The boutique in the age of the machine, part two

This is the second of five articles on what artificial intelligence will do to the practice of law. The first article, "Why the litigator will outlive the machine, part one," was published on July 7, 2026.  

I am writing this in the middle of a recruiting cycle. We have two open positions, and the honest truth is we ought to have three. The work is coming in faster than we can shake hands.

At the same time--and this is the part that keeps me up--I am learning each week, sometimes each day, that the machine on my desk can produce the first cut of a complaint, a demand letter, a set of discovery responses, even the bones of a brief. It can do this with the kind of speed that, if you saw it done by a human, would make you wonder what they had taken.

We treat every word of that output with extreme care and a healthy paranoia. AI hallucinates, and a hallucination in a court filing is not a quirk--it is a headline a lawyer spends years living down. But the direction of the road is unmistakable. AI is going to be part of the litigation workflow whether we welcome it now or whether we are dragged to it later. The firms that get on the bandwagon early will be a long way down the road before the rest have figured out which way to walk.

I limit my observations to the litigation boutique, because that is what I run. I suspect a good deal of this travels, but I'll let the AmLaw firms speak for themselves.

So, five things every boutique must reckon with:

1.      The paralegal has never been more important. The fashionable prediction is that AI will retire the paralegal. Perhaps the people saying this have never worked in a litigation firm. The paralegal is, and increasingly will be, the gatekeeper between the machine's confident-sounding output and the document that goes out the door with your name on it. Every citation must be checked. Every fact must be verified. Every quotation must be matched to a real source that actually says what the machine claims it says. The cost of getting this wrong is not a slap on the wrist--it's public embarrassment with a long shelf life on Google. The paralegal's role has not shrunk, it has sharpened.

2.      Every attorney must be AI-native. I do not mean every attorney must be a prompt engineer. I mean every attorney must know, in their bones, when to reach for the machine and when not to, how to ask it the right question, and how to spot the moment it has confidently led them off a cliff. The attorney who spends six hours doing what the machine could have started in six minutes is going to discover, quickly, that the client will not pay for the extra five hours and 54 minutes. Efficiency has taken on a new dimension, and there is no point pretending otherwise.

3.      Training in technology is now part of training in lawyering. Litigation firms have always trained their associates in writing, research and oral advocacy. There has always been some technological piece--the research platforms, the formatting, the eternal struggle with the table of authorities--but it sat in the back of the room. It is now sitting at the head of the table. Our firm has been ahead of the curve on this since we opened in 2013. We were running on Asana when most law firms had never heard the word. Workflow systems are now routine; AI is simply the next layer. But the layer must be taught, deliberately and consistently, across the whole team--because a firm in which one lawyer is fluent and another is fumbling is a firm that produces inconsistent work product, and inconsistent work product is how reputations come undone.

4.      Cybersecurity has gone from a quarterly concern to a daily one. When we moved to Microsoft Office 365, we spent a great deal of time climbing the Microsoft security dashboard, locking down what needed locking down. That was the easy part. AI brings new perils that no dashboard yet covers. Drop the wrong document into the wrong chat window, and you may have just violated a protective order. Paste a privileged communication into a model that retains it, and you have done something the case law has not yet sorted out--and the case law on this is being written, right now, by people who do not yet know they are writing it. Every member of the team must understand: the machine bears as much peril as promise, and like any instrument of that kind, it cuts in directions you did not intend if you handle it carelessly.

5.      This is the one that matters most: do not let the machine think for you. This is the subtlest danger, and I suspect the one that will separate litigation firms that thrive from the ones that quietly slip. The early phase of a case--the strategy, the theory of the case, the choice of which doors to open and which to leave shut--is precisely the part of the practice where the human mind matters most. It is the part where creative, open-minded, slightly contrarian thinking finds the argument the other side did not see coming. If you let the machine draft your strategy, you will get the strategy the machine would draft, which is to say the strategy any competent lawyer might also draft, which is to say the strategy that loses to a better one. We do not let a lawyer at our firm begin a brief inside the machine. They sit down first--with a whiteboard, or a legal pad, or their feet up on the desk--and they think. Then, and only then, do they reach for the tool. The machine can help draft. It does not decide.

So where does this leave the boutique? In a remarkable position, actually. Smaller firms can adapt faster than larger ones--there is no committee, no twelve-month rollout, no consensus-by-attrition. We can train the whole team in a week. We can rewrite our protocols in a day. We can make the machine an instrument of the firm's judgment rather than a substitute for it. The boutiques that do this well will not merely survive the transition. They will be the firms most other firms are, in a few years, trying to look like.

In the meantime, I am still hiring. We need every good lawyer we can find. The machine, for all its cleverness, has not yet learned to take a deposition.

The third article takes up a question that troubles me more than any of the above: what artificial intelligence will do to the judicial system, and what, if anything, might be done about it.