AAA Opens AI Arbitrator to Commercial Disputes in $29 Billion Market

The American Arbitration Association opened its AI Arbitrator to two-party, documents-only business-to-business commercial disputes on November 9, 2026, extending a system that previously handled only construction cases. The expansion targets a large pool of work: AAA's commercial-services page reports more than $29 billion in B2B claims and counterclaims in 2025. The commercial cases remain untested with the tool, and the efficiency figures AAA cites come exclusively from construction matters.

The announcement matters because it moves AI-assisted adjudication from a niche construction docket into the mainstream of commercial arbitration, where costs and delays are chronic complaints. Among AAA's awarded commercial cases closed in 2025 with claims above $1 million, the median time to award was 18.7 months. Separately, 52% of closed B2B cases settled before an award. Those figures describe the broader commercial caseload, not AI Arbitrator matters — a distinction AAA itself draws and one that readers should keep in view throughout.

How does the AI Arbitrator actually work?

Under the workflow, parties submit position statements, evidence, exhibits and legal authorities. The AI summarizes the claims, the documents and the law, and each side can validate or correct the summary of its own submissions. The system also builds a timeline and structured case summaries, which parties can review while sending any feedback to the human arbitrator.

Parties do not see the AI's key-issues analysis or its draft award. A specially trained AAA human arbitrator reviews those materials along with the party feedback, may revise or rewrite any part of the draft, and exercises independent judgment before finalizing, signing and issuing the award. AAA states plainly that the tool is not an autonomous decision-maker: the arbitrator is responsible for detecting and correcting deficiencies and remains accountable for the outcome.

Participation is consensual. Both parties must opt in; without that agreement, a case otherwise subject to AAA arbitration moves to traditional AAA administration. The service is also confined to documents-only commercial and construction disputes without live witnesses or complex factual issues, so the expansion does not encompass every commercial case. In plain words: the AI drafts, a human decides, and only parties who agree to the process are subject to it.

What did the construction phase prove?

The service debuted on November 3, 2025 for two-party, documents-only construction cases. That version was trained on AAA construction awards and refined with expert-labeled examples. AAA's technical overview for the commercial expansion describes domain-specific training on its commercial and construction awards and case data, refined through expert labeling and calibration rather than based only on generic legal text. For the new rollout, AAA says the technology was trained on cases from more than 20 industries, including healthcare, financial services, energy and technology.

Diana Didia, AAA's executive vice president and chief technology and innovation officer, explained the sequencing: "By introducing the technology first for a narrower set of construction cases, we could continue testing and evaluating its performance, refining its capabilities, and strengthening the governance processes supporting its use."

The construction-first approach was cautious by design, but it also means the performance evidence AAA can point to is narrow. Accuracy is evaluated against benchmark cases with human-issued awards and monitored through system evaluation and arbitrator review. AAA frames the goal as alignment with experienced-arbitrator reasoning and human parity, rather than identical outcomes — a carefully hedged standard that leaves undefined what "parity" means in practice.

What are the costs and deadlines?

The AI-Led Arbitration Rules, amended and effective February 18, 2026, set the arbitration cost at $2,500 for a claim and separately for a counterclaim, covering AAA administration and the human arbitrator's compensation. Parties may also agree at the outset to an advisory, non-binding award instead of a binding one.

Once submissions and party feedback are complete, the award is due within five business days after the feedback deadline. AAA can extend that date only in unusual and exceptional circumstances. That deadline applies to the final stage only; it is not a commitment to resolve the entire dispute in five business days, and the source is explicit on this point. For parties weighing the option, the fixed $2,500 fee is the firm number; the end-to-end timeline remains an open question.

How fast and cheap is it, really?

AAA's FAQ reports early results of 20–25% faster resolution and cost savings of at least 35%. Those figures are expressly limited to two-party, documents-only construction disputes. They do not establish the time or cost performance of the newly eligible commercial cases, and AAA does not claim they do. This is the central evidentiary gap in the announcement: a construction track record attached to a commercial product.

The savings figures also come from AAA itself, the party with the strongest incentive to promote the service, and the FAQ does not specify the sample size or methodology behind them. Independent verification is absent from the announcement. For now, the commercial expansion rests on a defined price, a human decision-maker, and efficiency evidence borrowed from a different docket.

What are the data and governance safeguards?

AAA says the platform operates in a closed AI environment: case information and AI-generated outputs are not used to train or expand the underlying model, and foundational OpenAI models — the technology's base — do not ingest case data. The disclosure confirms, incidentally, that the system is built on OpenAI's foundation models rather than AAA-built architecture. Case-data controls align with SOC 2 Type 2 standards, an auditing framework for security, availability and confidentiality of service providers.

An AI Governance Committee oversees compliance, ethics and model outputs. Human arbitrators must disclose circumstances that could reasonably raise doubts about their impartiality or independence — a standard disclosure duty extended into the AI-assisted context. What AAA does not detail is how the committee reviews model outputs in practice, how often arbitrators substantially revise AI drafts, or what happens when an arbitrator and the system diverge on the merits. Those operational questions are the ones commercial parties will need answered before opting in at scale.

What is the so-what?

The commercial expansion pairs a defined filing cost and an accountable human decision-maker with efficiency evidence that remains specific to construction. The open question for the broader rollout is whether commercial cases can achieve comparable savings while preserving the parties' consent and review rights and the arbitrator's responsibility for the final award. AAA has now built the machinery to answer that question; the caseload of the coming months will supply the data.