Institutional AI tooling for arbitration moved from “experimental” to “operational” between 2024 and 2026. The American Arbitration Association now publishes and operates a small portfolio of AI tools — each with a different audience, scope, and governance posture. This article compares three: the AAA-ICDR AI Arbitrator, ClauseBuilder AI, and the Chat Book. We built the Chat Book with the AAA; the other two are AAA-developed.
This is a comparison page, not a marketing page. Where we cannot verify a specific product detail from AAA-ICDR public materials, we say so rather than guess. Readers evaluating these tools for institutional or counsel use should treat the AAA-ICDR’s own published documentation as the authoritative source for AI Arbitrator and ClauseBuilder AI feature claims; this page is a framework for thinking about the three tools side by side, not a substitute for institutional materials.
The three AAA-ICDR AI tools at a glance
The table below sets out the comparative dimensions worth thinking about before selecting or recommending a tool. Several cells are necessarily summary-level; the institutional materials carry the detail.
| Dimension | AI Arbitrator | ClauseBuilder AI | AAAi Chat Book |
|---|---|---|---|
| Audience | Tribunal / parties (institutional pilot) | Parties drafting arbitration agreements | Practitioners, advocates, students, self-represented parties |
| Function | AI-supported decision-making within defined scope | Generation and customisation of arbitration clauses | Source-cited Q&A over AAA’s practitioner handbooks |
| Core technology | AAA-internal (institutional pilot) | AAA-internal | RAG-based, built with Edtek |
| Output type | Procedural outputs / decisions in scope | Drafted clauses | Cited answers grounded in handbook materials |
| Citation | Per institutional governance | Per institutional governance | Page-level citations into the source handbook |
| Refusal on out-of-corpus | Per institutional governance | N/A (generation, not retrieval) | Yes — refuses or caveats when the source does not support |
| Available to | Specific AAA-ICDR programme scope | Public (parties drafting agreements) | Open to all (free) |
| Stewardship | AAA-ICDR (with AI Governance Committee) | AAA-ICDR | AAA, built with Edtek |
The three tools serve different needs and should be evaluated against the need, not against each other. A counsel preparing for hearings has different needs from an institutional secretariat administering cases or a party drafting an arbitration agreement at the contract stage.
AI Arbitrator: scope, training data, when it can decide a case
The AAA-ICDR launched its AI Arbitrator on 3 November 2025 (announced 17 September 2025) for two-party, documents-only construction arbitrations — no live witnesses, no complex multi-party or cross-border issues. The tool was built with QuantumBlack, AI by McKinsey and trained on more than 1,500 AAA-ICDR construction arbitration awards, refined with expert-labelled examples and feedback from the AAA-ICDR Construction Panel. Architecturally it is a multi-agent system running on the AAA-ICDR Case Management Platform.
The workflow is more constrained than the name suggests. Parties upload submissions; the AI summarises the factual and legal arguments; parties validate the AI’s understanding of their case; the AI drafts a proposed award; an AAA-trained human arbitrator reviews, revises if needed, and issues the final award. Participation is opt-in and requires both parties’ consent. The training-data disclosure stops at “more than 1,500 construction awards” — corpus details and case-eligibility criteria beyond the documents-only construction scope are not public.
Andrew Barton, the AAA-ICDR Vice President overseeing the AI Arbitrator, told Akin Gump in November 2025 that the institution is “approaching AI Arbitrator as a platform, not a one-off tool,” positioning it “to expand responsibly as additional use cases and case complexities are brought into scope.”
The scope is bounded for both regulatory and practical reasons. The product is narrowly scoped, not a general AI-arbitrator product. Substantive decisions still flow through a human arbitrator’s review and issuance. The institutional governance posture (committee oversight, defined deployment scope, transparency about use to the parties) is meaningfully different from individual counsel adopting AI tools privately; institutional AI tooling carries institutional accountability.
ClauseBuilder AI: who it serves and its limitations
ClauseBuilder AI (Beta) launched on 12 June 2024 as a generative-AI evolution of AAA-ICDR’s long-standing ClauseBuilder tool. It uses natural-language processing built on Vellum AI workflows, drawing on a database of over 500 curated AAA clauses tied to variables (case size, industry, location, rules) to produce customised arbitration and mediation clauses — including AAA’s new AI-specific clauses. The tool remains officially in Beta.
The strength of the category: party-side tools that help non-specialists draft enforceable arbitration clauses reduce the rate of malformed clauses that cause downstream procedural disputes. An ill-drafted arbitration clause is one of the most common reasons arbitration goes badly at the threshold stage. Better drafting at the contract stage prevents predictable disputes about whether the clause covers the matter, which institution applies, which seat and language govern, and so on.
