Artificial intelligence (AI) is rapidly transforming industries worldwide – and now, it is reshaping the field of international arbitration. The American Arbitration Association (“AAA”) and its international division, the International Centre for Dispute Resolution (“ICDR”), are pioneering this revolution with the launch of an AI-powered arbitrator dedicated to construction disputes. Beginning in November 2025, this initiative aims to make arbitration faster, more affordable, and more transparent for lower-value, document-only construction cases.
This initiative marks a significant milestone in alternative dispute resolution (ADR), with the AAA-ICDR promising to deliver “fast, cost-effective, and trusted dispute resolution” in industries where time and efficiency are of paramount importance.[1] The AI arbitrator will start with document-only construction cases – an area where timely decisions are essential to keep a project moving.[2]
Still, even with its promise, this innovation raises important questions about transparency, accountability, and how much reasoning we can realistically expect from a machine. Can technology ever mirror the judgment and sensitivity that human arbitrators bring to complex disputes? Who takes responsibility if the technology gets it wrong? And can an algorithm match the experience and intuition of a seasoned arbitrator? At its core, this is about how much control we as humans are comfortable giving to machines.
What Is the AI Construction Arbitrator?
The AI Construction Arbitrator is a machine-learning system designed to assist in resolving low-value, document-only construction disputes. Developed by the AAA-ICDR in partnership with QuantumBlack, AI by McKinsey, it reviews pleadings, evidence, and contract documents to prepare a draft award that a human arbitrator then reviews, finalises and, most importantly, signs.[3] The system was trained on more than 1,500 real construction awards from the AAA’s archives and is supposed to replicate the actual decision-making process by real human arbitrators.[4] According to AAA-ICDR, it uses a structured legal prompt library and advanced conversational AI models to deliver legally sound and explainable recommendations. Human arbitrators remain in the loop, reviewing and validating each draft before it becomes an official award.[5]
While this approach may bring greater efficiency and consistency, it also raises understandable concerns. The AAA-ICDR says real arbitrators remain firmly in charge through its “human-in-the-loop” system, where a real human arbitrator has the final word. Still, it is fair to wonder how much arbitrators will come to rely on AI – and whether that reliance could slowly shift decision-making away from humans.
The system is built around “explainable AI” and a library of legal prompts, and is meant to keep its reasoning grounded in established legal reasoning. But as many observers point out, even the clearest algorithm cannot match the context, empathy, and discretion that experienced arbitrators bring to complex disputes.[6] The AAA-ICDR’s model is certainly a major step forward, yet it also opens a much-needed discussion about how far automation should go in shaping human decisions and judgments.
Transforming Construction Disputes: Efficiency and Cost Savings
Construction disputes are often slow and costly. Projects stall, deadlines slip, and expenses rise as parties battle their claims in courts or arbitrations. To tackle that problem, the AAA-ICDR chose construction cases as the starting point for its AI Arbitrator initiative. Construction awards usually include detailed reasoning, which gives the developers a way to trace how real arbitrators think – how they weigh evidence, analyse arguments, and reach their conclusions.
