For nearly a century, the American Arbitration Association® (AAA®) has been a leader in alternative dispute resolution (ADR), committed from its founding to designing systems that make fairness in ADR possible. In November 2025, the AAA launched its AI Arbitrator—an AI-powered system that evaluates the merits of claims, generates explainable recommendations, and prepares draft awards. This innovation reflects how the AAA continues to improve ADR while staying rooted in its core principles. In this episode, Frank Rossi, AAA COO and CRO, joins host Bridget McCormack and Zach Abramowitz to share how the AAA built its first-of-its-kind, human-in-the-loop AI Arbitrator.
Drawing on nearly three decades at the AAA, Frank recalls the moment it became clear that the AAA was uniquely positioned to take on this project. “We have always been the pioneer,” he notes. Bridget walks through the choices that shaped the service, including how the system helps parties feel heard and understood from the very beginning of a case—an insight that became the center of the project.
The conversation offers a close look at trust, fairness, and what it means for an institution to pursue innovation with both confidence and responsibility.
Key Takeaways
- How the AI Arbitrator Works & Why Feeling Understood Matters: The AI Arbitrator reads parties’ submissions, extracts the claims, elements, evidence, and legal arguments, and reflects its understanding back by asking a simple question: “Did I get it right?” The parties further provide feedback on the AI Arbitrator’s summaries of their submissions, and the human arbitrator assigned to oversee and validate the AI Arbitrator’s work and issue the final award can see that feedback. Bridget notes that research shows parties accept outcomes more readily when they feel heard and understood, and the team found that the system’s ability to restate claims accurately, using the parties’ own structure, can improve procedural fairness in ways traditional systems cannot match at high volume.
- Why the AAA is Uniquely Positioned to Build the AI Arbitrator: Frank explains that the AAA has defined arbitration standards for generations, and its caseload experience, neutrality, and not-for-profit mission give it responsibilities no private company can match. With its longstanding history of trust, the AAA was uniquely positioned to introduce an AI Arbitrator responsibly.
- Why Construction Was the Right Starting Point: Construction offered a uniquely strong foundation for the AI Arbitrator. The AAA administers thousands of construction cases each year, many with substantial and standardized document sets that created an ideal dataset for building the system. Arbitrators in the field were enthusiastic partners, and parties already operate in a highly tech-enabled environment. As Frank, who grew up in a construction family, explains, modern job sites have transformed so dramatically through technology that he sometimes hardly recognizes them — another sign that construction was the right place to start.
- Opt-In Participation, Human Oversight & Transparent Results: The AI Arbitrator is completely opt-in and used only when both parties agree; otherwise, the matter moves forward traditionally. Frank emphasizes that the process is party-driven and allows for party feedback on the AI Arbitrator’s summaries. Also, because the system was trained with direct input from experienced arbitrators, its reasoning closely reflects how seasoned neutrals review cases.
- Frank’s Path from Skepticism to Conviction: When the AAA began exploring the AI Arbitrator, Frank had reservations, but his perspective shifted as he saw strong interest from long-standing customers, steady software progress, and growing excitement within the AAA. Speaking with customers and hearing them identify potential applications across arbitration, case evaluation, predictive analytics, and mediation further convinced him the AAA is moving forward with purpose.
Final Thought
The AI Arbitrator is the kind of project that works when the institution behind it knows what it is trying to protect. It accelerates early analysis but keeps the neutral in charge, and it makes room for parties to feel heard before the outcome is written. It is built from data, shaped by judgment, and backed by nearly a century of trust in the AAA. It is not a leap. It is a deliberate step into what arbitration can be when innovation and fairness move together.
Transcript
Zach Abramowitz: Hey there, and welcome to the AAA AI Podcast, Bridget, we have a distinguished guest with us today. Frank, do you want to introduce yourself to the audience?
Frank Rossi: Sure. I'm waiting for the distinguished guest to show up though, but I guess I'll substitute for now. I'm Frank Rossi, Executive Vice President, Chief Operating and Revenue Officer here at the AAA ICDR. I’ve been here just about 30 years, and I've had this role for a couple of years before that I was CFO for about two and a half decades. So, it's been a fun ride, and it looks like it'll be a continued fun ride for the next several years.