The limitations are documented in AAA’s own materials. ClauseBuilder AI explicitly does not include Employment or Consumer Clauses, and AAA recommends counsel review before adopting any output. The tool generates clauses but does not validate user-submitted clauses against AAA standards. Parties drafting their own arbitration clauses without legal advice should treat AI-assisted clause generation as a starting point, not a finished product — the same caveat that applies to any AI legal drafting tool.
Chat Book: source-verified retrieval over the practitioner handbook
The AAAi Chat Book is a different category of tool: a source-cited retrieval system over AAA’s case preparation and presentation materials. A user — arbitrator, advocate, student, self-represented party — asks a question in natural language; the system returns an answer grounded in specific passages from the AAA’s practitioner handbook, with citations to the page and section the answer derived from.
The technical underpinning is the hallucination-proof RAG architecture described elsewhere in this knowledge base. Specifically for Chat Book:
- Retrieval is calibrated against the AAA practitioner corpus. Generic similarity thresholds do not work; the parameters are tuned for the specific corpus shape.
- Page-level citations point the user to the source location in the handbook for verification.
- Refusal on out-of-corpus questions. If the handbook does not address the question, the system says so rather than answering from general knowledge or the model’s training data.
- Configurable abstention behaviour. For deployments where reviewer time matters more than refusal precision, thresholds can be relaxed; for deployments where every output must be defensible, tighter.
The Chat Book is freely available and open to anyone — a deliberate choice by AAA to make arbitration knowledge more accessible. The technology pattern is reusable: the same architecture works for other institutional bodies that want to make authoritative content more accessible without losing the citation chain that makes it authoritative in the first place.
We are not the right people to compare Chat Book against AI Arbitrator and ClauseBuilder AI on internal technical specifics, because we do not have access to the other two tools’ implementations. What we can say with confidence: the three tools solve different problems, and the right tool depends entirely on the user’s task.
Governance: institutional AI committees and external guidelines
A common thread across these tools is institutional governance. The AAA-ICDR’s AI Arbitrator operates under the oversight of an internal AI Governance Committee — referenced on adr.org’s AI Arbitrator product page and in Akin Gump’s launch alert — and within the broader AAAi Standards framework. AAA-ICDR describes the Standards as “the culmination of a structured governance initiative the AAA-ICDR launched in 2024” and organises institutional AI use around six pillars: ethical/human-centric values, privacy and security, accuracy and reliability, explainability and transparency, accountability, and adaptability. The Standards build on AAA-ICDR’s earlier “Principles Supporting the Use of AI in Alternative Dispute Resolution” (1 November 2023). The composition of the AI Governance Committee is not publicly disclosed.
External professional bodies provide complementary guidance. The Chartered Institute of Arbitrators (Ciarb) published its Guideline on the Use of AI in Arbitration in March 2025 and issued an updated version on 5 September 2025; the Guideline is soft law, non-binding unless parties or the tribunal adopt it. Institutional AI is not an individual technology decision.
Three governance considerations matter when evaluating any institutional arbitration AI tool:
Transparency to the parties. Parties should know when AI tooling has been used in the conduct of their case, what was used for what, and what review the human tribunal or counsel applied. Institutional governance frameworks generally require this; individual counsel using AI tools privately should adopt the same disclosure norms.
Scope discipline. A well-governed deployment has a defined scope — what the AI is used for, what it is not. Out-of-scope use risks both procedural challenges and broader institutional credibility. Tribunals and parties should expect institutional tools to honour their stated scope.
Audit and reconstruction. AI outputs that are part of the case record (procedural decisions, party-facing analyses, cited authorities) need to be reconstructible. Institutional AI tools that lack a defensible audit log should be treated cautiously regardless of their other features.
For the broader landscape of institutional rules and guidance — Ciarb, ICC, LCIA, SIAC, HKIAC, AAA-ICDR — see AI for International Arbitration.
Choosing between the three — decision tree
The right tool follows directly from the user’s role and task.
You are a party at the contract drafting stage. ClauseBuilder AI is the relevant tool. The output is a starting point; involve counsel for non-trivial commercial contracts.
You are a practitioner preparing for a hearing or working on a case. The Chat Book is the right tool when you need quick, cited access to AAA’s authoritative procedural guidance. For substantive research over your own case file, a private deployment of the same RAG pattern (Edtek Chat or similar) is the right tool — the Chat Book is over AAA’s published materials, not your case documents.
You are a tribunal secretariat or AAA-ICDR programme participant. AI Arbitrator tooling, where institutional rules permit, may be available for the scope it covers. Follow the institutional governance framework.
