The divergence of law firms from lawyers

LLMs' absorption of legal task performance will drive law firms towards commoditized service hubs while raising lawyers to unique callings as trustworthy legal guides — so long as we do this right.

In the private legal sector, lawyers and law firms have always been inseparable from each other.

• A law firm is the commercial vehicle for the delivery of lawyers’ services, and it was never intended to be anything else. In practical terms, law firms cannot produce anything of value to clients other than through the work of lawyers. Take the lawyers out of a law firm and you leave behind, in a commercial sense, an empty shell.

• Equally, every lawyer who carries out a legal task for a client needs a law firm to do so. In fact, a lawyer effectively “creates” a law firm just by providing legal services; even sole practices are considered law firms. Lawyers precede law firms, but they also generate them and are indispensable to their existence. It’s a symbiotic relationship.

Generative AI is going to weaken and potentially dissolve that relationship. Law firms will become capable of generating output that can be sold to clients with no lawyer involvement at all.

Right now, it’s possible for an ordinary person to obtain from an LLM like ChatGPT-5 the performance of a legal task — the provision of legal analysis, the production of a legal instrument, the delivery of legal advice — that previously could only be acquired from a human lawyer.

I’m not saying a person should do that. The LLM’s output might be effective and reliable, or it might prove disastrously off-base. But many people are already using LLMs in this way, and in the absence of other accessible options for legal assistance, they will continue to do so.

The flip side, however, is that a law firm can also apply an LLM — a legal-domain-specific version, built to function at a “lawyer standard” — to perform a legal task and then sell the output to a client. The LLM, in that situation, would be the primary performer of the legal task. This is something entirely new. (Yes, some law firms sell the work of paraprofessionals, but not on anything close to the scope, scale, and capacity of LLMs.)

Any law firm currently using AI to generate legal output will place a lawyer in the production loop, both to ensure accuracy and to protect itself against ethics investigations. But as Generative AI gets better at performing legal tasks, that oversight will become more perfunctory, and past a certain point, it will taper off altogether. A law firm eventually will generate and sell legal work product without the active involvement of any human lawyer.

As Large Language Models continue to gradually displace lawyers from legal task performance, they will thus bring about a divergence, a parting of the ways, between law firms and lawyers. LLMs will make it possible, for the first time, to have a law firm that sells legal services to clients without any involvement by lawyers. I know there are regulatory barriers to that possibility, and these might remain in force for some time yet. But in practical terms, that possibility is already here, and that’s momentous.

I wrote last month about the displacement of lawyers from legal task performance and its transformative effect on how lawyers understand themselves and identify their purpose. But the impact of LLM task-displacement on law firms might be even more profound.

At a conceptual level, law firms are understood and defined entirely in terms of lawyers. Their inventory is the billed hours of their lawyers’ effort, their capital base is the equity contributions of their lawyer-owners, and all their metrics of importance (productivity, utilization, realization, growth) are expressed in “units of lawyer.” From a traditional profitability point of view, the more lawyers in the law firm, the better.

Within the next year or two, the opposite trend will start to take shape. As LLM capability continues to move up the legal value ladder, law firms will find it economically less feasible to continue giving lawyers tasks that their competitors — and their own clients — are giving to AI instead. The lawyers themselves likely would balk at such assignments too, given that nobody really enjoys carrying out work that a machine has shown it can do. Legal task performance will require fewer “lawyers per task” than it did before.

This won’t automatically mean the end of the billable hour, by the way. Although lawyers won’t be engaged in the performance of legal tasks and the generation of legal output nearly as much as they once did, they can (and I believe will) raise their game to provide services founded in advice, advocacy, guidance, strategy, counsel, and accompaniment. Lawyers can bill those services on a hourly basis if they like (and if their clients agree), or they can bill them on a fixed-price basis, or on a relationship basis, or on the phases of the moon, I don’t really care.

But the steep reduction in lawyer contributions to legal task performance will strike at the heart of three other, elemental aspects of law firm operations.

• Profitability: Lawyers make money by selling the time they spend performing tasks for clients. But law firms make profits by selling the time of less experienced lawyers for more than those lawyers are paid. Associate leverage is the essential source of law firm profitability. Without leverage, law firms are just lawyers selling their time; with it, they’re profit factories. So if associates aren’t billing hours, because there are no such hours to be billed, where will firms find profits?

