In AAAi Podcast Episode 20, Bridget McCormack and Zach Abramowitz recorded live from Legalweek in New York where one thing was clear: the conversation around legal AI has shifted. The technology is no longer theoretical. It’s been tested, it’s value is clear, and user behavior is already changing.
The industry is adopting AI, and the focus now is on how to deploy it in a way that builds trust and scales safely. That’s where governance comes in.
Bridget’s Legalweek panel, featuring leaders from Freshfields, Moderna, and OpenAI, focused on how governance is actually being implemented inside organizations. The conversation reflected a consistent theme: the teams moving fastest are the ones that have figured out how to govern what they build.
Key Takeaways
- Governance, done right, accelerates speed: Governance has traditionally been seen as a constraint. Leading teams are reframing it as an enabler. At the AAA, governance drives speed because it is integrated across teams and continuously evolving. It is not a static risk framework updated once a year. It is a living system that adapts alongside the technology. When designed this way, governance creates momentum instead of friction and builds the trust required for real adoption.
- The risk equation has changed: Legal organizations have long managed risk by slowing down decisions. That approach is becoming outdated. Today, the risk of inaction can outweigh the risk of early adoption. Delays, whether from extended testing or layered approvals, can quickly become liabilities as technology and markets move forward. Forward-looking organizations are adjusting by engaging legal teams early. By solving for the most skeptical stakeholders first, they build more trustworthy systems and unlock faster, broader adoption.
- Adoption requires a clear, repeatable playbook: AI adoption does not happen through access alone. It requires structure. At the AAA, that starts with leadership commitment. If the C-suite is not actively using the tools, adoption stalls. Next comes enabling the broader workforce through access, training, and encouragement. Finally, organizations need to let champions emerge organically. These early adopters, often from unexpected places, help turn experimentation into sustained, organization-wide change.
Final Thought: Governance Is What Turns AI Into Infrastructure
Episode 20 reflects a turning point in the conversation about Legal AI. Legal AI no longer has to prove that it works. Now, it’s about building systems people trust within organizations ready to use them.
The next phase of legal AI will not be defined by better models. It will be defined by better governance.
Transcript
Zach Abramowitz: So, Bridget, this is our third Legal Week in a row recording podcasts right here in New York at the conference. This is the third year in a row that you've presented at this conference, and yet this year feels very different to me from three years ago. It feels very much like where we thought this might head, but with some surprises. It is just an incredibly exciting time. I wanted to start, before we get into big-picture takeaways about Legal Week, with the panel that you hosted. There were some very serious heavy hitters on this panel, which was standing room only. I was there, and many people I saw were taking notes. Talk to us a little bit about who was on this panel and the topic
Bridget McCormack: It was a great panel. It was on governance, but more specifically on how AI governance and legal oversight is critical to AI adoption. It turns out governance is actually a piece of the formula for success in integrating AI into your business. Nobody knows what they're doing with it because it just requires some new governance muscles. For example, the governance has to be living, because as quickly as the technology is changing, the governance also has to change.
Bridget McCormack: Actually, the Legal Week organizers asked if we had any expertise in that. In fact, at the AAA, we have a lot of expertise in it. We're three years into building AI across our operations and products, each of which requires slightly different governance. As a result, we have a pretty sophisticated governance approach that we're proud of. We’ve done some talking about it and giving free advice to other businesses that are trying to figure this out. So when the Legal Week organizers said, "Do you guys know anything about governance?" we're like, "In fact, we know quite a bit."
Bridget McCormack: We even have a survey in the field now. We can talk about that in a future episode when we get the results back. I had a great panel. I had Henry Hagen, who is Associate Counsel at Moderna. Moderna was one of these very early company-wide adopters of AI: OpenAI's products in particular. The CEO at Moderna had the legal team lead adoption, so they had to go first. I think that was super smart. Then I had Galia Amram, who is Associate General Counsel at OpenAI. Galia and I have ended up on panels together in places about AI across the world, including most recently Tel Aviv. Galia focuses on safety in her role in-house at OpenAI.
