“Cravath To A Million Squillion!”

Sorry, but we’ve seen this dance so many times before that, you have to admit, we all know the script by heart.  Cravath (or other blue chip firm [insert here]) bumps up associates’ salaries across the board by an average of $10,000–$20,000/year—in this case, to a base for first-years of $225,000 starting in January—and the top 50 or so law firms follow suit [i] in hours, days, or max a couple of weeks.  The American Lawyer asserts without elaborating that the group of firms matching this new normal scale “remains a fairly limited club…with PEP above $2-million per year and [each with] more than a billion dollars in gross revenue.”

Long-time readers of Adam Smith, Esq. know that we rarely comment on developments like this, for a few reasons that we hope are relatively self-evident, at least upon a moment’s reflection.

We have a strong, bordering on non-negotiable, preference for publishing original material—stuff you won’t find anywhere else—and the coverage of stories like this is guaranteed to be exhaustive (and exhausting).

And perhaps most importantly, we find it difficult to discern any thematic meaning in these episodic spasms of “call you and raise you” on the associate salary front.  (Real meaty “news” on this front to us would be something like a firm declaring that they decline to match “the going rate” because their training and intensive skills-development workshops provide their associates different and superior long-term career value than a few thousand dollars a year after taxes.  Never happen.)

But this time is different.

Why?

We’ll offer up three reasons.

First, the air may finally be getting thinner up here.  So long as essentially any credible AmLaw 200 firm could match Cravath if it chose to without drastically affecting its partners’ year-end take-home, some combination of prestige, envy, associate recruiting competition, and plain old chesty pride led many firms outside the “Top 50” to follow suit.  (For now, I will invite you to define the requirements for qualifying for entry into that privileged set as you wish, but dollar-driven economics doubtless has to be part of the threshold bar.) In other words, it was more a matter of choice and preference than exigence and necessity.

Second, and sticking with the P&L implications of this for a moment, I commend to you if you haven’t seen it our series on “The Maroons and the Grays,” which among other things posits that clients’ demand for law firm services is price-elastic for the great run of legal issues that arise in the day-to-day operations of sophisticated enterprise.  These matters are handled by law firms in the “grays” category.

The model, as we introduced it at the outset of the first column in our series on this, begins as follows:

We have never subscribed to the belief that law firms operate in a fundamentally undifferentiated industry—that each law firm competes with every other law firm—and that given the magic alignment of expertise, cost, and personal rapport, clients’ choice of law firms is essentially unbounded.  Not so.  When it comes time for a client to choose a law firm to address an issue, the client knows the short-term and long-term gravity of the issue to the company, has a rough sense of what the monetary and/or reputational risk exposure is, knows the venue/locale (if that’s germane), has a good idea of which firms have handled lots of similar matters and which haven’t, and indeed has in mind a multitude of other considerations from the quantitative to the subconscious.  These combine to produce a short list of “plausible” law firms for the matter.

Stated differently, law firms do not exist in their own economic walled garden, exempt from the normal economic dynamics of competition, substitutes, price/quality tradeoffs, reliance on familiarity and personal relationships, and the anchoring effects of search costs.  And, to run a law firm effectively you have to be a gimlet-eyed realist about your firm’s position in the market.

Only a distinct minority of law firms—the “maroons”—qualify for being in clients’ consideration set when truly consequential legal matters arise; matters receiving boardroom level attention with potentially long-lasting and even existential consequences for the corporation.

For clients of the maroons, price is not a material object.  But for the grays, price surely is far more than material in the firm selection process; it is consistently in the top three reasons why/why not to select that particular “gray” firm.

Stated in economic terms, client demand for the maroons is all but price-inelastic; for the grays, it’s always quite importantly  price-elastic.  Which means the cost base (associate comp, remember where we came in here?) matters.

Third, ChatGPT and the next and the next and the next wave of Generative AI.

This is new.

The first perspective we mentioned above (how high is too high to match comfortably?) is intrinsic to setting benchmark pay levels and has been going on forever, and the second (existential issues vs. running the trains on time) has not changed.  Ever has it been thus.

But just one data point about how law firms are already experimenting with GAI—today, in the equivalent of GAI’s nursery school years—gives us all a strong navigational indication as to where it likely will be going in Law Land.

