Since its founding in 1926, the American Arbitration Association (AAA) has provided dispute resolution services to help parties settle their differences out of court. But for more than two centuries, the American civil justice system has remained largely unchanged. In spite of this, the AAA has been committed to building better systems and approaches to transform how the world moves through conflict. Now, thanks to the AAA’s partnership with QuantumBlack, AI by McKinsey,1 McCormack is advancing a new solution to scale resolution. The AI Arbitrator is designed to support the AAA’s case management process, adding a layer of support that may help resolve cases at least 20 percent faster, offer cost savings starting at 35 percent, with a goal to realize a tenfold caseload capacity. McCormack sat down with McKinsey Partners Albert Bollard and Eric Goldberg to discuss the collaboration with QuantumBlack, how the AAA managed to avoid pilot purgatory, and why AI is uniquely suited for dispute resolution.
This interview has been edited for length and clarity.
A legal system in need of an overhaul
QuantumBlack: Can you tell us about the American Arbitration Association?
Bridget McCormack: The American Arbitration Association is a 100-year-old nonprofit that provides dispute resolution services to anybody who needs them. That includes a lot of large complex B2B, as well as C2C [customer to customer] or B2C, disputes. We also provide mediation services and other upstream ways of resolving disputes that keep people out of court.
QuantumBlack: According to the AAA, 92 percent of the American public either lacks access to legal counsel or doesn’t show up with representation in court.
Bridget McCormack: The American civil justice system has turned into a bit of a market failure. I saw it firsthand as the chief justice of the Michigan Supreme Court. The traditional court systems often drain resources, stall progress, and limit resolution. When it was designed over 200 years ago, it worked fine, because every person with a dispute that needed resolving had access to a lawyer. That’s just not true anymore. Four industrial revolutions have passed, and the legal profession has not updated its operating system.
And so now, for example, 92 percent of Americans with a civil justice problem cannot afford legal help. Think of the small businesses, independent contractors, everyday people. Many of them try to figure it out on their own, which is why most state courts have more parties without lawyers than parties with them. That’s a challenge for people, the community, and the courts. When the AAA was founded, it was an innovation and a disruptor, because at that time, the only way to resolve disputes was in a courthouse.
But the AAA paved a new way to resolve disputes out of court—through arbitration—although its mission was much bigger than just resolving disputes. The founders of the AAA believed that by giving individuals, businesses, and governments the power to resolve their own disputes instead of just turning them over to the courts, you would grow trust and confidence in institutions and forge a stronger society.
Using AI to make all parties feel heard
QuantumBlack: What were your aspirations when you took the helm of the AAA?
Bridget McCormack: When I joined the AAA almost three years ago, it seemed pretty clear to me that AI was going to disrupt the business and practice of law, as well as dispute resolution. I viewed this technology as almost a perfect match within the legal vertical. Because in addition to being a magic word calculator—our profession is one in which words are currency—it also has the ability to make parties feel heard.
And if there’s one thing that matters in dispute resolution, it’s that the parties feel like they have been heard and understood. In fact, there’s a lot of scholarship that finds that parties in a dispute resolution process who believe they were heard and understood are far more likely to accept an unfavorable outcome and still grow trust in the system.
AI has the ability to produce a clear, reasoned, predictable result in a dispute, which was obviously a warning shot for somebody running a dispute resolution organization. So we decided we needed to figure out how we could build a trustworthy, governed, reliable, AI-native, dispute resolution framework, process, and decision-maker—a process that could scale our ability to resolve disputes, while still ensuring that parties felt heard. And that’s what we’ve been building over the last ten months with the QuantumBlack team.
Finding the right partner
QuantumBlack: People don’t trust AI, especially when it comes to their jobs. So how did you resolve that trust issue externally and build trust and skill sets with internal users? Is that an area McKinsey helped you with?