Bridget McCormack: Finally, I had Anna Gressel, who I think is a really impressive, forward-thinking leader on governance. She's a partner at Freshfields and their Global Co-Head of AI. She's advising all kinds of enterprises about AI governance, so she has experience across the enterprise spectrum where businesses are succeeding and where they’re not. She's super thoughtful about the topic, and it ended up being a fantastic combination of perspectives on a topic everyone is thinking and worried about.
Zach Abramowitz: Yes, and sometimes at conferences, a panel is just something you do because you brought everyone together and there has to be some content. But really, the most important part of the conference is getting people together in between. I actually thought this one felt incredibly meaningful. I would say that Anna in particular, the very fact that she's doing what she's doing in terms of advising companies on their AI rollout, felt like a data point in its own right. There is a partner at Freshfields who is the Global Co-Head of AI, who's not just working on making sure their lawyers internally use AI.
Zach Abramowitz: She's actually taking an active role and brought a perspective that was not just "we're going to be so careful" to the point of non-adoption. That is something you and I spoke about years ago: cutting red tape and making sure the path to adoption wasn't held up by endless committees or cumbersome policies. Her role, and the fact that it exists, is such an interesting data point. You've had 24 hours now to digest. What are your top takeaways following this panel?
Bridget McCormack: One was that the theory of the panel turned out to be right. My view was that governance is part of what’s allowed us to move fast. Our governance is different from governance across other enterprise risks. It’s integrated across teams, iterative, and changes constantly. It isn't the static ERM program you're used to across the rest of your enterprise risks, but I think it’s why we’ve been able to move as fast as we have. We are building a clause builder product with AI, which has different risks. But I don't know if that's true in other places. Moderna is Pharma. It's a highly regulated place where you could imagine maybe a different approach than we're building a clause builder product with AI, there are different risks. But I think it turned out to be right that getting governance right is critical to success. You heard that from Anna, who sees it across enterprises, and Henry, who is in the middle of one of the most successful enterprise-wide adoptions in a highly regulated industry.
Zach Abramowitz: Just to reflect on that: governance is not a side issue. If you view it as a key part of adoption, you understand that the biggest issue people have is trust. We’ve spoken before on this podcast about elevator technology working for 10 years until the Otis experiment: until the ability to break was demonstrated. That trust level is so key with new technologies that can cause a bit of anxiety that feel frightfully new. I thought that was a really interesting point. Anna specifically made the point that you don't want to over-govern or risk non-adoption. She said if you compare the risk of adopting AI to the risk of non-adoption, non-adoption can be existential. Your competitors and the market are moving. The world is changing. Non-adoption is not an option.
Bridget McCormack: That was the second takeaway for me: how non-adoption is the bigger risk. That has been my view in the business I’m in for a long time, but we heard that clearly from every panelist. Anna and Henry in particular said it quite loudly. Galia's job is really important but pretty narrow: she’s focused on the governance of the actual models that the rest of us are now governing in our building of products on top of them.
Bridget McCormack: But Anna and Henry both said, "as opposed to what?" You are a lawyer, you can always think of a way to get slightly more comfortable if you wait a little longer, do another set of tests or send it to another committee. But the risk of not moving is more significant than moving. Hearing that from lawyers in particular is still not super common. I think there is still a lot of reservation about risk with this technology because it is different from previous ones. You and I talked to probably more lawyers with that view than most people, but I also talked to lots of other lawyers and I think there's still a lot of reservation about risk with this particular technology because of some things that are different about it than previous technologies.
Zach Abramowitz: And the fact that they started adoption with legal. Maybe that's the recipe. Maybe the recipe for correct adoption is to start with the people who might be the first to shut down AI in the organization. Because they're typically the folks that have to focus on risk, right?By starting there, let's build trust where it might be the most difficult, and then empower those people to let others in the organization trust the AI . I think that this is one of the things that I've seen in legal departments is a lot of the folks who are looking at AI tools maybe in 2024, and that was like their mandate, if you compare now in 2026, they're actively involved in creating governance policies and guardrails for various agents being rolled out across their organization. But now again, having listened to the panel yesterday, my takeaway was maybe Moderna specifically got it right by starting in Legal.
