A flood at the courthouse door, part three

They say a rising tide lifts all boats, and it will. The productivity gains AI brings to litigation will lift every firm that can afford the latest models, the well-equipped solo practitioner, and even the litigant who has no lawyer at all but has learned to ask the right questions. That is good news. But a rising tide only helps if you happen to be a boat. It is a different story for the fixed structures standing on the shore.

And no structure is more fixed than a courthouse. A court cannot hire 40 new judges by Friday. It cannot build a new wing out of an appropriation it does not have. It cannot, by an act of will, read twice as many motions in the same number of hours. So "rising tide" may be the wrong image. What is bearing down on our state and federal courts is closer to a tsunami. The productivity and insight AI brings to litigation will create two nightmares for the judicial system. But there is hope, and I will get to it.

The first nightmare is the self-represented litigant, supercharged. For as long as there have been courts, you could spot a pro se filing the moment it arrived. When I clerked for a federal district court, I could pick the unrepresented party's work out of the stack without looking at the names above the caption. The briefs lacked the polish and prose of a professional shop, even if the merits were there. That is ending. The engines now in that litigant's hands draw on terabytes of precedential pleadings, and they will produce work product that does not look much different from what the firm across the aisle filed that morning. The person who used to arrive with a grievance now arrives with a complaint--well-pleaded, properly captioned, full of authority--and a tireless AI counsel at his side, ready to proceed with fire and fury.

The second nightmare is the represented party. Clients and their firms have always had to weigh resources before filing a pleading or a motion, or taking any other step in a case. A party did not move for summary judgment on a long shot, because the long shot was expensive to chase. AI is about to drive the cost of every phase of litigation down dramatically, and behavior changes when the price of a gamble falls toward zero. If you can now draft and file a motion for summary judgment for a fraction of what it used to cost, why not take the chance? The same goes for discovery motions, discovery requests and every other piece of work product a case generates. The multifold productivity gain AI hands the lawyer will reach the courthouse as one thing: a sharply, and I mean sharply, higher workload.

What is to be done?

At a minimum, courts must be able to use the same technology the parties and their counsel have already started to integrate. It makes no sense for both sides to arrive armed with engines that read and draft at superhuman speed while the judge in the middle is left to keep pace with a pen, a law clerk and a pot of coffee. Give the courts AI for the work that is genuinely clerical--triaging dockets, flagging duplicative filings, organizing the record, drafting the routine order a judge then reviews, corrects and signs. The machine can be the court's clerk. It cannot be the court's judge, and I will come back to that. Beyond technology, state legislatures and Congress must fund the judiciary, and do it now, ahead of the wave rather than years behind it. A system that has been underfunded for a generation cannot absorb a flood. It needs more judges, more clerks and more courtrooms, appropriated quickly, before the backlog hardens into a denial of justice for everyone in line.

Funding alone will not fix a system whose incentives run the wrong way. Legislatures should also consider building new deterrents into the rules we already have. That might include broadening the fee shifting afforded by offers of judgement under Federal Rule of Civil Procedure 68 and its state equivalents. Another possibility is encouraging courts to more effectively sanction frivolous filings and lawyers who needlessly multiply proceedings. These sanctions are discretionary and rarely imposed to their full extent.

But there is one "solution" I want to head off. Some technologist will surface, any day now, the idea that parts of the judicial system can be put online and handed to computers altogether. The loudest voices will belong to a certain kind of billionaire (trillionaire?) who feels cheated by our jury system, has little respect for judges, and would happily see his disputes decided by a machine he believes he can understand and, perhaps, one day own. Letting AI adjudicate contested questions of fact would be a mistake. These systems are nowhere near able to weigh the fine nuances of a factual dispute--to decide who is telling the truth when two credible people swear to opposite things. And the jury is not an inefficiency to be optimized away. It is a body of citizens standing between the state and the individual, and the role it plays in American courts is essential to our democratic commitment to self-rule. Let the machine sweep the courthouse floor. Do not let it sit on the bench, and do not let it empty the jury box.

There are less drastic measures worth naming, too. Courts can meter the flow at its source with graduated filing structures and pre-motion-conference requirements--the judge who insists on a short letter and a phone call before any motion is filed already understands the idea. They can require disclosure and certification, codifying with real teeth the standing orders now spreading across the country that require counsel to disclose AI use and personally certify that a human checked every citation. They can divert volume away from the trial judge by expanding the use of magistrate judges, special masters and recalled senior judges, and by leaning harder on early mediation and court-annexed arbitration. And for the smallest disputes, they can build real online dispute resolution--platforms that help parties reach their own settlements with human oversight, which is very different from a machine handing down a verdict.

So I am not pessimistic. The same wave that threatens to swamp the courthouse also carries the tools to bail it out, if we are wise enough to put them in the right hands and keep them out of the wrong ones. Fund the courts. Give judges the machine as a servant, not a master. Price the marginal motion so litigants feel its weight.

In the fourth article, I will turn from the courthouse to the invoice and take up the question that quietly worries every managing partner: what artificial intelligence is about to do to the billable hour. Parts one and two can be found on the Daily Journal website.