Behind the AI Arbitrator: The Governance Work Supporting Its Expansion into Commercial Disputes

When the American Arbitration Association® (AAA®) introduced the AI Arbitrator, it brought a new approach to dispute resolution, combining AI-powered analysis with the judgment and oversight of an experienced human arbitrator. Initially available for eligible two-party, documents-only construction disputes, the AI Arbitrator is now expanding to eligible two-party, documents-only business-to-business cases across industries.

From the early stages of development, the AAA took an intentional phased approach to introducing the AI Arbitrator, beginning with a small subset of cases before expanding to additional industries. This expansion required extensive training, testing, and evaluation to assess the technology's performance across different dispute types, while maintaining the legal, procedural, and ethical safeguards that underpin traditional arbitration. Throughout this process, the AAA brought together practicing arbitrators, legal experts, technology experts, and AI governance professionals to evaluate the technology and integrate governance throughout its development, deployment, and ongoing operation.

A key part of the AAA’s expansion approach was preserving the role of the parties and the human arbitrator throughout the process. Under the AAA's AI Led Arbitration Rules, both parties must agree to use the AI Arbitrator and have opportunities to present their positions, review AI-generated summaries of their submissions, and provide feedback for the arbitrator to consider. The human arbitrator reviews the case materials, party feedback, and AI-generated recommendations, revises as needed, and issues the final award. This process allows AI to support arbitration while preserving human judgment and oversight.

The Work Behind the Expansion into Commercial Disputes

Preparing the AI Arbitrator for commercial disputes involved more than making the existing technology available to a broader range of cases. Commercial disputes span a wide range of industries, contractual relationships, legal questions, and factual circumstances, and the AAA needed to confirm the technology could perform across these varying circumstances.

As part of this process, the AAA assessed cases based on factors such as the number of legal questions, choice-of-law considerations, and the number of parties and claims involved, allowing the team to evaluate the AI Arbitrator across cases of varying complexity. Experienced arbitrators worked through previously closed commercial cases using the AI Arbitrator, reviewed its work, and provided feedback at different stages. To examine consistency, the teams ran identical case materials through the system multiple times, made refinements, and retested cases after refinements to evaluate changes in the output. Members of the AAA's Legal team also independently evaluated cases using the AI Arbitrator, examining the AI's analysis and comparing arbitrator feedback with the underlying case records. This cross-functional approach to testing, evaluation, and refinement helped inform changes the technology needed before the AI Arbitrator was ready to be introduced for commercial cases.

An important part of this evaluation was determining whether the AI Arbitrator's analysis and outputs were supported by the case record and underlying legal reasoning. In traditional arbitration, two experienced arbitrators may reach different conclusions based on the same evidence, contract provisions, and applicable laws, and both decisions may be reasonable. Applying that same principle when reviewing the AI Arbitrator’s analysis and outputs, the AAA considered whether a reasonable human arbitrator could reach the same conclusion based on the case record and supported reasoning, rather than evaluating the technology against a single expected outcome.

Throughout testing and refinement, the AAA's Legal and AI Governance teams worked closely with developers to assess the technology from multiple perspectives. The Legal team helped distinguish legal reasoning issues from technical concerns, reviewed testing feedback, and evaluated proposed refinements. Legal and appropriate subject-matter teams also reviewed material changes to the AI's instructions or legal guidance before deployment. By integrating these perspectives throughout development, the AAA incorporated legal, procedural, and governance considerations directly into the technology's refinement, rather than treating them as separate review stages.

Governance Continues Beyond Launch

For the AAA, governance extends throughout the entire lifecycle of a product, from initial development through deployment and ongoing operation. The collaboration among Legal, AI Governance, and development teams continues after launch, helping guide refinements and evaluate their potential legal, procedural, and practical implications. The AAA's governance framework provides for ongoing monitoring and evaluation of the AI Arbitrator to identify areas that may require additional testing, review, or refinement. This includes:

  • Assessing the quality and consistency of the AI Arbitrator's analysis, including how its reasoning compares with that of human arbitrators.

  • Examining how the technology treats claimants and respondents and evaluating outputs across different types of parties.

This ongoing oversight matters for parties, counsel, and organizations evaluating whether AI-enabled arbitration fits their disputes. Beyond understanding how the technology works, they need to understand how it is tested, how parties participate, how information is protected, who makes the final decision, and what safeguards remain in place after launch. Data protection is a critical part of the AAA’s governance approach. Case materials and AI-generated outputs are not used to train or improve the underlying AI model, and access to case information is limited and secured under the confidentiality provisions of the AAA's AI Led Arbitration Rules. Together with the system's reliance on party-submitted materials, these safeguards establish clear boundaries around how case information is accessed and used.

As the AI Arbitrator expands to commercial contract disputes, the AAA continues to apply the same principles that have guided its development: rigorous testing, ongoing evaluation, data protection, and meaningful human oversight. By embedding governance throughout the technology's lifecycle, the AAA is building on its century of dispute resolution experience while continuing to evaluate and refine how AI can responsibly support arbitration.

Learn More about the AI Arbitrator

November 18, 2026

Discover more

Behind the AI Arbitrator: The Governance Work Supporting Its Expansion into Commercial Disputes

Building AI Governance That Works: Eight Lessons from Putting Principles into Practice

Beyond the Transcript: What AI Speech-to-Text Quality Really Means for Legal and Arbitration Professionals