Bridget McCormack: I have to assume that was part of why they started there. If you start with the team most likely to say "this is too risky," your chances of success go up significantly. At the same time, you get that team thinking about what guardrails every other team needs, and they get a head start instead of playing catch-up. When lawyers feel they're playing catch-up, their first go-to is often "no, don't do it." That's just what we were taught in law school. They were put in a position to figure out how to say yes from the beginning.
Bridget McCormack: My third take away is that agents are making governance a "do-over" almost. Not a do over but it’s requiring us to think again about what governance looks like in a world where agents are doing things autonomously. Not every agent is one you have to worry about with the same precision. Some are looking across internal documents for an outcome, which isn't as risky as a product people are using. Or in Moderna’s case, even more so, a product that may result in life or death. Where if agents are making decisions, you're going to need to figure out what your governance looks like.
Zach Abramowitz: as opposed to reviewing an NDA
Bridget McCormack: Exactly.
Zach Abramowitz: Not the same risk factor.
Bridget McCormack: Exactly. Lawyers aren't always good about making that distinction.. But in an agentic future where we have agents on our teams, conducting commerce or business for us, we'll have to figure out the right governance structure for each kind. The human in the loop, the fallback for every governance talking point that falls apart when you have agents at scale negotiating agreements, 24 hours a day and they never get hungry or tired. That doesn't work. So you're going to have to figure out new approaches.
Zach Abramowitz: I'm glad you mentioned this because AI is moving really fast. Many people are still being given the governance frameworks and mental models of two or three years ago. You and I discussed this in the podcast with Jen Leonard that we did, recently that AI draft me my email might have been a really good use case in 2023 and 2024 when you're trying to learn about where could I use AI? Today I don't think I'm drafting very many emails using AI If anything, my emails have now become a little bit more human. Less AI applied and I think in the same way, if you look at like Human in the Loop. "Human in the loop" was built for a technology that was phenomenal but deeply incomplete. We saw that with fake citations and hallucinations. If your entire governance model three years later is still just "human in the loop," you're talking about a framework that is kind of not relevant anymore, especially considering the impact agents can have. Which was a big part of the panel discussion. Part of what an agentic structure does is actually make the model better. In some cases, you don't necessarily need some of the same governance structures that you would've used in the past. If the calculator says two plus two equals four, you don't need a human to make sure that four is correct. It changes the governance structure.
Bridget McCormack: I think that's right. None of us have fully wrapped our minds around what this agentic future looks like. I'm thinking about this in terms of the dispute resolution layer. When we have millions of agents negotiating and executing contracts, what happens when there's a dispute between the agents? It's a totally different set of stakeholders and disputes that can be resolved with different processes. I don't think anyone is thinking about that quite yet, but there will be interesting governance questions around it.
Zach Abramowitz: Mindset is such an important feature as well. One of the things that I've seen is that early adoption is not just about knowing how to use AI, but it really is wrapping your mind around AI. This is underappreciated. When you begin to really wrap your mind around AI, you start to understand it, but you also kind of start to see the dot, dot, dot. Now that we have three years of progress to look back on, it's becoming a very real question: "Now what does this mean?" Galia mentioned that she's in child safety at OpenAI: one of the most sensitive areas. I could easily have seen her saying to be really careful, but her message to lawyers was: "If you have not sunk your teeth into this yet, do not wait. Go home right now and do it." That is one thing to hear Sam Altman say. But to hear that from the person dealing with the riskiest and most sensitive issues at OpenAI was a really interesting point.
Bridget McCormack: She was pretty clear-eyed about that. She's responsible for some very important safety work at OpenAI, but she also has a firsthand view of the power of the technology to make legal work better. So I get it.
Zach Abramowitz: I wanted to highlight one other point from the notes I took during the panel. I think the AAA has done this effectively. It's crazy that we're three years out and the AAA has one of the best operating playbooks for general adoption in an organization, a hundred-year-old organization. We just celebrated the centennial. So it’s really exciting. You've done a fantastic job as an organization creating a real AI culture. One of the things that Henry Hagen from Moderna suggested this idea of a GTC: a Generative AI Champions Team, about 10 to 15 percent of the organization who are leaning in hard.