• Development: And for that matter, where will firms find suitable work for juniors? The beauty of associate leverage was that clients effectively paid for the development of new lawyers, whose leveraged labour on less complex tasks doubled as law firm professional development. Of course, this was always experiential immersion more than actual “training.” But most firms don’t know any other way to develop lawyers other than through repeated legal task performance. If machines are performing their tasks, how will associates learn?

• Value Metrics: The greatest contribution any lawyer can make to their law firm is to bring in substantial, sustainable client business. But a close second is to pump out billable work at a steady rate, and that’s not an optional activity for lawyers at any stage of their careers. If less work is performed by lawyers billing their time, how will firms assess the relative contribution of lawyers to the firm’s success? I’m not certain most law firms can even process the concept of measuring value in non-hourly terms.

The time and effort lawyers put into the performance of legal tasks and the generation of legal output is central not only to how law firms operate, but also to how they’re conceptualized. LLMs’ entry into the legal service production chain is going to force a fundamental reorientation of what law firms are and what they’re for.

This could go a number of different ways. Here’s one of the most interesting possibilities.

Suppose we reach a point where legal task performance is so thoroughly within the capacity of Large Language Models that lawyers are not required for legal task performance at all. They’re not needed as translators or intermediaries between the client and the solution (advanced prompting takes care of that), or for quality control (superior LLM engineering takes care of that). I don’t think this is the likeliest outcome, but it’s certainly within the realm of possibility.

In the event this does come to pass, however, what would we even need “a law firm” for? If LLMs develop this breadth and depth of capacity, then anybody would be able to use one for legal assistance — an in-house lawyer, a government employee, a single mother. Legal task performance would essentially devolve down into an app, or an online hub. You’d no more go to a “law firm” for these products or services than you’d walk to an “Uber station” to order a ride.

What about professional insurance? Doesn’t the involvement of a lawyer mean that a legal service purchaser has access to an errors and omissions fund to cover losses from negligence or incompetence? Yes, it does — but who says insurance can only be accessed through the lawyer? I buy temporary auto insurance when I rent an SUV at the airport counter, but I buy the policy from Enterprise, not from Toyota. Lawyers aren’t the only source of potential coverage if a legal transaction goes wrong.

The divergence of law firms from lawyers could lead to the extinction of “law firms” as we understand them — and to the consequent flourishing of “law firms” as something entirely new: branded, accessible, super-highly-engineered legal solution engines, perhaps universal in scope, more likely narrow in market and subject-matter focus. If most legal task performance becomes commoditized — and I think there’s a pretty decent chance it will — then so will law firms, insofar as their only real market function is to perform legal tasks.

But as law firms dive down along one path, lawyers could soar up along another. What we now consider the highest and finest callings of the legal profession — advice, advocacy, strategy, and accompaniment — would be the only ones left for lawyers, once LLMs disintermediate them from legal task performance. Clients would be able to get legal task performance off the shelf; but they could only get a lawyer’s unique trustworthy guidance, counsel, and loyalty from that particular lawyer.

And a lawyer wouldn’t need “a law firm” to engage in these activities. A law firm would only slow the lawyer down. Stewardship of a client’s affairs doesn’t involve hourly billing or leveraged labour or the endless circulation of drafts and revisions. Nor would the provision of strategic watchfulness and advice, or the fulfillment of critically important societal roles, or the envisioning and pursuit of Legal Moonshots. And those are just the post-AI lawyer functions we can glimpse from this side of the transition. There would certainly be others.

Like I said, this could go a lot of different ways; certainly, the picture I’ve painted here is among the brightest options. And a lot would need to go right for it to happen — starting with a complete overhaul of lawyer formation, of course. But the forces at work here are not hypothetical, their pace is accelerating, and the commoditization of most legal task performance seems more likely than not at this point.

There’s a real opportunity here for the legal profession to divest itself of the worst trappings of its old legal practices — the hours, the targets, the drudgery, the tournaments — while prioritizing and upgrading all the best parts of being a lawyer. But it’s not going to happen by accident. What lawyers make of their coming divergence from law firms will shape the future of legal services for everybody. It’s incredibly important that we get this right.