At first, the AAA-ICDR’s AI Construction Arbitrator will focus on documents-only construction cases – typically lower-value disputes similar to those handled under the AAA Construction Rules fast-track procedures. Such disputes are well-suited for AI testing because they rely heavily on the documentary record, schedules, quantification, and contractual mechanics, rather than requiring the resolution of novel points of law. According to the AAA-ICDR, the system could cut arbitration costs by 30–50% and shorten case timelines by around 25–30%, offering real relief for both contractors and clients who need quick, practical decisions to keep projects moving.[7]
Human-in-the-Loop: Safeguarding Trust and Transparency
A key part of the AI Construction Arbitrator initiative is its “human-in-the-loop” approach. After the system reviews the submissions and drafts an award, a human arbitrator steps in to review, edit, and finalise it. This process makes sure that human judgment remains at the heart of the process while using AI to handle the time-consuming analysis.[8]
As Bridget Mary McCormack, CEO of the AAA-ICDR, explained, the process is transparent at every step. When parties submit their materials, the AI system “deconstructs their submissions, identifying claims, evidence and legal frameworks.” As McCormack further explains, this analysis is presented back to the parties for validation.[9] This validation step is interesting – addressing one of the most frequent complaints in arbitration: the feeling that decision-makers overlook what are, in the parties’ view, key arguments. According to AAA-ICDR, the AI Construction Arbitrator’s transparent breakdown ensures parties know they have been heard and understood before any decision is rendered.[10] Once all submissions are complete, a human arbitrator from the AAA’s permanent panel is appointed through a traditional round-robin system, maintaining the same disclosure and conflict-checking procedures that apply in any other AAA case.[11]
The AAA-ICDR maintains that the AI Construction Arbitrator’s reasoning is grounded in law rather than simple pattern recognition, claiming it has been trained on expert-labelled construction awards to replicate authentic arbitral reasoning and align with established legal standards. Yet, how deeply such training can capture the nuance of human legal interpretation remains uncertain.[12]
The design is intended to keep awards consistent with principles of fairness, confidentiality, and impartiality, but these assurances rest largely on institutional confidence rather than independent scrutiny. While the model reportedly adheres to the AAA-ICDR’s 2025 Guidance on Arbitrators’ Use of AI Tools – which insists that AI should assist, not replace, human decision-making – the real test will be whether that balance holds once the system is widely deployed.
Expansion Beyond Construction Disputes
Although the AI Construction Arbitrator starts with construction cases, as its name suggests, the AAA-ICDR plans to expand its use to insurance and payer–provider disputes in 2026 – areas that also deal with a high volume of smaller, document-heavy claims.[13] As the technology develops, it could eventually reach commercial and consumer arbitrations, so long as there is the parties’ consent, and fairness and transparency are preserved. If used carefully, AI could transform the way lower-value disputes are handled, allowing businesses to resolve routine cases faster and free up human expertise for more complex ones.
AI in Arbitration Practice
The AAA-ICDR’s AI Construction Arbitrator did not appear overnight – it is the result of years of steady innovation within the arbitration world. Over the years, arbitral institutions and practitioners have quietly adopted a variety of digital tools – from AI-assisted case management and document review systems to chatbots that guide users through procedural steps, AI-powered arbitrator selection platforms such as the AAAi Panelist Search, and software that can summarise submissions or respond to rule-based questions.
These developments are part of a broader trend. Organisations like the SCC, CIArb, and the ICC have each released or are developing AI guidelines to ensure that technology in dispute resolution is used ethically, transparently, and with proper human oversight. The emphasis across all these initiatives is clear: technology should support the arbitral process – not replace it.
Yet, as recent cases show, that line can be difficult to draw. As previously discussed, in LaPaglia v. Valve Corporation, one party asked a U.S. court to overturn an award, claiming the arbitrator had relied on ChatGPT to help draft the decision. The case exposed a deeper tension between efficiency and integrity: if parties start to believe that awards are shaped by undisclosed AI input, confidence in arbitration itself could begin to slip. It also raised a practical concern – that transparency about the use of AI may soon be just as important as disclosing conflicts of interest or prior appointments. The AAA-ICDR’s approach, with its human-in-the-loop design and disclosure requirements, is meant to prevent precisely that kind of problem. The lesson from LaPaglia v. Valve is clear: as technology advances, ethical boundaries must evolve alongside it. As arbitration becomes more digital, the challenge will be to preserve trust and procedural fairness without slowing down the efficiencies that make technology so appealing in the first place.
Ethical Considerations: Balancing Innovation with Oversight
The rise of AI in arbitration is certainly promising, but it also raises significant concerns. While technology may improve speed and consistency, it raises questions about bias, accountability, and enforceability. For instance, several arbitration laws explicitly require that arbitrators be natural persons. Article 1450 of the French Code of Civil Procedure[14] and Article 10(1) of the UAE Federal Arbitration Law No. 6 of 2018 are clear examples,[15] and similar provisions appear in the arbitration laws of Saudi Arabia (Article 14 (1)),[16] Egypt (Article 16 (1)),[17] and other jurisdictions. These statutes presuppose human capacity, integrity, and judgment – qualities that current AI systems cannot fulfil. This means that any award produced through an AI-assisted process could face challenges under the New York Convention if courts view it as inconsistent with national law.
