The Advent of AI-Assisted Arbitration

The American Arbitration Association is at the forefront of artificial intelligence-assisted arbitration. For those involved in construction disputes who may be concerned about delay, lost time, and costs in lower-value cases, the platform is notable for both what it offers and what it does not supplant. The process offers a defined, rule-based framework to accelerate the resolution of straightforward claims, conserve costs, and produce reasoned awards that remain, in the end, the product of a human arbitrator’s judgment.

A wealth of information is available about the platform on the AAA’s website at https://www.adr.org/ai-arbitrator/. And the AI Led Arbitration Rules are available at https://www.adr.org/media/rzthzgco/ai-led-arbitration-rules-2026.pdf.

The steps are straightforward, with no witnesses and no final hearing involved. First, parties submit case materials and validate AI-generated summaries of their respective positions. Second, the AI program analyzes the submissions, identifies key issues, and generates a structured case narrative and analysis for parties to additionally validate or comment on. Third, the AI arbitrator generates a draft award with reasoning tied to the parties’ submissions. Of note, the AI arbitrator is proprietary—trained on actual arbitration awards and reflecting logic employed by arbitrators in a multitude of settings. Finally, a human arbitrator reviews the analysis, incorporates party feedback, and evaluates the outcome, then finalizes and issues the award. Considering documents-only construction cases, which are typically smaller-value disputes, the AAA estimates approximately 35 percent cost savings and 20 to 25 percent time savings.

The first check on avoiding a flawed result rests with the parties and their counsel, where they ensure that the submissions and outputs are correct. The second check on avoiding a flawed result rests with the appointed human arbitrator, where he or she corrects errors in the analysis and result.

Thus, as the process develops, it will not replace lawyers’ services. Attorneys can play an active, and perhaps critical, role in navigating the platform, ensuring the accuracy and sufficiency of submissions (and avoiding submission of irrelevant matters), and guiding their clients through the process. The prospect of a faster outcome does not necessarily mean forgoing the value that counsel can provide. Similarly, the process will not replace the independent judgment of an arbitrator. As noted above, a human arbitrator is still involved. The AI program drafts a proposed award, while a human arbitrator decides the result, whether agreeing with or modifying the proposed award generated by the AI program. It is the pre-hearing administrative matters, hearing time, and study time, and attendant costs, that are saved by the parties. Furthermore, not every case will be suitable for AI-assisted arbitration, which is one reason that the platform has been promoted as well suited for lower value claims. It is those cases, where the cost burden can rapidly eclipse the claim value, where AI-assisted arbitration may be best suited. For the other cases, human arbitrators are likely to be used in the long term.

However, for all of the potential benefits, there are trade-offs. In disputes where live testimony and the accompanying subjective credibility assessments are critical, an AI-led arbitration will not offer that opportunity. And while the human arbitrator review is designed to lessen the risk of a flawed result, it will take extended real-world application and use to test how consistently those safeguards perform. Users of AI-led arbitration will be joining others at the frontier of dispute resolution.

The risk of the unknown can be mitigated. For those interested in but leery of the process, an AIassisted arbitration may be used by agreement to produce an advisory, non-binding arbitration award—a potentially useful tool if settlement is a preferred outcome or to otherwise jointly evaluate risk. R-13 of the AI Led Arbitration Rules provides:

R-13. Non-Binding Arbitration

(a) The parties may agree at the start of the case that the human Arbitrator will render only an advisory, non-binding arbitration award in their case instead of a binding award.

(b) If the parties proceed with non-binding arbitration pursuant to this Rule, they agree that:

i) any non-binding arbitration award rendered shall not be introduced or cited as evidence in any other arbitration, litigation, or any other adjudicative body, and ii) any non-binding arbitration award shall not be entered as a judgment in any court, unless the parties agree otherwise.

Additionally, the AAA offers a Resolution Simulator (https://www.adr.org/resolution-simulator) to replicate the process as may be utilized by one side, with a result that may similarly assist in evaluating risk and developing case strategy. As the AAA announced earlier this year:

Designed for single party use in documents-only construction disputes, the Resolution Simulator delivers explainable, informational insights grounded in the AI Arbitrator’s structured legal reasoning, analysis, and decision logic. Parties may use it to understand how an arbitrator may resolve a dispute, develop their legal or negotiation strategy, or level-set internally with clients or counsel. It may also inform alternative paths to resolution, such as settlement or mediation. Ultimately, the Resolution Simulator brings legal teams insight before escalation, empowering strategic decision-making and informing a path to resolution.

AI-assisted arbitration is poised to continue to expand in scope and scale and may be appropriately considered in various instances. Attorneys will still have a role as valuable team members, and parties can rest assured that results are being decided by a human arbitrator, not simply being turned over to an AI program.

For more information, please contact R. Carson Fisk.