Zach Abramowitz: In the AAA, I think it's more than that. I've seen this from talking to case managers who are vibe coding solutions today. There's really not just a top-down but also a bottom-up culture of adoption. This is not the old days of a single developer at the top of the organization and anything we want developed has to go through them and there's a long sort of laundry list and I'll get to your project when I get to it. Today, the power is in the hands of anyone in the organization who chooses to go in that direction.
Bridget McCormack: Janet Miranda showed me yesterday an app, she vibe coded on Base44 over the weekend to help her more quickly capture leads at Legal Week. She literally built an app over the weekend.That’s what people at AAA are doing because they've been empowered to do it. I think of them as our "lab." I believe there are three pieces to this, with a hat tip to Ethan Mollick for the names.
Bridget McCormack: First, your organization has to have leadership commitment. If your C-suite isn't a hundred percent bought in and using the technology themselves to understand where it's going, you're going to stall. I think there's just no way around that.
Bridget McCormack: Second, you have to enable the crowd to use it. And that means everybody across your organization, they won't all be champions, but the only way you find your champions is by giving your crowd permission. Not only permission, but some training and encouragement however you encourage your employees to do things,
Bridget McCormack: Then you find your lab. You find your champions, because you never know where they're going to come from. There'll be a case manager who figures out all the different ways this can make case management better, not only for case managers, but even for users of your system. And you'll find somebody in marketing who figures out that you can do things you never could do before. And they become your champions. And then you can have them go sit with people who are having a harder time figuring out how the technology changes what they do. And it's a game changer. It's definitely our playbook
Zach Abramowitz: What you said may sound simple for people who are listening to this,but I hope they pay deep attention because that order is also critical. Yes. If you can't convince your senior team and get them on board, you're going to stall. So before you try to diffuse the technology throughout the organization. Time has to be spent with the leaders. And, if they're not seeing the eureka moment, you got to keep showing them. It's not a one time conversation. I think a lot of people think we had our C-suite training. I think that there's a lot of persuasion that goes into it and then giving the technology to everybody, not just sort of selecting you're going be the champion, or you're gonna be the champion. Empowering everyone and then seeing who emerges as the champions. And you said another thing that I've now seen at so many orgs and I have to continue now counseling other clients, along these same lines, which is it's not necessarily the people that you expect. Sometimes it is. But sometimes the people who are getting involved and building tools and taking advantage and not just trying to automate what they do, but
Bridget McCormack: seeing the potential?
Zach Abramowitz: How can I become a Superman? Yeah. How can I do something that I never did?
Bridget McCormack: Now I can do someone else's job in addition to my job.
Zach Abramowitz: Yes. And you can't assume just because someone has not necessarily taken initiative in your organization before. This might be their moment. Sometimes, it really comes from places that you wouldn't expect.
Bridget McCormack: A hundred percent.
Zach Abramowitz: Moving beyond the panel. What are some of your early takeaways from Legal Week itself? You know, this is the first year at the Javits Center, so it feels different.
Bridget McCormack: Feels different. I like the Javits Center better. I don't know about you. First of all, I'm not lost all the time. I could never figure out where I was in the Hilton. I literally couldn't get to the same place twice; someone had to always show me. But it feels like we've grown up. Legal tech has grown up a lot. Big time. The exhibit floor I thought was way easier to get around and see what's happening and see who's where. And then the rooms for presentations were better. I thought everything about it was better.
Zach Abramowitz: So it's like the user experience of AI being better has also extended to the in-person events.
Bridget McCormack: Yeah. To the in-person events.
Zach Abramowitz: Right. It didn't feel like the kind of crowded throng that it always felt like in the Hilton.
Bridget McCormack: Less janky. It's less janky. Whenever anything gets less janky, I like it better.
Zach Abramowitz: Now what was interesting was the folks you had on the panel, let's go back to that, none of them talked about using legal-specific applications.
Bridget McCormack: It's true.
Zach Abramowitz: That's fascinating. At Moderna, the partnership was with OpenAI. Anna didn't talk about this at Freshfields, but their primary partner in AI has been Google. They're unique in that respect. And obviously Galia at OpenAI. But even at the AAA, you know, you have built the AI Arbitrator and Resolution Simulator also on the foundation models. And you've built that bespoke. So my question is looking at the exhibit floor, and maybe this is too simple a narrative, but in five years from now, who's still going to be there?
