot herders?) in 5 years. Will they be, like Arent Fox, encamped on a frontier to nowhere?
There remains little obvious reason for a 2026 lawyer to revise their strategy with generative AI as a core goal. The strategic mistake is to assume a productivity tool can be so overweighted as to derail other considerations. Ongoing accuracy issues, the opaque nature of the tool, and the cost to license model access are all drags against adoption.
But fundamentally, as it so often has been with the legal profession, it’s not yet clear what improvement it makes to how lawyers practice law. It fails the basic risk balancing that skill development and expertise acquistion and professional ethics require when utilizing a new technology. In a profession that uses slight automation despite having widespread, more reliable automation tools (like macros, PowerAutomate and workflow tools, etc.) that only do what they are configured to do, I think generative AI will remain a hard sell.
The largest law firms may adopt artificial intelligence but we will never see them absorbing any of the large untapped market that is currently self-represented. It is far more likely that we will see them create an expertise and training vaccuum that will no longer be filled by new law school graduates. They will do so on the back of a still-poorly tested technology that has very questionable financials. If there is an AI bubble, law firm’s with inverted strategies based on it may find themselves in the same position as the businesses who have speed run this curve and who found themselves having to hire human intelligence to achieve their strategic goals.