Article | |

AI arbitrators

AI arbitrators
  1. What has happened?
  2. Is the decision-making process really automated?
  3. What's next?

Now Reading

What has happened?

The American Arbitration Association-International Centre for Dispute Resolution (AAA-ICDR) is set to launch an AI arbitrator for construction disputes in November 2025. This marks a pivotal transition toward automated dispute resolution in commercial law.

The AI arbitrator will rely exclusively on documentary evidence rather than live testimony or hearings. It has been trained on more than 1,500 previous construction awards and refined using expert-labelled examples. Construction is a sector with a high volume of technical disputes, which are ideally suited to data-driven resolution. We previewed the emergence of AI decision-makers in our 2024 article, 'Is AI justice on the way? AADR: Automated alternative dispute resolution'. Our view then was that highly technical areas (e.g. cases decided by expert determination) would be a likely area for the development of automation of disputes. This is consistent with AAA-ICDR's decision to launch its AI arbitrator for construction disputes.

Is the decision-making process really automated?

While the AAA-ICDR's introduction of an AI arbitrator is an impressive and promising step towards the use of automation in dispute resolution mechanisms – which should ultimately improve the efficiency and accessibility of dispute resolution – it is important to note that the AAA-ICDR is still envisaging that the AI arbitrator will ultimately be subject to human oversight.

AAA-ICDR has explained that the AI arbitrator will be used to generate the initial decision, which will then be subject to human review prior to finalisation. As a consequence, awards will not be reliant solely on the decision-making of the AI arbitrator, and so the process is not fully "automated". This hybrid model is a way of simultaneously ensuring that there is due process and trust from parties, while utilising the benefits of AI to improve speed and consistency. Concerns that parties may have with this approach include automation bias (i.e., the tendency for humans to be biased towards AI output) and hallucinations or error which may be missed by human review – the various drawbacks were explored in our previous article.

What's next?

AAA-ICDR is also planning to expand support for additional industries and more complex or higher-value disputes in 2026. We anticipate that other sectors that generate significant volumes of data-heavy and standardised disputes (such as insurance, consumer contracts, or some areas of intellectual property) would be prime candidates for such expansion.

We expect that AAA-ICDR will apply any learnings it derives from the roll-out of the construction disputes AI arbitrator to its expanded offering. Nevertheless, it remains to be seen how reliable the AI's outputs will be in practice, and whether there will be significant adoption of the tool by commercial parties. It would be interesting for parties to understand, in respect of each case, whether or not the AI arbitrator's decisions have in fact been confirmed or overturned following human review (and if so on what basis). However, AAA-ICDR has not yet provided any indication that it will provide such information to parties, or indeed what (other) information about the decision-making process will be made available to parties using the tool.

Get in touch

Read David Bufton Profile
David Bufton
Back To Top Back To Top chevron up