In my previous article on AI in ADR, I emphasized both the opportunities and risks associated with the use of artificial intelligence in alternative dispute resolution (ADR) and suggested that it is possible to move too fast. Lest I sound like a skeptic or, worse, a Luddite, my 35 years in legal technology has taught me that it is more important to focus on the goal—or benefits, if you will—offered by new technologies and then figure out the best way to get there for a reasonable cost.
Exploring AI's Potential in ADR
With a basic understanding of AI and large language models (LLMs), it isn't that difficult to come up with a long list of potential applications, or use cases, in the ADR industry. In fact, the AI Steering Committee at JAMS devised dozens of interesting use cases before we agreed on a much smaller list of actual projects that align with our vision of how our clients, neutrals and associates can benefit from AI.
A window into the world of mediation in the not-too-distant future can help illustrate some potential opportunities. Because mediation involves such a heavy dose of human interaction and psychology, AI is less likely to replace skilled mediators any time soon in most types of disputes. But here are some situations in which it can help.
Enhancing and Streamlining Processes
Company A enters into an agreement with Company B to have Company B distribute its product. Company A's in-house legal team constructs the distribution agreement using a generative AI engine that has access to all prior written agreements the company has entered into, as well as a database of all documentation associated with prior disputes arising out of those agreements. Having been trained in the use of AI technology, the lawyers utilize their experience in prompt engineering and employ retrieval-augmented generation (RAG) to allow the search results to cross-reference examples from outside sources. The resulting contract contains a dispute resolution clause created with an online tool supplied by an ADR provider, calling for mediation of disputes or, failing that, an arbitration before a panel of three arbitrators.
Months later, Company B makes a written claim against Company A, asserting that its product does not meet the standards defined in their agreement. Lawyers for each side go to the website of their preferred ADR provider, where they find a chatbot waiting to help with their selection of a mediator. They use natural language questions to query the institution's roster of neutrals to find someone who has the requisite experience and background to help resolve their dispute. Fortunately, both parties are able to agree on a mediator who can meet with them quickly so they can get past this dispute and back to business.
In preparation for their mediation, the lawyers again query internal and external LLMs to hone their arguments and put together their mediation statements. They carefully review their briefs to ensure that there are no erroneous or irrelevant facts produced by AI hallucinations. They submit their materials to the mediator through the ADR institution's client portal.
Because of the distance between the two companies' offices, they decide to use an online video platform instead of meeting in person. The ADR institution has already sent out automated engagement agreements, mediator disclosures and fee schedules, and directs the parties to its client portal to select a mutually agreeable date for the mediation. Once that date is secured, the portal automatically sets up the video call and communicates the details to the parties.
The mediator, having received the materials electronically, reads through the documents while asking their AI mediation assistant to summarize the history of the dispute and the arguments made by each party. The assistant also gives a preliminary assessment of the strengths and weaknesses of each party's arguments, as well as a range of potential outcomes based on similar cases in the past.
Swift Resolution
The day of the mediation arrives, and the parties join their mediator online. After an initial discussion, the parties enter their virtual breakout rooms. The mediator moves back and forth between these rooms, making notes on their tablet about offers and counteroffers. As the day progresses, the AI mediation assistant refines its analysis and begins to suggest potential next offers and likely outcomes. Toward the end of the day, the parties ask for a mediator's proposal to settle the dispute. The mediator uses their past experience and instinct as well as the AI assistant's input as a point of reference. The AI assistant predicts an 80% probability that the parties will accept the mediator's proposal.
Eventually, the parties accept the proposal, and the AI assistant generates a draft settlement agreement, which the parties edit and finalize. The agreement, mediation data and mediator's notes are automatically anonymized to eliminate personally identifiable information before being added to the ADR institution's internal and private LLM.
Fortunately for both companies, the entire process—from Company B's claim to the resolution of the dispute—took three months instead of an estimated nine months without the use of AI and other automation, eliminating litigation. Both parties are able to get back to business after Company A made adjustments to its production processes. The mediator is now available to take on more cases because of the speed of the mediation, and the lawyers for both parties can move on to their other responsibilities without the distraction of a long-running dispute.
After all, that is what commercial mediation is all about: reaching a swift resolution to disputes so that relationships can be preserved and everyone can get back to business. While other technologies such as online dispute resolution, clause generators and client portals have advanced the practice of ADR, AI has the potential to have a far greater impact on ADR.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.