Zach Abramowitz: Yes. And I like to really credit you, I don't know if you want credit for this, but as the person who handed me a book given to you, by one of the former CEOs of the AAA, the work of Frances Keller American Arbitration. You said, if you want to know a little bit more about the history of the organization, check this out. And then I spent a summer just absolutely immersed in the book and the works of Francis Kellor. So, we certainly owe that to you.
Frank Rossi: We're going to have to get you a Frances Kellor custom t-shirt Zach.
Zach Abramowitz: It's already in the works. So, we're here to discuss what I think is really a significant moment, the launch of the AI Arbitrator. It's been covered already a little bit in the news, but for those who are not familiar, let me start here with you, Bridget. What is the AI arbitrator?
Bridget McCormack: The AI Arbitrator is a new service the AAA launched last week. That is a human in the loop AI native arbitration service for documents only construction cases. AI Arbitrator is shorthand for a series of agents that operate across the arbitration process and make it significantly faster and significantly less expensive. It is our first offering of its kind and we're very excited about where we can go from here.
Zach Abramowitz: It's not only your first offering of its kind. Isn't it the first offering of its kind, outright?
Bridget McCormack: Yes, absolutely. We're definitely pioneering in this area. We started with a caseload that we know can really benefit from some of the efficiencies that the technology will bring to the process. But as I said at the beginning, our service has a human in the loop, throughout the process because that's important to us, as we hear back from users about what else they want to see.
Zach Abramowitz: Effectively, this means that two parties in a dispute and you mentioned construction cases, and I want to get to why we started there but effectively, two parties to a dispute can now, by agreement of both parties, choose to have the option instead of having a dispute resolved primarily by a human, they are having a dispute that is essentially resolved by AI.
Bridget McCormack: Yes, but as I said, there is a human in the loop throughout the process for the front end of the process. It's the parties themselves, the AI agents, that are extracting information about claims and elements of claims and evidence, supporting those elements and legal frameworks supporting their arguments. The agents then have the parties affirm that the agents correctly understood it. From there, when the agents start reasoning and drafting awards, we have a panel of our distinguished construction arbitrators who then step into the process and supervise the work of the agents.
Zach Abramowitz: Frank, Bridget mentioned before that you're starting with construction cases only, and document only cases as well. Can you explain what the thought process was? Obviously, you're not going to try to boil the ocean and roll out AI to resolve any and all disputes that come into the AAA. Why did we think construction was the best place to start? And construction documents only cases?
Frank Rossi: Yes, we thought construction document only cases were the right place to start because for a few reasons. One, we had a relevant and meaningful data set. We do a large number of construction cases annually in the thousands, and even documents only in the hundreds. So, we had data, we had documents, briefs, arguments that we could utilize, to develop the software. We also had willing participants in construction arbitrators. And we also had some parties we think were interested in this. The construction industry is very tech future oriented. I grew up in a construction family and if I walk onto a construction site today, I probably wouldn't even recognize it with all of the technology that's in play to properly manage an effective project. So, we thought this was an industry, and a community that would be very much open to leveraging technology, for disputes as well.
Frank Rossi: I also want to circle back to Bridget's point on humans in the loop. I would also add, we know that there are some people that also want some support. And so, we also have staff behind the scenes, case support specialists who will be there. I don't know if it will be a help button or how it's working, but we also have staff that will help support people if they run into any kind of issues with the AI arbitrator.
Zach Abramowitz: Yes, and I think it's a really interesting element of this because, you know, typically when we think about things like tech and automation. The thought is, oh, we're going to lose the human component. I like to remind people that in the early days of Netflix, when they were actually still sending out DVDs, they used to have people who you could call up movie buffs and ask them to recommend movies to you, and, they would talk through all kinds of films and, help you decide on the best movies to watch. And I think that's sort of emblematic here where it's like on the one hand, you're automating a certain portion of the process with technology, but on the other hand, you're also making the experience feel more VIP red carpet, with a kind of a hand holder there as well.
Frank Rossi: Yes, I think that's right. We do want to reduce the human touch points, as much as possible, because that's what makes it efficient. That's what makes things move forward more quickly. But we also want to have that safety net there for people when they need it.
