The New Legal Skill Set
Judge Maritza Dominguez Braswell on AI Guidance, Attorney Development, and How JAIC Is Transforming Judicial Education
The rapid emergence of generative AI is forcing a fundamental reevaluation of how legal work is done and how justice is delivered. This technology presents a complex duality: it offers the potential to bridge the long-standing access-to-justice gap through unprecedented efficiency, yet it demands a rigorous new framework for ethics, verification, and judicial oversight.
Judge Maritza Dominguez Braswell is at the forefront of this new frontier and has spent the past few years educating legal professionals and judges on generative AI and its impacts. In addition to her role as a United States Magistrate Judge for the District of Colorado, she curates a monthly newsletter for judges around the country, teaches AI and law and the University of Denver, serves as Co-Chair of the District of Colorado’s AI Committee, is an active member of The Sedona Conference Working Groups 1 and 13, and a co-founder of the Judicial AI Consortium (JAIC).
Judge Braswell spoke with Everlaw about AI guidance, the skills young attorneys should be prioritizing most, JAIC, and more.
Prior to joining the bench, you co-founded a tech-forward litigation firm. Can you talk about how you leveraged legal technology in your work during that time, and how that experience has informed your perspective from the bench?
I founded a small litigation boutique firm over a decade ago. We were small, but mighty. We did a lot of work in New York, primarily in the financial services space, which meant we were essentially competing for work with big law firms.
We had to figure out how to be nimble and leverage technology to stay ahead. A lot of our work involved financial services, more specifically, the residential mortgage-backed security space. Even though we were lawyers, we were pouring through loan files during claim review.
We collaborated with a tech firm to process that review faster. They developed very customized databases that helped us triage information, and we co-developed decision trees to work through the files. This helped us cut down review times significantly, from years of work down to months.
Additionally, around 2015 or 2016, we were exploring AI tools in the research space. It was very underdeveloped at that time, and wasn't anywhere near where it is today, so we ended up not adopting any of those tools, but we were already looking into them.
"I want to encourage folks to learn how to use AI. But you can't learn to use it responsibly unless you're engaged, with your sleeves rolled up, trying new things."
I've always been really interested in finding new solutions to old problems. This idea of, “This is how we've always done it and this is how we will always do it,” just never quite fit into my head. So, when I hear someone say that, it's immediately a flag for me to say, "Let's dig a little deeper. Let's figure out where we can find some really great innovation opportunities.”
In terms of how it informs my work on the bench today— soon after stepping into the role, I saw opportunities. For example, I kept hearing the same pain point from practitioners when I was out in the community: too many practice standards to track. I recruited an intern to help me pull everyone’s practice standards together, we presented the collection to the team, found the overlap, had some discussions, and a few months later we all signed onto one uniform set of practice standards.
When modern AI made its big splash, I saw another opportunity to think differently about how we do our work, both as lawyers and as judges. I dove deep into the AI space pretty quickly and started reading, learning, experimenting, absorbing as much as I could.
You’ve positioned yourself as an advocate for generative AI and have prioritized creating educational content for both attorneys and judges on its use, writing frequently for Thomson Reuters, developing a podcast featuring conversations with other judges, creating a judges-only AI newsletter, and more. What do you see as the value of creating this content, and what do you hope legal professionals will get out of it?
I really do think AI is going to be transformative. Whether we participate in that transformation process or not, it's going to change the way justice is delivered. It's going to change the legal profession and the ecosystem that we all operate in.
The only question is, who's going to be part of it? Who's going to have a hand in shaping all of that? I really think it's important for both lawyers and judges to be active in that space. Because if we're not, then the space is going to be redesigned by folks with different priorities. I don't assume that tech companies are going to have ill intent, but they're going to be focused on things like efficiency, cost, and speed. In the law, we need to be focused on fairness, equity, and access for everyone. And those two mindsets are different.
"Don’t begin with AI, and don’t end with it. Instead, use it in the middle of your process to augment, not automate or replace."
It's important to involve judges who don't have a vested interest in the particular outcomes of a case, but rather want to deliver good process, and fair consideration. We want to have enough time to hear everyone as best we can and those are the types of motivations that need to be injected into this redesign process.
That's what I really view this as. It's an opportunity to redesign how we deliver justice. We've had an access to justice problem for decades, and we've been nibbling around the edges trying to figure out how to solve it, but we can't quite figure it out. AI presents an opportunity to really step back and say, "You know what? It's a different world today. People now have access to knowledge at their fingertips. How do we design a system that delivers justice in that context?"
You’ve advocated for an “AI sandwich approach," where humans begin and end the legal work while AI supports the middle of the process, with the idea that this helps legal professionals use AI responsibly without compromising ethical standards. Can you speak a little more to that approach?