A few days ago The American Lawyer published an article by three of the senior leaders of Sidley’s eDiscovery team describing one of their very first experiments with GAI as applied to a real world discovery assignment.  (See Replacing Attorney Review? Sidley’s Experimental Assessment of GPT-4’s Performance in Document Review, available here .

Sidley’s experiment strikes me, as someone with a quasi-experienced amateur’s familiarity with research protocols, as a structurally sound test.  They took a closed matter having to do with the Anti-Kickback Statue and extracted 1,500 documents, one-third of which had been deemed to be responsive and two-thirds non-responsive to a subpoena request.  Next they asked GPT-4 to analyze and label the documents using the same review instructions given to the human lawyers. Also in line with attempting to “mirror” the guidance the human reviewers had been given, they gave GPT-4 a modified prompt to clarify ambiguities in the instructions that “provided the same additional information” given the attorneys in the original review.

Bottom line:

“Once these experimental adjustments were made, GPT-4 performed well.”

More specifically, where GPT-4’s confidence was “high” in how it categorized a given document, “its performance was quite strong [in] properly tagging documents.”  Indeed, so much so that “in its current form, GPT-4 may be reliable as a substitute for human review.”  (emphasis supplied)  And even where GPT-4 told the research that it was less confident, it tagged almost two-thirds of documents correctly, but/and that once GPT-4 had the opportunity to “re-review” those same documents with supplemental instructions given to the actual attorneys during the real-world review—which strikes me as the only fair way to proceed—the result was “higher overall recall and precision.”

But wait, there’s more.

The authors report that compared to the traditional technology-assisted review [TAR] tools, which rely on a manually harvested set of “training” documents, GPT-4 performs its own independent analysis of each document, which “leads to far more consistency in coding than we traditionally see with human review.”  Moreover, unlike TAR, GPT-4 can generate explanations about why it made the responsiveness determination it did, “which will likely expedite the QC process by allowing a human reviewer to more quickly confirm that a document is responsive.”

Keeping in the forefront of our minds that this is the first publicly reported BigLaw experiment with GAI and we’re all in the earliest of days on this front, it’s premature to draw any drastic conclusions.  But I confess that I find GAI’s ability to perform associate-level work impressive to the point of representing a conceptual fork in how lawyers and law firms perform a fundamental day to day function.

The advance of technology is relentless (I assume you’ve heard of Moore’s Law?), but if you’re a squishy and highly oxygenated grey-matter powered intellect, the good news is that—so far at least—in human and economic history, every advance that has displaced jobs has rather quickly been accompanied by new opportunities for human beings to perform functions and fill occupations hitherto impossible or unknown–in equal or greater numbers and at equal or greater income levels.

This is a structural and economic observation, not a personal or family/society observation.  Not too many laid-off coal miners will end up with the retraining to construct massive wind turbines.  And we already know of more than one Maroon BigLaw firm drawing up plans to substantially cut their entry-level hiring classes of associates–“substantial” in this case meaning by one third plus.

Assuming that  GAI proves to have half the potential being claimed for it, will the BigLaw firm of the future really be populated with 50% lawyers and 50% everyone else?  Does ~30% lawyers, ~30% BigData/GAI gurus, and ~40% “everyone else” sound plausible to you?

Those of you who may have already begun investigating the feasibility of using ChatGPT in your own firms have probably already discovered that installing and using it costs real $$.  Talking to a source last week who’s deeply embedded in this world said a figure of $50,000/month as a baseline license fee was customary.   Now, $600,000/year may strike you as far-fetched or implausible for a law firm to take on, but, as they say, “do the math.”  A mid-level to senior associate will cost ~$400,000/year in salary alone–excluding bonus, benefits, support staff costs, real estate square footage, etc.  Call it, all-in, $600,000.  (You with the green eyeshades out there–this is not a second decimal point accuracy calculation, so stop caviling.)

Given some allowance for Kentucky Windage, for all purposes indistinguishable.  And did we mention that the silicon version never gets sick, goes on vacation, gets tired or grouses at working weekends?  And is  100% current on new developments and always will be?

Discovered at an archaeological dig in Greenland as global warming melted the icesheet.

[i] How Some Big Law Firms Can Afford Associate Salary Raises, ALM, published online at  https://www.law.com/2023/12/12/how-some-big-law-firms-can-afford-associate-salary-raise