Zach Abramowitz: Bridget, I've been at some of your talks where you've discussed why AI is such a great match for ADR and there's a quote that you like to mention from Ilya Sutskever. Can you talk about why the technology is such a great match for dispute resolution?
Bridget McCormack: Yes, I mean, there are some obvious reasons. You've heard us say that we're going to be able to offer this dispute resolution service at a lower price point and it'll produce faster results. And obviously that means more people can have access to this option. And a lot of parties right now can't afford to get their disputes resolved and they often give up on them. And so more dispute resolution, more alternative dispute resolution is good for economies. It's good for communities.
Bridget McCormack: We predicted that at the front end. But there's another aspect of it that I don't think I fully understood at the front end, the Ilya Sutskever quote that I often use in presentations is, the thing that surprised me most about the models is the way they make, feel, understand, and that absolutely, you know, resonates with anyone who uses it. It turns out to be such a value add in dispute resolution because there's this robust literature that won't surprise anybody that goes something like this. In a dispute resolution process, when parties feel like they were able to say their piece, they feel like they were heard and understood. They will accept outcomes that are negative to their position and still grow trust in the system. Because of that sort of procedural fairness or procedural justice, it's sometimes called.
Bridget McCormack: In this process, our agents across the front end of the arbitration process, like I said, communicate with the parties and say. Here's what I understand your claims are. Here's what I believe the elements of each claim is. As you explained it to me, here's the evidence I think that supports it based on what you gave me. Here's what I think your legal arguments are. Did I get it right? And the parties get to say, yes, you did. And that's just a game changer. That's not something courts can ever provide given the volume that they're deciding. Judges are never going to be able to provide that kind of feedback.
Zach Abramowitz: Right. And so in a sense, we think of AI and immediately we think, oh, less human, robots. But in fact, what you're suggesting is that the AI is actually going to create a more human process, meaning it focuses on the parties themselves and makes them feel heard, makes them feel understood.
Bridget McCormack: I really think that's going to turn out to be one of the things that parties value about this, that they can be very confident that their dispute and their complaint is understood. before they find the outcome.
Zach Abramowitz: You and I interviewed, recently the founder of Dioptra, one of the cofounders, Farah Gasmi. She had a very interesting thing that she said to us, which was that you can get into an organization, by discussing AI in terms of savings and efficiency. But she said that won't actually make the product stick. What makes the product stick in her experience is whether it gives the attorneys a better work experience. And so I think what's critical here is that, as you're suggesting, it's going to be a much shorter turnaround time. Yes, it's going to be much less expensive, you know, economy, justice, speed are sort of the three pillars of the AAA. It should essentially be less biased as well, but ultimately, you know, it's not about just the efficiencies. It's about a better dispute resolution experience, one in which the parties feel heard and feel understood and, and sort of regain trust in the process.
Frank Rossi: It's an interesting notion, right? I hadn't really considered it until Bridget was just talking here. You know, inherently, humans get distracted, right? I know I do it, I tune people out, if they're starting to drone on, and I'd never really considered that the AI will listen to your entire argument. Might not get distracted by a buzzing phone or thinking about what they're having for dinner tonight or what their commute's going to look like. So, yes, I'd never really considered that the AI might be a better listener than humans, so it kind of throws me for a loop a little bit.
Bridget McCormack: They never get hangry. They never get hangry. And I can actually tell you that some judges do get hangry.
Zach Abramowitz: Yes, and you're referencing the Kahneman famous study of Israeli judges that suggest that they're more likely to give parole the closer they are to when they've had their last meal. And I think that sometimes we obsess about, well, are you losing the human components? Assuming that the human component is only a good thing. When in fact sometimes the human component could actually make someone more biased. I like to think within Lady Justice and I know that dispute resolution is different from courts of justice and I think that's an important distinction. If we think about Lady Justice, you know, by making her blind and effectively what we're doing is making her slightly less human. We're saying actually. Yes, some things about your humanity are not great in the court. We want you to sort of overcome that. And I think that to Frank's point, there are many places where the AI might actually be better at certain components of this than the humans, despite the fact that you still have humans in the loop here. Frank, you've been at the organization, you said 30 years in one month?
Frank Rossi: In one month, 30 years. 29 years, 11 months.
Zach Abramowitz: So. You, you've been here for a while.