You are evaluating which tools to recommend or adopt. All three may be relevant for different parts of your practice. They are not substitutes for each other; they are different tools for different jobs.
Roadmap: 2026 expansion of institutional AI tooling
In its early-testing claims for the AI Arbitrator, the AAA-ICDR has projected 20–25% faster resolution times and at least 35% cost savings versus traditional arbitration (as reported in Adams & Reese and Faegre Drinker summaries); the AAA-ICDR has cited a wider 25–30% time savings and 30–50% cost savings range in other forums (A&O Shearman, Gowling WLG). These are AAA-ICDR’s own pilot/early-testing projections, not independently audited results — the methodology, sample size, comparator baseline, and measurement definitions are not publicly disclosed.
The institutional AI landscape in arbitration has expanded notably between 2024 and 2026. The general direction is clear even where specific institutional roadmaps are still being published:
- More institutions are deploying source-cited retrieval tools over their published materials, following the Chat Book pattern, because the upside (accessibility of authoritative content) is high and the downside is controllable with the right architecture.
- Party-side drafting tools are evolving from form-fill helpers toward generative drafting with appropriate disclaimers.
- Tribunal-side AI tooling is moving cautiously, with institutional governance leading rather than following the technology.
- Cross-institutional norms — disclosure, audit, scope discipline — are converging through bodies like Ciarb and through individual institutional rules.
For specific institutional roadmaps and policy positions in 2026, the institutional publications are the authoritative source. Where ranges and figures are quoted here, they reflect institutional pilot claims rather than peer-reviewed results.
Related reading
- Convert a Book to an AI Chatbot — the pattern the AAAi Chat Book uses, generalised for any publisher or institutional body.
- AI for International Arbitration — institutional rules and frameworks at ICC, LCIA, SIAC, HKIAC, AAA-ICDR.
- Hallucination-Proof RAG Architecture — the technical underpinning of the Chat Book.
- On-Premise RAG: Deployment Guide for Regulated Sectors — the deployment model for institutional AI work with confidentiality requirements.
Frequently asked questions
What is the AAA-ICDR AI Arbitrator?
An institutional AI tool deployed by the American Arbitration Association — International Centre for Dispute Resolution for use within defined institutional scope. The specific operational scope, training data, and governance framework are described in AAA-ICDR’s own published materials, which are the authoritative source for current programme parameters.
What is ClauseBuilder AI?
An AI-enabled evolution of AAA-ICDR’s ClauseBuilder tool, helping parties drafting commercial agreements construct enforceable arbitration clauses. Useful as a starting point for parties drafting their own arbitration clauses; not a substitute for legal advice on non-trivial commercial contracts.
What is the AAAi Chat Book?
A free, AI-powered tool built by the American Arbitration Association in partnership with Edtek. It transforms AAA’s case preparation and presentation materials into a conversational interface, giving practitioners, advocates, students, and self-represented parties source-verified answers grounded in the handbook with page-level citations.
How do these three tools differ?
They solve different problems. AI Arbitrator is institutional tribunal-side tooling within defined scope. ClauseBuilder AI is party-side drafting support at the contract stage. Chat Book is practitioner-facing retrieval and Q&A over AAA’s authoritative published materials. The three are complementary, not substitutes.
Can the Chat Book pattern be applied to other institutions?
Yes. The technology — source-cited RAG with page-level citations, refusal on out-of-corpus questions, configurable abstention thresholds — is general-purpose. Any institutional body with authoritative published content can deploy the same pattern over their own materials. The pattern is described in detail in our guide to converting a book into an AI chatbot.
Does the Chat Book replace AAA’s published materials?
No. The Chat Book is an interface for accessing the materials, not a substitute for them. Every answer cites back to the source location in the handbook for verification, and the materials themselves remain the authoritative content. The Chat Book makes the content more accessible; it does not replace the underlying publication.
What governance applies to institutional AI in arbitration?
Institutional AI deployment is governed by the institution’s own committee structures (e.g., AAA-ICDR’s AI Governance Committee), broader professional body guidance (e.g., Ciarb’s published guidelines for arbitrators), and the disclosure and human-oversight expectations of the specific institutional rules. Specific governance frameworks are evolving; institutional publications carry the current detail.
Are these tools available outside AAA-ICDR proceedings?
The Chat Book is publicly accessible to anyone — its corpus is AAA’s published practitioner materials. AI Arbitrator and ClauseBuilder AI are AAA-ICDR institutional tools whose availability follows the institution’s programme scope. The pattern of each tool (source-cited Q&A, party-side drafting assistance, institutional decision-support tooling) is reusable by other institutions with their own deployments.