Bridget McCormack: or there's just one big Claude booth One big Law booth? Exactly.
Zach Abramowitz: One BigLaw booth. Exactly
Bridget McCormack: There's like Gemini and Claude and OpenAI and that's it.
Zach Abramowitz: How are you thinking about this right now? And again, I know it's impossible to predict the future on this.
Bridget McCormack: Yeah, it's hard to say. I mean, obviously from where we sit, what we're building, we alone have the expertise and the data to build the tools that make dispute resolution better.And so building on top of the foundation models is a no brainer. There's nothing that another layer on top of the foundation layer would give us that would help us build better.
Zach Abramowitz: There isn't a tool out there called RB.
Bridget McCormack: No. I don't know RB yet, but we need to be RB that's us. But we're a little bit different. We're not a law, we're not a law firm. So, you know, law firms sometimes need some comfort level with another vendor having built the guardrails and the UI, UX for them. That makes it easier for adoption and easier conversations around risk. And so maybe it
Zach Abramowitz: feels fancier also, w
Bridget McCormack: I think it’s part of the culture. A couple of them are, I think, doing a pretty good job getting into the fancy market. So I don't know what that looks like in five years because everybody read the viral like X post from - the Claude law firm guy.
Zach Abramowitz: Zach Shapiro. Yes
Bridget McCormack: Zach Shapiro. Sorry, I should have known, how did I not remember? He's a real guy, running a real practice with Claude and it's unbelievably successful and easy. Right. So there's going to be more of those, and smaller practices like his, where lawyers say, I'm not sure I need anything other than the frontier model. It feels like the Wild West for legal businesses, like what they might buy. On the other hand, you and I have talked about the startups building bespoke platforms, processes, services for legal businesses who have never had any love from legal tech. Right,
Bridget McCormack: And that feels like there's a "there" there to me. Because there are definitely pockets of legal where lawyers definitely still have files and file cabinets and they have long paper. They have a longer way to go. They're not gonna start using Claude to build the tools to fix their criminal defense practice or their immigration practice.
Zach Abramowitz: A few years ago I was staying in downtown LA, right near the courthouse. And my hotel was basically like a litigation war room for multiple firms. I remember every night seeing the same person coming in with his big accordion folder. Amazing. And I was just like, that is the AI that's being used right now by the litigators. In certain respects, I think it remains kind of an old-school business.
Bridget McCormack: Absolutely. It's not a surprise that there's so much capital interest in legal businesses right now. The confluence of what technology is enabling and a little bit of regulatory cracks means no surprise that capital is finding its way to legal businesses. I don’t know what that's going to look like in terms of AI tech for lawyers. I can't wait to see if we walk the floor in 10 years who's still going to be there. Will it just be Dario and Sam Altman? That's it. I don’t know.
Zach Abramowitz: I think the other thing that I started noticing earlier this year is we're getting a lot more companies that are focusing on AI for litigation. Now, I don't mean just in the discovery sense, which is where it's always been. For years, the joke about the Legal Tech show was that it should just be called the eDiscovery Show. These new tools are not focused on discovery, but really on litigation storytelling and strategy.
Zach Abramowitz: I flag that because I think the first use case for AI was contracts. That's where everyone went with their brain. Boring use case. But contracts are also not necessarily taking full advantage of what a language model can do. Because a large language model is so much
Bridget McCormack: more creative.
Zach Abramowitz: deeply creative. Yes, totally. And, I think, this is again, where I hope the frame is shifting a little bit instead of just thinking about, oh, I use AI to do the sort of mundane, boring tasks. So that I can focus on high level strategic work. Have you ever thought about maybe using the AI opposite?
Bridget McCormack: Opposite, yeah.
Zach Abramowitz: To help you with high level strategic work. So I think the move into litigation is also very much like AI 2.0 and it is a real sign of maturity. Obviously you've thought about AI in the context of dispute resolution and disputes long before a lot of these companies have emerged. But I think now we are starting to see a meeting of the minds and I think more people over the coming year, more litigators are going to start taking advantage of these tools because it's really powerful.