Frank Rossi: You call me old without calling me old. Zach
Zach Abramowitz: You've got a long career in dispute resolution. Why do you think it was important for the AAA to be the pioneer here in this? Why did it not make sense to just wait for some startup to do this. Why was it critical that this come from the AAA?
Frank Rossi: I think AAA's always been the pioneer, especially in this field. Innovation is a word that's thrown around, very, very quickly these days. But we've been innovative for as long as I've been here, developing websites, online filing, introducing bot automation in some of our processes, a long time ago. So I think it was just natural for us to move right into the AI space, with AI Arbitrator, certainly Bridget pushed us a little bit there. She came in and was all in from day one. Maybe not day one, maybe day 30, but she was all in. I think what makes it important that we did this is that we're a not-for-profit. We've been around for almost a hundred years. We have done a lot for the ADR field. Many of the processes that are in place today, we created, and they were quickly copied by others, and so I think it was important for us to take this on because of that a hundred years of trust, because we know what we're doing. We're not looking at this as just a cool technology, we're just leveraging technology to improve our customer services.
Frank Rossi: So, you know, I don't want to necessarily bad mouth software startups or tech companies or anything like that, but I feel like we're doing this for the right reason. And it's not really profit driven. It's just the latest and greatest in AAA's long history. Improving our processes and improving the customer experience and improving the field for everyone.
Zach Abramowitz: Yes and, advancing the cause of the mission, isn't that the original mission statement? And as I mentioned before, you gave me his book, American Arbitration by Frances Kellor, that, I think everyone in the organization has since fallen in love with, one of the things she talks about is what she refers to as the science of arbitration. And she really discusses the American Arbitration Association as a kind of lab that should constantly be working to figure out how we can bring dispute resolution to more people? Because the more people that have access to it, that will create essentially a greater society. But let me stick with you for a moment because again, you have been here for what will be in one month, 30 years.
Frank Rossi: Yeah, keep calling me old Zach. Go ahead.
Zach Abramowitz: You know, you mentioned Bridget was all in on day one. You probably had some reservations, and I think we've seen this with others. Can you describe to me some of what your initial reservations were and at what point you were converted and said, oh, I'm all in on this. Can you take us through that process? Because I think for a lot of people who are hearing for the very first time, AI is going to resolve my dispute, may share some of those reservations and I think it'd be useful for them to hear from you, who may have had some skepticism about this. You know, what ultimately persuaded you that, oh my gosh, this is the direction that we need to go and we need to kind of go all in.
Frank Rossi: Sure. So, some of my initial skepticism or reservations was a little bit out there. I grew up with sci-fi, you know, Terminator, and I just thought. Are we contributing to the ultimate demise of humanity here? But that gets quickly put aside. I think I started worrying a little bit about market reaction. We have built a lot of these processes. We have over 5,000 arbitrators. We have tens of thousands of customers. How would they react to this? Would they view this as just us jumping on some AI bandwagon or some tech bandwagon?
Frank Rossi: And if I'm being honest, in the early days I thought this was going to cost a lot and we're just not going to be able to pull it off. I haven't been CFO for two and a half years now, but sometimes it's hard to turn that switch off in my head. And I just started worrying about all the money that would flow out the door and, and whether that money should be invested elsewhere. What ultimately helped me turn the corner, there were a few things. One was seeing the excitement throughout the organization about doing this, and I thought, okay, a lot of the people that I thought would be skeptics are not skeptical. So what, what am I missing that they're seeing?
Frank Rossi: I started having conversations with them about why they were excited, and that got me excited. Part of it was talking to longstanding customers and hearing their excitement about possible use cases for the technology, things that we might not even have thought of. Then when I started seeing the progress of the actual software development and seeing that we were actually making headway, against technological challenges and that we might actually pull this off, then my guard was let down even further. And the more I talked to customers, that's part of my job now to be out there talking to customers. The more I realize there is excitement about this, I don’t know that anybody wants to be the first user of this, have the first award. Personally, I would, it's kind of like opening a bar and having that first dollar bill hanging up over the cash register. I would want to have the first arbitration award, under the AI Arbitrator, and I'd frame it and I'd talk endlessly about how I was first. So I'm not sure when the tipping point will come, when people will really jump onto this.
