The line is moving from AI helping lawyers argue a case to AI helping write the first version of how a dispute should be resolved.

WHAT’S HAPPENING

The American Arbitration Association is operating an AI Arbitrator that reviews submissions, identifies and analyzes key issues and prepares a draft arbitration award with reasoning.

The AI does not issue the final decision. Under AAA’s rules, a human arbitrator reviews the draft and can approve it, edit it or rewrite any part of it before signing and issuing the final award.

The system is being used for eligible documents-only disputes, initially focused on two-party construction cases. Participation is voluntary: both parties must agree to use the AI-led process.

WHY IT MATTERS

AI in law is no longer limited to searching cases, summarizing documents or helping attorneys prepare arguments.

Here, AI is participating much closer to the actual point of decision: evaluating the dispute and drafting the proposed outcome.

The human arbitrator still controls the judgment. But the first structured version of that judgment can now originate with a machine.

That creates a different question for the legal system: How much can AI shape a decision before assisting the decision-maker begins to look like influencing the decision itself?

WHO BENEFITS

Parties with relatively straightforward, document-heavy disputes could benefit from faster and less expensive arbitration.

AAA says its early results show 20–25% faster resolution times and cost savings of at least 35% in two-party, documents-only construction disputes. Those figures are AAA’s own reported early results, not an independent industry-wide finding.

Human arbitrators may also spend less time organizing large amounts of information and more time reviewing the reasoning and deciding whether the proposed result is correct.

WHO LOSES

The biggest risk is automation bias: the possibility that a well-structured AI recommendation receives more deference from a human than it deserves.

That concern is already recognized elsewhere in arbitration. The Chartered Institute of Arbitrators says arbitrators should not surrender their decision-making authority to AI, should independently verify AI-generated information and remain responsible for every aspect of an award.

Efficiency only works if human review remains genuine rather than becoming a rubber stamp.

WHAT HAPPENS NEXT

The important question isn’t whether AI can assist arbitration anymore.

It already does.

The question is how far this model expands—from straightforward document-based disputes into larger and more complicated cases—and what safeguards follow it.

AAA says additional uses are planned. And its current system establishes an important precedent:

AI can help construct the proposed decision. The human still owns the final judgment.

For now, that distinction is the line keeping artificial intelligence on the assistance side of the bench.