Deals Are Not Chess Games. They Are Marriages.

Arthur C. Brooks draws a distinction that I find myself returning to often. Complicated problems, he says, can be solved with enough computational horsepower. They have answers. You get there through effort, logic, and the right tools. Complex problems are different. They cannot be solved — only navigated. They involve human beings, relationships, meaning, irreducible uncertainty. You never fully arrive. You just keep going.
‍

The canonical example of a complicated problem is scheduling a flight. The canonical example of a complex one is a marriage.
‍

AI is extraordinary at complicated. Everyone who works with it knows this by now. It drafts, summarizes, cross-references, catches inconsistencies, and produces output that used to take hours in a fraction of the time. In earlier posts, I have written about how this changes the nature of legal apprenticeship and what it means for the profession's relationship to burnout. Those questions are real. But they stop short of something important. A transaction is not only complicated. It is also complex. And that second layer is where the outcome is actually determined.
‍

A business deal is not just a problem to be solved. It is a condition to be managed — between people who want different things, for reasons they will not always tell you.

‍

Consider what actually happens in a negotiation. Two or more parties come to a table. Each has a stated position and an unstated one. Each has a principal whose psychology shapes every instruction the lawyer receives. In a venture financing, the founder may say she wants the best terms, but what she actually wants is not to feel taken advantage of by the fund she is about to spend the next decade partnering with. The lead investor may say he wants founder-friendly governance, but what he actually wants is enough structural protection to survive the moment this goes sideways. Neither of those things is in a term sheet. Both of them determine whether the deal closes.
‍

The diligence, the documentation, the structure — that is complicated, and AI handles it well. But underneath all of it runs a current that is something else entirely. And complexity of that kind does not yield to computational horsepower.
‍

Legal entities are more complex than people

Here is where I think the AI-in-law conversation misses something important. A corporation is a legal person. But it is not a simple one. The law is the framework and the language in which all of it finds expression. It is an aggregation of multiple people, multiple psychologies, and multiple incentive structures acting simultaneously. When I represent a company in an acquisition, I am not representing a single decision-maker. I am navigating the dynamics between a CEO who is excited, a CFO who is cautious, a board that is divided, a management team that has not been told the full picture yet, and a seller who is watching all of it through a data room. The legal entity has a single signature line. What sits behind it is enormously complex.
‍

Even big proponents of AI have been careful about one domain: therapy. The reason is not that AI cannot produce empathetic-sounding text. It can, and does. The reason is that therapy is not a delivery of information. It is a relationship, and relationships are irreducibly human. The therapeutic value is not in what the therapist says. It is in the fact that a real person — with their own history, limitations, and genuine presence — is in the room with you.
‍

I think transactional law has more in common with that than the legal industry typically acknowledges. Not because deals are emotionally charged, though they often are, but because the work is fundamentally relational. It is not about producing the right document. It is about creating the conditions under which two parties who are fundamentally uncertain about each other decide they trust each other enough to move forward. That is complex. No model resolves it.
‍

Judgment is not a better algorithm

There is a version of the AI-in-law argument that says: we will use AI to handle the complicated parts, and lawyers will focus on judgment. I have made versions of that argument myself. But I want to be more precise about what judgment actually means in a transactional context, because I do not think it is simply a higher-order form of pattern recognition that AI will eventually reach.
‍

Consider what AI actually does at its best. It beat the world chess champion in 1997. It can solve protein folding problems that stumped biochemists for decades. It can price derivatives, optimize logistics networks, and identify legal risk across thousands of documents in the time it takes to pour a cup of coffee. These are staggering achievements. Chess, in particular, feels complex — the number of possible positions exceeds the atoms in the observable universe. But it is not complex in Brooks's sense. It has fixed rules, a finite board, and a clear winning condition. Given enough compute and enough games, it yields. AlphaZero mastered it by playing itself millions of times, converging on optimal play through repetition at a scale no human could match.
‍

A deal has none of those properties. There is no board. The rules shift mid-game. The winning condition is not agreed upon in advance, and it means something different to every person at the table. No amount of self-play gets you there, because the thing you are navigating is not a system. It is a set of specific human beings, in a specific moment, with specific fears and histories that no training data fully captures.
‍

This is the more fundamental limitation. AI is not an oracle. It is an extraordinarily powerful engine for arriving at conclusions that reflect the average of everything it has been trained on. That is genuinely useful — market-standard terms, precedent analysis, risk flags — and I rely on it. But the average is, by definition, not this deal. It is not this founder, this investor, this dynamic, this moment. Judgment in a transaction is precisely the capacity to depart from the average in ways that are right for the specific people in front of you. That is not a skill that accumulates through more data. It accumulates through presence, through repetition, and through the specific discomfort that comes from being personally responsible for an outcome that affects real people.
‍

Nowhere is this more apparent than in governance — in the negotiation of board composition, protective provisions, and control mechanics that will govern how these same people relate to each other when things get hard. That is worth its own conversation, and I will return to it. But the point here is simpler: the document is not the deal. The deal is the relationship the document is trying to formalize. Getting that right is not just complicated. It never was.
‍

Brooks says that complex problems cannot be solved, only navigated. That navigation requires a human being who can be present to the full texture of a situation, including what is not said, what is not in the document, and what the people involved do not yet know about themselves. A deal, at its best, is a relationship that becomes an agreement. It is complicated, yes. But it is also something more. And that something more is where lawyers earn their place — not just as drafters, but as navigators. That part is not going away.

‍
^{Lizbeth Flores is a corporate partner at PAG Law PLLC, where her practice focuses on cross-border technology, venture capital, and M&A transactions across Latin America. She has been named by Latinvex as one of Latin America's Top 100 Female Lawyers.}

‍

^{Disclaimer: This publication is provided by PAG Law PLLC for general informational purposes only and does not constitute legal advice or create an attorney-client relationship between PAG Law and the reader. The content reflects the views of the author as of the date of publication and may not reflect subsequent developments in law, regulation, or policy. Readers should not act or refrain from acting on the basis of any information contained herein without seeking professional legal counsel tailored to their specific circumstances and jurisdiction. PAG Law expressly disclaims all liability with respect to actions taken or not taken based on any or all of the contents of this publication. This material may be considered attorney advertising in some jurisdictions.}

‍