One of my pro bono activities for years was serving as a small claims court arbitrator in Manhattan. My best guess is that I handled about 1500 trials, and it turned out that I was quite good at getting the parties to settle, which was generally a better outcome than issuing an award to one side, which had the problems of collecting and aggravating hostilities between the parties. It was far more art than science.

The American Arbitration Association announced that it’s introducing AI arbitration into its mix.

“The legal industry has long been cautious about adopting new technologies, but as demand for fair, efficient, and accessible justice grows, innovation is no longer optional – it is essential,” said Bridget Mary McCormack, president and CEO of the AAA-ICDR.

“Now is the time to embrace AI to drive positive change through speed, efficiency, and accuracy. Our AI arbitrator can deliver timely and transparent outcomes that meet the speed of today’s society. For nearly 100 years, people have trusted us to resolve disputes fairly, and today we carry that trust forward by transforming arbitration to meet the demands of a digital-first world.”

A long time ago, Keith Lee made a very good point about the relationship of law and innovation.

It’s not that lawyers are anti-technology, it’s that they are anti-bullshit.

Pushing vapid hype is nothing new, and nothing that hasn’t proven wrong before. Can an AI arbitrator “deliver timely and transparent outcomes that meet the speed of today’s society”? If speed is the only measure, then AI will probably do far better than human arbitrators. After all, AI can process and spit out a ruling in seconds, maybe less. And boom, dispute resolved. After all any outcome is an outcome, whether it’s right, wrong or off-the-wall.

But speed isn’t the only criteria for good arbitration. Notably, arbitration came into existence with the intention of being an inexpensive, prompt mechanism to resolve disputes. Naturally, it didn’t always turn out that way, promptly assuming most of the trappings of courts and getting bogged down in the procedural nuts and bolts that made litigation a nightmare. Speed matters, and AI would almost certainly make arbitrations move far more quickly than they do under humans. But speed isn’t the only thing that matters.

Grounded in legal reasoning as the foundation for its decision-making, the AI arbitrator was trained on actual arbitrator reasoning from AAA-ICDR construction cases and calibrated and trained with human arbitrator input. With each step of the dispute resolution process, the AI arbitrator will evaluate the merits of claims, generate explainable recommendations, and prepare draft awards that will be benchmarked to maintain alignment with expert human legal judgment. A human-in-the-loop framework embeds human arbitrators to review reasoning, evaluate and, if needed, revise AI-driven outcomes before a decision is finalized, and validate results, safeguarding trust, transparency, and due process.

Initially, the use of AI will be limited to document-only construction cases, which have a particular need for swift resolution. But then there will be a “human-in-the-loop” to oversee the AI arb’s outcome. After all, AI is known to hallucinate, and at least based on my experience testing legal AI, it tends to get some very basic concepts and doctrines very wrong. Does a decision by a Boston state trial court take precedence over a Fifth Circuit federal decision? AI wouldn’t know because it only cares which serves to better support its outcome.

If the speed of AI becomes subject to human review, has the most significant benefit of AI decision-making been undercut? Is this any different than having a human arb handle the case, using AI to draft a decision which the arb then reviews, assuming the arb isn’t too lazy to bother and just lets AI out on its own?

“With our AI arbitrator, we set out to create a solution that can emulate human judgment, and with human arbitrator validation can provide a whole new path to dispute resolution,” said Diana Didia, EVP and chief technology and innovation officer of the AAA-ICDR. “By drawing on nearly a century of ADR expertise and our experience bringing innovative AI solutions to alternative dispute resolution, we built a platform that delivers consistent, transparent results where human legal reasoning is the foundation for the AI’s functionality. Just as importantly, we placed ethics at the center, supported by strict governance, AI standards, and robust validation frameworks, because trust and accountability must guide every innovation in justice.”

Lots of pretty words, all to suggest the wonder of technology making arbitration faster, consistent and better. While it starts with document-only construction cases, if parties embrace it, expect it to be rolled out to others, and eventually all cases. It will likely be fast. It will likely be less expensive. But will it be any good? Alternative dispute resolution is more art than science if its to fulfill it’s purpose. Artificial intelligence may eventually prove capable of fulfilling the science, but whether it can perform the art remains to be seen.