AI in Arbitration and Mediation: Real Uses, Limits, and What Your Firm Must Change

AI in Arbitration and Mediation: Real Uses, Limits, and What Your Firm Must Change

The arbitration and mediation are among the legal practices where AI is being tested most rapidly. Not because of fashion, but because of cost pressure, the growing duration of procedures, and the volume of documents push institutions and firms toward solutions that once seemed dispensable.

The ICCA Congress held in Madrid from April 12 to 15, 2026, dedicated a good part of its agenda to this change. And the most relevant arbitration institutions in the world have already taken concrete steps that are worth understanding before deciding if, how, and when to integrate AI into your firm's practice.

Why Now

The LCIA published in its 2024 cost analysis that the average duration of its procedures increased from 16 to 20 months. The amounts in dispute grow, teams are smaller than a decade ago, and clients, especially corporate ones, pressure to shorten deadlines and improve predictability.

In September 2025, the ICDR (international division of the American Arbitration Association) announced its "AI Arbitrator" for construction disputes based solely on documents. In internal tests, the tool delivers resolutions 20% to 25% faster, with cost savings of 35% or more, and a potential reduction for parties of 30% to 50%. The model was trained with more than 1,500 real annotated awards.

There is no autonomy. The human arbitrator reviews, adjusts, and signs. But it marks a before and after in what an arbitration institution can offer.

Three Concrete Applications Already Operating

1. Analysis of Arbitration Precedents and Jurisprudence

Arbitration jurisprudence has a characteristic that complicates it: many awards are confidential. This makes searching for precedents costly and dependent on specialized databases that many firms do not have licensed.

The Jus Mundi platform has built the largest public repository of international awards with annotations. The LCIA formalized in 2025 a strategic collaboration with this platform. The HKIAC did the same, signing an agreement to use Jus AI in preparing case summaries.

For a firm, this translates into tasks that previously occupied a week of an associate (mapping relevant jurisprudence on a clause in a specific sector) now being resolved in hours, with greater coverage and lower risk of relevant omissions.

2. Outcome Prediction Based on Patterns

This is where caution is needed with language. AI does not predict what a specific arbitration tribunal will decide in a specific case. What it does is analyze historical databases and show statistical probabilities about how disputes with similar characteristics have been resolved.

Tools like LexMachina or Solomonic offer analytics on tribunal behaviors, arbitrator history, procedural strategies that have succeeded or not, and associated outcomes. The correct use is not a closed probability figure. It is information about compensation medians, reasoning patterns, trends in accepting certain arguments.

The 2025 International Arbitration Survey by Queen Mary and White & Case shows that 52% of respondents anticipate that arbitrators will rely more on AI in the next five years, and 91% expect to use AI for factual and legal research.

3. Document Management in Complex Procedures

An international arbitration in sectors such as construction, energy, or insurance can involve tens of thousands of documents. Technology-assisted review is not new, but the entry of generative AI has expanded what can be done:

• Automatic classification by relevance and document type.
• Extraction of key clauses and obligations in contracts.
• Timelines of events built from correspondence.
• Detection of inconsistencies between documents and testimonies.
• Preparation of evidentiary indexes.

With the appropriate configuration, a team processes in two weeks what previously required two months. And consistency improves, which is usually the real problem when volume exceeds a certain scale.

The Limits That Matter

Legal criteria, strategic reading of the case, negotiation, assessment of witness credibility, the decision about whether the client should settle or proceed: all that remains human. It is not an abstract principle, it is a real technical limitation of current models.

Sector guidelines have formalized this point. The Ciarb published in 2025 its Guideline on the Use of AI in Arbitration. The Silicon Valley Arbitration and Mediation Centre (SVAMC) published its AI Guidelines in 2024. Both insist that AI can assist, but that the arbitrator cannot delegate their decision-making function. The ICC Note to parties and tribunals from 2021 already established the same logic regarding arbitral secretaries.

What Must Change in a Firm Working in Arbitration or Mediation

Madrid has consolidated itself as an international arbitration seat. In the Queen Mary Survey 2025 it appears for the first time in the global top, alongside London, Paris, Singapore, or Geneva. The CIIAM (formerly CIAM) has managed 68 cases since 2020 for an aggregate amount of 424.9 million euros. And the Law on Efficiency of the Public Justice Service, in force since April 2025, promotes appropriate means of dispute resolution in the Spanish system.

In that context, there are decisions that a serious firm must make in the coming months.

Review the Actual Document Flow

Quantify how many hours are invested in each procedure in classification, search, and timelines, and separate what can be automated from what requires criteria. Without this diagnosis, any decision about technology is speculative.

Decide on Infrastructure and Confidentiality

Arbitration lives on confidentiality. This rules out many generalist tools based on public models. It forces considering solutions where data never leaves a controlled environment, or implementing prior anonymization layers. It is a technical debate that is worth addressing with someone who understands both AI and the regulatory framework (GDPR, EU AI Act, professional secrecy).

Formalize AI Use Clauses

Terms of reference and procedural orders are starting to include clauses about AI use by parties and the tribunal. Your firm needs its own criteria to negotiate these clauses and to know what to reveal about its own practices.

Train the Team

AI in arbitration is not delegable to "the technology team." Associates and partners who handle cases need to understand what they can and cannot do with these tools. Without that training, integration fails due to distrust or improper use.

The Next Step

Arbitration is a field where a generic tool rarely fits well. The parties have agreed to strict confidentiality, documents are strategic, and deadlines are tight. A solution built or adapted to the actual flow of the procedure and the internal policies of the firm or the institution that administers it usually works better.

If you want to explore what makes sense for your specific firm (what processes to prioritize, what infrastructure you need, how to integrate AI respecting confidentiality and regulatory framework), at Artekia we can help you see it in a no-commitment conversation. Let's Talk.

---

Sources

• Queen Mary University of London & White & Case, 2025 International Arbitration Survey.
• Faegre Drinker Biddle & Reath, "Use of AI in Arbitral Institutions" (Feb 2026).
• Hill Dickinson, "AI arbitrators: a new era in international arbitration" (Sep 2025).
• Kluwer Arbitration Blog, ICC YAAF workshop on AI in arbitration.
• Ciarb, Guideline on the Use of AI in Arbitration (2025).
• Silicon Valley Arbitration and Mediation Centre, AI Guidelines.
• LCIA, Costs and Duration Analysis (2024).
• CIIAM / CIAM-CIAR, statistics and activity 2020-2025.
• Law 1/2025, of January 2, on measures in the field of efficiency of the Public Justice Service.
• Global Arbitration Review, "Spain: Madrid rises as an emerging arbitration hub" (2026).
• Crowell & Moring, "Consideration of Artificial Intelligence in Arbitration Terms of Reference" (Jan 2026).