Bridget McCormack: Yes there are businesses, lawyers have dispute portfolios and there are some disputes for which really high-end AI tools, storytelling, could be a game changer. There are lots of others where I think all you want is like the AI arbitration tools we're building to get them resolved. Lawyers have dispute portfolios. There are some disputes for which high-end AI tools and storytelling could be a game changer. There are lots of others where I think all you want is like the AI arbitration tools we're building to get them resolved in a significantly more efficient way. Everybody's going to be able to start thinking about their disputes across a portfolio: which ones need high-end creative strategic tools to take a $50 million case and make it a $500 million case, and which are better funneled into "let's get these resolved in 30 days."
Zach Abramowitz: I want to ask you before we conclude about the recent launch of Resolution Simulator. I think it's a very important part of the AI Arbitrator which you've released. I have my own thoughts about it, but I wanted to let you explain to the audience what this is and where it came from.
Bridget McCormack: The Resolution Simulator we're pretty excited about. It actually didn't take a lot of work because it's built on the AI arbitrators Platform and we took a lot of time and care to build that in an agile way over 10 months. And it's really a bunch of agents who operate across the dispute resolution process. But the resolution simulator is really a direct response to what we were hearing from the market, which was can we take the AI arbitrator into our own environment and just figure out the value of our case so we know at the front end is this the case we want to put a bunch of time and money into? Or is this one we probably can move more quickly towards some kind of settlement? We heard enough of that, that we were like, sure, that's actually not that hard for us to build given what we already built. Now parties are going to be able to do that.
Zach Abramowitz: I think it's such a game changer for a few reasons. Number one, it gets past the point of how can AI automate the things that we already do and begin to get you into the world of what I would not have even thought of doing?
Bridget McCormack: Yes
Zach Abramowitz: Once upon a time. But now that I have AI, we can do that. So if there is an AI Arbitrator by definition, now there is a way of like doing that simulation,
Bridget McCormack: right?
Zach Abramowitz: And it's one thing to say, okay, well, here's our evaluation of your case and here's why we think it's strong or not. The incentive of a lawyer might not be the same incentive of a client when it comes to resolving a dispute.
Bridget McCormack: Absolutely.
Zach Abramowitz: And now there's both a way for the client or the lawyer to have a little bit of data and evidence on how this is going to be decided? And I think that, running a simulation once. I think I might want to run simulations of this like a hundred or a thousand times to see hey listen, I know you think the case is so strong. Why is this coming out with us not getting our way 90% of the time? Right? And maybe we ought to reconsider that. And to me it's such a game changer. I think the other thing is the AI arbitrator requires the consent of two parties.
Bridget McCormack: Yes exactly.
Zach Abramowitz: And it's new. There's not been anything like this in history. So for both parties to, to agree to it, to get
Bridget McCormack: comfortable,
Zach Abramowitz: they need to see how to get comfortable. They need to be able to see how it works. And I think this gives that perspective that the simulation that we saw was pretty great. Why aren't we just considering this as the way to resolve this dispute, right? This doesn't, right, this doesn't need a human arbitrator. This should be quick in and out.
Bridget McCormack: It significantly reduces the information asymmetry that, I love lawyers, some of my best friends are lawyers, but that lawyers have put a little bit of a wedge in between them and the public they serve for a long time. I think ultimately that's to the benefit of everyone: clearly to the benefit of the public to have better information earlier about what the options are for disputes they're dealing with, and ultimately better for lawyers as well because there's an enormous market that is underserved that now we have technology to build business models that will allow them to serve it more happily. I think we have to work through some things over the next few years, but at the end of it, it's good news for everybody.
Zach Abramowitz: Bridget McCormack, I look forward to getting together again at Legal Week next year and seeing how far things have come.
Bridget McCormack: Can’t wait
Zach Abramowitz: This has been really fun. Thanks for tuning into today's episode of the AAAi Podcast. Please consider leaving us a rating and review on your favorite podcasting platform. And don't forget to follow and subscribe so that you never miss an episode.