The American Arbitration Association–International Centre for Dispute Resolution (AAA-ICDR) has announced the launch of an AI-native arbitrator for low-value construction disputes.[1] Starting November 2025, parties will be able to opt into a procedure where their case is supported – though not decided – by artificial intelligence.
This bold experiment deserves attention. Below, we explore how the AI arbitrator works, its potential benefits and risks, and what this means for the future of dispute resolution.
How It Works: Human-in-the-Loop, Not Robot-on-the-Bench
Let’s get one thing straight: this is not iRobot. The AAA-ICDR’s AI arbitrator is not a sentient machine making autonomous decisions. Instead, it operates within a human-in-the-loop framework. The AI supports – but does not replace – the human arbitrator.
Here’s how the procedure unfolds:
- Initially, the AI arbitrator will be available only for documents-only construction disputes, likely capped at $25,000 or less. These cases tend to be technical, repetitive, and data-heavy – ideal conditions for machine learning.
- Parties will still appoint a human arbitrator. The AI arbitrator addition requires mutual consent. Without it, the case proceeds as usual.
- The AI is trained on reasoning frameworks derived from over 1,500 annotated construction cases. It analyzes submissions and exhibits, deconstructs claims and legal arguments, and presents the human arbitrator with a dashboard of organized summaries, timelines, and claims analysis. It also generates recommendations and drafts the arbitral award – subject to human review and finalization.
- While the current system is limited to low-value, documents-only cases, the AAA-ICDR is already exploring future iterations that could handle more complex disputes, including those involving witness testimony and cross-examination.
All in all, the AI arbitrator is a tool in the arbitrator’s hands, not a substitute for one.
Benefits and Risks
There are compelling reasons to welcome AI into arbitration:
- AI can process information much faster than humans. It can scan thousands of cases and hundreds of pages of submissions in seconds, identifying patterns and relevant precedents with precision.
- The AAA estimates that the AI arbitrator could reduce costs by 30–50% in eligible cases. That’s significant – but the real promise lies in access to justice. According to AAA-ICDR President Bridget Mary McCormack, 92% of Americans can’t afford legal help for civil disputes. Many simply give up. By lowering the cost and complexity of arbitration, AI could unlock resolution for thousands of otherwise abandoned claims.
- Construction disputes are a smart starting point. They’re often technical, document-heavy, and time-sensitive. Delays can stall projects, inflate costs, and disrupt supply chains. The volume of similar cases also provides fertile ground for machine learning.
But the rollout raises important questions:
- Will courts uphold awards drafted by AI – even with human validation? We believe yes, especially since party consent is required. But this will be tested.
- Can parties truly understand how the AI reached its conclusions? That’s harder. AI can provide reasoning, but it often remains a black box. Then again, so do many human decisions. The difference may be more philosophical than practical.
- What obligations do human arbitrators have when relying on AI-generated reasoning? There’s a psychological tendency to over trust AI, which could lead to errors going unnoticed. The AAA-ICDR’s Guidance on Arbitrators’ Use of AI Tools emphasizes that arbitrators must retain full control – but as AI becomes more sophisticated, the line between assistance and delegation may blur. This blurring may also result in legal challenges to an award.
Conclusion: A Necessary Step Forward
It’s encouraging to see arbitral institutions embracing AI. This is the right mindset – despite legitimate concerns about transparency, accountability, and control.
In a recent article published in Germany,[2] David Weiss and Joerg Risse proposed an even more radical idea: a Robo Judge for passenger rights disputes. In their model, the AI would decide cases autonomously, with parties retaining the right to appeal to a human judge. Compared to that, the AAA-ICDR’s AI arbitrator is a cautious first step.
But it’s a step in the right direction. Long-term, autonomous AI decision-making should be possible – if parties want it. The AAA-ICDR’s initiative doesn’t get us there yet, but it brings us closer.
[1] AAA-ICDR® to Launch AI-Native Arbitrator, Transforming Dispute Resolution.
[2] BB 2025, 1731.