Frank Rossi: But I am energized by the excitement that I hear from customers about all the different ways they think they can use this, whether it's for binding arbitration, a case evaluation, predictive analytics, mediation, there's all kinds of use cases that that can keep us busy for a long time and, and further growing the AI the trade.
Zach Abramowitz: Bridget, Frank mentioned that you were all in on AI. I think it's important that people understand the circumstances that led to this. So, you were hired by the American Arbitration Association in September of 2022. But you only started in February of 2023. You said you were going to take some time off and read some books in between then. But instead what happened?
Bridget McCormack: Instead, Sam Altman ruined my little sabbatical and instead learned everything I could about large language models, how we got here, the 70 year history and what they were likely capable of. Because it seemed pretty clear to me that we had a magic language calculator for a profession whose currency is language. And so I was pretty convinced this was going to have an impact on the business law and the practice of law, and I wanted to make sure I could start to formulate my thinking about what it would mean for our business and our users. So that's what I, that's what I did
Zach Abramowitz: But, but now the AI arbitrator, even though you were all in from, you know, day one, as Frank mentioned before, it was not the first AI project you worked on, and I think it's important that people understand you know, some of the context and history here. Do you talk about, before going for the sort of holy grail, some of the steps the organization took to sort of begin working on those AI muscles?
Bridget McCormack: Yes, it took us a couple of years to feel like we were ready to build what I think will be a transformative service across the industry. We spent the first two years figuring out what we could build ourselves, what our users would want, what ways the technology can make a difference across our operations. We had everybody in the game. We have enterprise ChatGPT licenses for people across our organization and we have people in every role, using it and figuring out new ways it'll have an impact on what we do and, and the way we serve our users.
Bridget McCormack: We gave our panelists some of the tools and we had them start experimenting with it to figure out how it would impact their end of the process. From all of those ideas, we started building point solutions. You know, we believe that action leads to information. So we were not waiting, a few years for the white paper to tell us we could start. We built point solutions, some of which are still in use and quite popular. I think the first one, cost builder AI that we built, is not only still in use. People really like that product. We built others that we moved on from, they just didn't turn out to be useful. But that was the only way to learn what would make sense in terms of some of the more transformative projects.
Zach Abramowitz: Right. I think it’s; I think it's critical. I think it's critical with what Frank mentioned beforehand, because, you know, there was this concern, or we are just jumping on the newest fad or, you know, is, is this like another kind of Web3 moment? I think it's important to remember. AI in the form that we're talking about has now been available for nearly three years. And so it was not in year one that you set out to build out the AI Arbitrator. It wasn't even in year two, it was after a significant amount of research and monitoring the market and being able to say, you know what, no, we think this is quite real.
Zach Abramowitz: And at this point there's a much greater risk in waiting. Especially for an organization that, as Frank said, has been a pioneer from its inception. And so, you know, it's interesting coming up on the 100-year anniversary as an organization that you really went from zero to one was in 1926 and you're a hundred years later. Again, a kind of category defining moment. One thing I wanted to point out, because Frank mentioned it before, is this fabulous statistic that came out from Microsoft and LinkedIn in 2024, which is that 33% of companies surveyed had bought an AI tool, but 76% of the employees at those organizations were using AI at work. Recently, the MIT study had this in similar numbers, 40% to 90%. And if you look in most of the industry. ChatGPT and AI have been adopted more rapidly by employees than by employers. But in the AAA that is not the case. As Frank said, the organization has been great at inspiring many people to embrace AI and in that respect, I work with a lot of law firms, a lot of legal service providers and legal departments.
Zach Abramowitz: I've never seen that. That's the one thing that I think is unique about the organization is that the organization led and inspired its employees to do the same. And there are employees on both the senior level I've met and the front lines of the AAA who have really embraced AI.
Bridget McCormack: Yes. I think that's all right. I think our view was we don't know where the best ideas will come from. This is just a general purpose technology. It's not the kind of technology you can stick with the IS team, with the IT team and hope that they figure out how we should all use it and then they give us some training and then we all use it. it's electricity, it's the steam engine. So it made sense for case managers to figure out how it impacts their work, for marketers to figure out how it impacts their work, for the legal team to figure out how it impacts their work. it's an all in approach.
