How AI is Transforming Mexico’s Courts: The Case for Innovation

Originally published as an Expert Contributor column in Mexico Business News — Tech, Apr 15, 2026. Reproduced here in full.

Jorge has been a judge in a civil court in Mexico City for seven months and has not issued a single ruling.

Zero. None. Seven months, and his ruling count is still blank.

This should be a red flag. In the Mexican judicial system, backlog is the norm, not the exception. A judge who doesn't issue rulings sounds like a judge who isn't working, or who is hiding behind a mountain of unresolved case files. But Jorge is, by any metric you want to use, the most productive judge in his court. He gets through 11 or 12 hearings a day — more than any of his colleagues. His records are ready before noon. He leaves early. He has no backlog.

How is it possible that the judge who issues no rulings is the one with the least backlog?

To understand that, you first have to understand the courtyard.

Jorge's Court

In Jorge's court there are eight judges. Every morning, before the hearings begin, they're all there: sitting in a courtyard, reading case files. Four hours. Sometimes more. They read folios, underline, try to retain the names of the parties, the dates, the arguments. It's a work of memory and endurance. If you ask them why they do it that way, they look at you strangely. It's like asking a surgeon why they wash their hands. It's always been done this way.

Jorge doesn't sit in the courtyard. Jorge arrives five minutes before the first hearing. Sometimes in sneakers. He puts on his robe and starts.

What he does is feed each case file into an artificial intelligence model. Before the hearing he has a chronological summary, a table with the key points of each party, the applicable laws organized by topic. He walks in knowing more about the case than the lawyers themselves. If something comes up that he didn't anticipate, he calls a five-minute recess, consults, and comes back.

When he explained this to his colleagues, one of them looked at him as if he'd told her he read tarot cards. "That's illegal," she said. "Why are you doing that?"

A Lesson From History

In 1847, a Hungarian physician named Ignaz Semmelweis discovered that women were dying of puerperal fever in Vienna's hospitals because doctors were not washing their hands between the autopsy room and the maternity ward. The solution was so simple it was offensive: wash your hands with a chlorine solution. Semmelweis implemented it in his ward and mortality dropped from 18% to 1%.

His colleagues rejected it. Not because the evidence was weak — it was overwhelming — but because accepting that washing hands saved lives meant accepting that they, for years, had been killing patients. The solution wasn't difficult to implement. It was difficult to admit.

Semmelweis died in a psychiatric asylum at 47. It took medicine another 20 years to adopt what he had proven.

Sociologists who look at innovation have a name for this. They call it the "Semmelweis reflex:" the instinctive resistance to a new idea, not because it is wrong, but because its correctness threatens the professional identity of those who hear it. It isn't ignorance. It's something deeper. It's the psychological cost of admitting there was a better way and you weren't using it.

"That's illegal. Why are you doing that?"

AI's Impact

In September 2023, a team of researchers from Harvard, Wharton, and MIT published what remains to this day the most rigorous study on the effect of artificial intelligence on knowledge workers. They took 758 consultants from Boston Consulting Group — not interns, not juniors, but experienced consultants who charge hundreds of dollars an hour — and divided them into two groups. One was given access to GPT-4. The other was not.

The results were not subtle. The consultants with AI completed 12% more tasks, 25% faster, and with 40% higher quality. But the most revealing finding wasn't that. It was that the consultants who scored lowest at the start — the least experienced, the ones who struggled most — improved the most: a 43% jump in performance. AI didn't just make the good ones more productive. It turned the mediocre ones into good ones.

Behavioral economists have a concept for what AI did for those consultants. They call it "discretionary time:" the portion of your day not consumed by mechanical tasks, the space where you can actually think, create, decide. For most knowledge workers, this time is surprisingly small. AI didn't make them smarter. It gave them back time to use the intelligence they already had.

Now think about the eight judges in the courtyard. Four hours every morning reading folios. That isn't discretionary time. That is mechanical work dressed up as preparation. And when you finish those four hours, you arrive at the hearing exhausted, with a general sense of the case file but without the precision you'd need to do something truly useful with that knowledge.

Jorge recovered those four hours. And what he did with them is what turns this productivity anecdote into something far more interesting.

Eliminating Third Parties

Because when you truly know a case file — not from having read it in a rush, but from having broken it down, from being able to see each party's position with a clarity that normally takes weeks to achieve — you can do something that most judges simply don't have time to attempt.

You can sit with the parties during the hearing, in open court, and help them see where they stand. What they'll gain if they keep fighting. What they'll lose. How much time, how much money, how much wear and tear it will cost them. And you can do it with such precision — because you know every angle of the case — that the parties, time and again, choose to settle.

Jorge sends them to alternative dispute resolution with a concrete proposal. And it works. Every time.

Seven months. Zero rulings. Not because he avoids making decisions, but because the parties no longer need a third party to decide for them.

The most widespread fear about artificial intelligence in the courts is that it will replace the judge. That an algorithm will issue rulings. That justice will be automated, dehumanized, made cold. But what happened to Jorge is exactly the opposite. The tool didn't judge on his behalf. It freed up his time and attention to do the most difficult and most human thing a judge can do: convince two people to reach an agreement.

AI didn't replace the ruling. It made the ruling unnecessary.

Ethical AI for Justice

If you look at the timeline, what is happening in Mexican justice has a sequence that says more than any official speech.

In August 2025, Magistrate Juan Jaime González Varas did something no federal judge had done before: he issued a ruling — Civil Appeal 212/2025 — in which he disclosed that he had used artificial intelligence as an auxiliary tool to calculate collateral amounts. He didn't hide it. He documented it.

He published the methodology, the input data, the models he used, even the errors he found when running the same calculation on three different systems. And he established four minimum principles for the ethical use of AI in justice: proportionality, data protection, transparency, and human oversight. His legal theses were published in the Federal Judicial Weekly Gazette that same day. It was the first regulation of artificial intelligence in Mexico to arrive through a court ruling.

Months later, Jorge — who had probably never read those theses — began using AI to study case files and discovered something González Varas could not have anticipated: that the deepest benefit wasn't in the precision of the calculations, but in the time left over after making them.

On March 3, 2026, the Federal Judiciary issued Circular 1/2026: the first institutional document with formal rules for the use of AI in judicial activity. Principles of proportionality, transparency, human oversight. The same four principles González Varas had already established seven months earlier in a ruling.

First a magistrate did it and documented it. Then a judge did it and took it further than anyone expected. And finally, the institution wrote the rules for what individuals were already living.

Innovation never arrives top-down. It comes from those who dare to move first — and afterward, if we're lucky, the institution runs behind to give it shape.

The eight judges are still in the courtyard every morning. Not because they are bad judges — they are dedicated, responsible, and work hard. But they are trapped in a system that consumes all their time in the task of knowing the case file, and leaves them no room to do something with that knowledge.

Jorge still arrives five minutes early, in sneakers, with his little bag. He still issues no rulings. And the question no one has asked him — the one that truly matters — is not how he manages to leave early, but why the other eight haven't tried to do the same.

Jorge is a real judge. He uses AI exactly as the guidelines and circulars say it should be used. Without fear and without ignorance.

Jorge is mindful of privacy, uses dedicated accounts for this purpose, and takes all the necessary precautions for the correct handling of the information he manages.

If you're a judge, I urge you: Be like Jorge.