Inside complex commercial negotiations, AI is exposing the work that actually closes deals
I have been running complex commercial deals for long enough to know what gets them across the line. It is not the work that AI is being marketed to replace.
Most of the "AI-and-lawyers" commentary published in the last few years is written from outside the work of a hard-fought deal. The pieces describe what AI can do (draft, summarize, review, redline) and conclude that because those are large parts of what lawyers do, AI is coming for lawyers’ jobs. The more sophisticated commentary adds a qualifier: AI will absorb routine work, freeing lawyers to apply "judgment," "critical thinking," and "the human side of practice."
I agree. But the conversation stops there. Judgment, critical thinking, and other near-synonyms are used in the abstract almost as buzzwords. What is judgment, in practice, for a commercial lawyer? Why are some lawyers visibly better at the "human side of practice" than others? Can these things be taught? There's consensus in the commentary and in our practice that human capabilities are in high demand. But we have not defined what they are, nor have we unpacked what they look like in action.
The work behind these words is what's becoming most valuable as other parts of our practice are commoditized. Yet it is also the work the profession is least equipped to develop, because we have never described it well enough to teach it.
Visible vs invisible work
The visible work of commercial legal practice (e.g., research, document review, drafting, and redlining) has historically absorbed so much of a lawyer's time that the work above it has remained in the background. The invisible work was always doing the actual lifting; we couldn't see it because the visible work crowded the view.
AI is changing that. It's absorbing the visible layer quickly, and the invisible layer above it, the work that has always closed deals, is becoming the most legible part of what we do.
A useful frame comes from a recent podcast conversation between Tobi Lütke and Jess Hertz, the CEO and COO (and former general counsel) of Shopify. Lütke distinguishes crystallized intelligence (accumulated knowledge) from fluid intelligence, the live capacity to reason and decide in novel situations. AI, he argues, is making crystallized intelligence cheap. The relative value of fluid intelligence is going up.
This maps onto our work directly. Legal knowledge that used to take years to internalize is increasingly retrievable on demand. The capacity to apply judgment in the live moment of a negotiation is not. The first is becoming abundant. The second is becoming the job.
Defining invisible work
Sixty years ago, philosopher Michael Polanyi observed that we can know more than we can tell. He called this tacit knowledge: the kind that lives in practice rather than text. Economist David Autor later applied this to explain why decades of computerization had not led to the collapse of skilled employment. The codifiable tasks were automated; the tacit ones, like judgment, common sense, and the ability to read a situation, proved resistant.
A senior commercial lawyer in a complex negotiation does not primarily produce drafts. They intuit, from a counterparty's word choice, whether they are testing or telling. They recognize that a counterparty's real concern is not the clause being negotiated, but the precedent they fear setting internally. They take a deal that is "stuck" and reframe it – we have agreed on 90 percent; here is the path through the rest – without changing the facts. The deal moves because the story moved. These are not idiosyncratic moments. There is a pattern across them, and the pattern is teachable.
These are glimpses of a category. The category is the one Polanyi named, and it has stayed invisible because our profession has measured, evaluated, and rewarded the visible layer above it. We bill in hours attached to documents. We promote outputs that can be counted. We train juniors on this layer, not the layer above. The best senior lawyers do the invisible work superbly, by hard-won instinct. Juniors absorb it slowly, by osmosis, in the rooms they happen to be in. Nobody writes down what they're learning.
There is one observation that should have settled the question long before AI made it urgent. The most sophisticated commercial clients – principals who have done dozens of high-profile deals, who know their businesses better than any outside counsel could, who in many cases understand the commercial terrain better than the lawyers across the table from them – still rely on their lawyers for advice. Not for the drafting; they could find that anywhere. What they are looking for is harder to name and harder to replicate: the invisible work, the judgment underneath the document, the read of a room that no model can produce. They have always known this is what they are paying for. Our profession now needs to catch up to what its best clients have understood all along.
What excellence will look like
If judgment is the new differentiator, the next step is to define what it consists of at the deal table. I think this work can be done, and that the categories are recognizable to anyone who has been involved in enough complex deals. The lawyers who will define excellence in commercial practice over the next decade will not be the ones who learn to use AI faster than their peers. They will be the ones who use AI well enough to free themselves for the invisible work and then build that work into a craft they practise deliberately.
None of this can be turned into a prompt. All of it can be named, described, and developed. The work of doing that – putting language to the craft that gets deals done and building a vocabulary the next generation can apprentice into – is what will distinguish the lawyers and teams that thrive in this era from those that mistake AI fluency for excellence.
We have agreed that human capabilities matter. The next conversation is the one that defines them.