Bridget McCormack: We started experimenting with AI across our operations and fundamental core services and products almost three years ago. When I joined, we empowered everybody to start experimenting with how generative AI might affect their workflows. So everyone from case managers to the legal team to the engineering team to the marketing team was encouraged to use it and figure out where it would make a difference.
From there, we started building some smaller point solutions ourselves and with partners. That helped us learn what we could do on our own and what we were going to need help with. All of this learning was in service of significantly bigger builds than point solutions. And our AI Arbitrator, which is built on an AI-native case management platform, was that kind of disruptive large build we needed to get right.
We realized we needed to add expertise beyond just our engineering team. We had to rethink the way business development folks thought about these new products and services as well as the way legal teams thought about governance. So there were many areas that called out for the right partner, because this was something we couldn’t afford to get wrong.
We also knew it was going be a technologically complex build, so we needed the QuantumBlack team to lead us through this process. Beyond the QuantumBlack team, the expertise McKinsey brought to the table, in terms of building a governance structure that we can show our users to build confidence in what we’ve built, was critical—in addition to scoping out what the market might look like and the next steps. That governance makes explicit where human judgment is required, how decisions are reviewed, and how accountability remains with the arbitrator, rather than the technology.
But perhaps the most significant part of the collaboration was McKinsey helping us scale and staff up. So that when they left after we stood up the MVP [minimal viable product], we were in a position to build out the product for the next use case and the one after that.
Avoiding pilot purgatory
QuantumBlack: Many organizations get stuck in the mindset of, “Everyone give it a shot,” and not really thinking about how or if it will affect business. How did you avoid the trap of “pilot purgatory”: pilot programs running everywhere but not actually doing anything?
Bridget McCormack: I do think “pilot purgatory” is a great way to put it. And I talk to so many organizations that are still struggling with moving from trying whatever individuals think might be useful toward trying something with real enterprise value. And that’s complicated.
We were definitely pro experimentation, and we have a shared innovation platform across the AAA as well as a funded and structured innovation program. So we are used to moving many ideas through a structured innovation funnel, and that muscle helped us when we needed everybody to start experimenting with this technology.
We had over 100 ideas for things we could build, and just determining which ones made sense to pursue was hard. We had a set of metrics about which ideas would make a difference across our operations and bring value to our clients. And I think the best idea was what we’ve now built: an entirely new way for enterprises to resolve disputes and move forward.
There might have been a point solution that didn’t make it through our funnel, and that’s okay. We also had some ideas make it through the funnel that didn’t turn out to be useful. And some other ideas helped us learn what we were capable of but wouldn’t have been worth it, since we could just license that particular tool from somebody else.
But ultimately, we learned what we really needed to know to build the future. And that was creating a trustworthy, multiagent system that could produce reasoned, reliable, and transparent outcomes to settle disputes—a system that works in close collaboration with our expertly trained human arbitrators to scale thoughtful, carefully considered awards. And since we needed to get that build exactly right, we had to pick exactly the right partner.
Meaningful, measurable impact
QuantumBlack: What was some of the real impact for both users and the business?
Bridget McCormack: The impact is actually an easy story to tell because we can measure how much more quickly it will resolve cases right now, and we think it will resolve cases twice as fast.
With this MVP, we have a pretty good way to compare it, because we have a process right now where people can go through our core service with a documents-only construction case, and we know how long that takes and generally how much it costs. Some cases can cost more than others if the parties need more time.
And we are confident that right out of the gate, AI Arbitrator is going to resolve cases at least twice as fast, serving as a decision support layer embedded in the workflow while human arbitrators retain ultimate judgment. It should even help speed up more complex disputes that will require fully bespoke, more intensive human-led resolution pathways. And the cost reduction is going to be at least 35 percent. And it’s only going to get faster and less expensive for users as it works and learns.
In addition to that, it’s going to dramatically increase our overall capacity at the AAA. Right now, the AAA resolves or administers about half a million cases a year. But it does all of them in a human-heavy process, and with this tool, we think we could do ten times that every year. And this newfound ability to do significantly more work for more users is what’s most exciting of all.