The EU AI Act (Regulation (EU) 2024/1689) goes further, classifying legal decision-making as a “high-risk” activity that must include clear human oversight.[18] The AAA-ICDR may have addressed this approach: every final award must be signed and certified by a human arbitrator to preserve independence and due process.
Transparency is another concern. What certain observers call the “black box problem”, i.e., the difficulty of explaining how an algorithm reaches its conclusion, threatens the legitimacy of any decision-making process.[19] The AAA-ICDR’s use of “explainable AI” and validation steps is supposed to solve this problem, ensuring that parties can understand how evidence and arguments are assessed and ultimately, the decision reached. The challenge, of course, is that AI systems reflect the biases of the people and data behind them. The aim is not to erase bias altogether but to build safeguards that keep it in check, just as evidentiary rules try to do with human judgment.
Across the arbitration community, institutions like the SCC, CIArb, and AAA-ICDR have all published guidance underscoring the same principle: AI should support, not supplant, human judgment. Arbitrators are encouraged to use such tools responsibly and to disclose their use of AI in the interest of transparency.
Looking ahead, key questions remain unresolved: How will mistakes be corrected? Who bears responsibility for them? And what kind of regulatory or ethical frameworks should govern this space? For now, consent must remain the foundation – parties should be free to opt into AI-assisted arbitration, but not forced into it through one-sided contracts or fine-print clauses.
Conclusion
The AI Construction Arbitrator is an important step forward, but it also comes with real responsibilities. Used thoughtfully, it could make arbitration faster and more accessible. But its success will depend not on technology alone – it will depend on people, and on their willingness to use these tools carefully while upholding due process and transparency.
As the line between human judgment and digital assistance grows thinner, the legal community faces a simple but important task: to ensure innovation strengthens justice rather than weakens it.
[1] AAA-ICDR® to Launch AI-Native Arbitrator, Transforming Dispute Resolution, 17 September 2025.
[2] LawSites, AAA Readies November Launch of AI-Powered Arbitrator for Construction Disputes, 22 September 2025.
[3] A&O Shearman, AI as arbitrator for certain low-value construction disputes at AAA-ICDR, 3 October 2025.
[4] AAA-ICDR® to Launch AI-Native Arbitrator, Transforming Dispute Resolution, 17 September 2025. See also A&O Shearman, AI as arbitrator for certain low-value construction disputes at AAA-ICDR, 3 October 2025.
[5] AAA-ICDR® to Launch AI-Native Arbitrator, Transforming Dispute Resolution, 17 September 2025.
[6] A&O Shearman, AI as arbitrator for certain low-value construction disputes at AAA-ICDR, 3 October 2025; Hill Dickinson, AI arbitrators – A new era in international arbitration?, 29 September 2025.
[7] LawSites, AAA Readies November Launch of AI-Powered Arbitrator for Construction Disputes, 22 September 2025.
[8] AAA-ICDR® to Launch AI-Native Arbitrator, Transforming Dispute Resolution, 17 September 2025.
[9] LawSites, AAA Readies November Launch of AI-Powered Arbitrator for Construction Disputes, 22 September 2025.
[10] LawSites, AAA Readies November Launch of AI-Powered Arbitrator for Construction Disputes, 22 September 2025.
[11] LawSites, AAA Readies November Launch of AI-Powered Arbitrator for Construction Disputes, 22 September 2025.
[12] AAA-ICDR® to Launch AI-Native Arbitrator, Transforming Dispute Resolution, 17 September 2025.
[13] LawSites, AAA Readies November Launch of AI-Powered Arbitrator for Construction Disputes, 22 September 2025.
[14] Article 1450 of the French Code of Civil Procedure: “La mission d’arbitre ne peut être exercée que par une personne physique jouissant du plein exercice de ses droits.”
[15] UAE Arbitration Law (2018), Article 10.1(a)
[16] Saudi Arbitration Law (2012), Article 14(1).
[17] Egyptian Arbitration Law (1994), Article 16(1).
[18] AAA-ICDR® to Launch AI-Native Arbitrator, Transforming Dispute Resolution, 17 September 2025.
[19] B. Prastalo, Arbitration Tech Toolbox: AI as an Arbitrator: Overcoming the “Black Box” Challenge?, Kluwer Arbitration Blog, 23 August 2024.