Zach Abramowitz: And I think that's probably surprising to some people who may not think about the American Arbitration Association as being an AI or tech forward company just because it has been around for a hundred years. But I think the AI culture here is palpable.
Zach Abramowitz: Frank. Obviously one of the questions that people have, we talked about the reservations before. Well, can these decisions be appealed if I lose to AI, have you heard that concern? And, and I, I'd like to hear from Bridget on this also. How should parties feel confident that, okay, no, these decisions are binding and final, like any other decision that's issued by an arbitrator at the American Arbitration Association.
Frank Rossi: Yes, so first off, they are decisions issued by an arbitrator at the AAA ICDR, that's the human in the loop, the final check and balance. And so in the end, they're really not going to be any different from an award that would be issued under a traditional process. It will just be faster and cheaper to get there. Long run though, you know, there may be some cases where parties agree to accept the AI arbitrator decision without the human in the loop. And I think the way people get comfortable that that's enforceable, they're opting into this. So right now it's an opt-in provision. One party files a case, the other party has to agree to utilize the AI arbitrator. If the other party, if the respondent does not agree, then we just let them know that we're happy to handle their case under our traditional process. A little more money, more time, et cetera. But, that's always an option.
Frank Rossi: So it really is a party driven process and so, you know, I think because the AI Arbitrator was built, utilizing input from actual arbitrators that documented how they review cases, the reasoning they apply, that should give people confidence. That it is a thorough and well thought out process, but what you're asking is will these cases be appealed? Sure. I mean, anything can be appealed and I'm sure there will be some subset of AI Arbitrator cases that are appealed. Just like there are a subset of traditional cases that are appealed or are brought to a court for confirmation or affirmation. So I don't think the AI arbitrator awards are unique in any respect when it comes to losing parties or parties who are dissatisfied. The end result will simply find ways to try to overturn an award that's no different than litigation or arbitration today, in my opinion.
Zach Abramowitz: There's a pretty complete answer to Bridget. Anything to add on that? I know, if lawyers can find some wiggle room, they're going to be those who try. What have you heard on that point? How should people feel again, confident this is like any other decision issued by the AAA.
Bridget McCormack: The best way people get comfortable is once they see the work that it does. We are releasing a white paper that's been put together by an academic who we let under the tent and let him have at it to test our systems John Choi at WashU and that'll help, but I think that the proof is going to be when people use it and they are satisfied because they do feel like their claims were understood, evidence was understood. They see the reasoned decision. and there's not much to complain about. Of course, you know, we have many, many cases right now where people try to raise an issue when they're unhappy with the results. So I assume that'd be part of our future as well. I assume people will continue to be people even when AI processes bring more reliability to dispute resolution, but I think the proof is in the product's results.
Zach Abramowitz: Actually, I have a prediction I want to make on this, and I'm outside the organization. I think you're going to see the data point. I think people would not have guessed. And again, I know this hasn't played out yet, it's a prediction. I know we're in the end, it's the end of the year here. So, good time for prognosticating and, and prediction. I think you're going to see less cases appealed as a percentage. I think AI will be appealed less than human cases, and I know that people think that that's sort of counterintuitive. My sort of contrarian take is I expect that there are going to be less cases from AI led decisions that are appealed than otherwise. We'll have to see how that plays out. But you've got me now on record, Frank, where can people who are interested in opting in, I know with construction cases only at the moment, we're going to probably expand to other caseloads soon. But where can someone who's got a construction dispute learn more about this process and how can they take advantage? Because we're not talking about the future anymore, we're very much in the present. How can they learn more about the AI arbitrator and actually enter into a dispute that is AI first.
Frank Rossi: they can go to our website, adr.org there is a section on the AI arbitrator, https://www.adr.org/ai-arbitrator/ and that section includes contact information, for one of my Business Development Vice Presidents who is focused exclusively on the AI Arbitrator, education, outreach, and business development. His name is Andrew Barton. So they can reach out to him as well, Bartona@adr.org. they can reach out to me. I'm happy to talk to people as well, part of my job but I think the website is the first place to go. There are demos on there, all kinds of information for people who want to learn more about what this is all about.
Zach Abramowitz: Fantastic. I look forward to this time next year when we get to review the progress of AI arbitrators, and I think this is a genuinely exciting moment. So, thanks for joining.