I'm optimistic about the future. I want to encourage folks to learn how to use AI. But you can't learn to use it responsibly unless you're engaged, with your sleeves rolled up, trying new things. At the same time, we have to figure out how to do that responsibly. There aren't a lot of guardrails right now, there's no comprehensive federal legislation, there’s no top-down uniform directive for judges or courts. So we have to ask ourselves, “How do we do this responsibly? How do we engage, learn, and experiment in a responsible way?”
And it can be confusing because we judge this technology by standards that apply to traditional compute. We assume it offers a fixed answer the way a calculator might, or that it retrieves specific information the way Westlaw might retrieve a statute, but that's not how these tools work. They are probabilistic and there are different types of risks that warrant different types of guardrails.
One practical way to create guardrails for people who are beginning to experiment with modern AI is the sandwich approach. At least that’s what I call it. It's kind of silly, but it's a good way to remember to put your AI use in the middle. Don't start with AI because, one, AI is biased and can produce biased results. It can lead you down a slanted path without you even knowing it.
Two, it may induce overreliance.
And three, it becomes difficult to discern between AI outputs and your own fresh reasoning. So start with your own thoughts, wrestle with the blank page, and sit in your own first pass analysis before you introduce AI.
Next, figure out how AI can augment your work. Can it surface gaps in your knowledge, or identify inconsistencies in your thinking? Can it help you better understand the particular legal framework you're dealing with? Can it formulate questions you hadn’t thought of asking? Stakeholders you didn’t consider?
Then bookend that work with traditional research tools to verify the results. Given the risk of hallucinations, traditional tools remain critical for cite-checking and final verification.
So that’s it. Don’t begin with AI, and don’t end with it. Instead, use it in the middle of your process to augment, not automate or replace.
You have chosen to issue AI guidance rather than a formal standing order, noting that mandatory disclosure requirements might create a false sense of security. Why do you believe this more flexible guidance is better for attorneys?
I think AI standing orders create a false sense of security. They don’t address the underlying problem. I also think they can have a chilling effect in a way that worsens the access to justice problem. Big firms aren't going to stop using AI. They're going to keep pouring millions of dollars into this technology. The folks that it chills are the solo practitioners, or the pro se litigants, or even the smaller firms that may have enough funds to use some AI, but not other, more powerful AI. They're thinking, "I can't afford to get called out. I should just abstain altogether." I think that worsens the access to justice problem.
And more importantly, it distracts us from the bigger problems. To me, hallucinations are just scratching the surface. Lawyers have been making mistakes in briefs forever, way before AI. This is just another version of the same problem. And I can usually spot a hallucination pretty quickly.
I'm more worried about deepfakes, which are harder to spot. I don't have the technology to catch them, and the opposing party may not spot them either.
And, like I said before, I’m concerned about bias, especially for judicial use. I'm worried about these AI systems leading us in a certain direction without us knowing it.
Our job is impartial truth-finding, and putting together a process that will help us achieve that. A big part of that process is evidence. That's at the heart of it. If deepfakes throw a wrench in that, and AI interferes with, rather than promotes, impartiality, it could exacerbate the public trust crisis we already face.
So, no, I have not issued a standing order on AI to combat hallucinations. I chose to do something different, which is to offer guidance, highlight different risks, and remind people to give each other a little grace as we get through this transition period and learn and grow, together.
You co-founded JAIC as a “judge-led educational consortium for judicial officers who want to engage about AI seriously, without hype or speculation.” Can you speak a bit more about the work you and your colleagues are doing to educate your fellow judges on the importance of AI education?
I’d been doing an AI-focused newsletter exclusively for judges, and really enjoyed the experience. I thought to myself, “I'm deep on these issues, and others should benefit from it. Not everybody's going to have the time or the desire to research this stuff, so if I'm doing it anyways, I might as well share what I'm finding with folks.” It was very well received. I'd get emails after every newsletter I sent out that said, "Thanks so much for taking the time. This is great."
I thought about what might be more helpful than just a newsletter, and what could take this exchange on AI to the next level. I figured there had to be other people who were also digging into AI.
"I can tell when self-represented litigants are using AI. Sometimes they're not using it very well and I need to caution them. But oftentimes I see it making a positive difference."
I looked into creating a peer group. The Sedona Conference brings groups together on different topics, including AI. There are also ABA committees and other associations that bring folks together on this topic. But the reality was most of these weren't solely for judges.
Judges don't always feel comfortable sharing their perspectives in a mixed setting because we have different considerations. We have to be thoughtful about what we say and how we say it. We don't want something to be taken out of context and used to suggest a particular inclination.
Additionally, the educational platforms that were out there specifically for judges were not AI-focused. There wasn't really a place where it was just AI and just judges. So that's how the idea started.
I reached out to a few judges and asked what they thought about a co-learning space, and everybody very quickly responded that they loved the idea and would want to participate. I talked to Judge Schlegel and Judge Rodriguez, who are very active in the space, and the collaboration was effortless. Within a couple weeks we had JAIC off the ground.
Does JAIC have anything on the roadmap for the next six to 12 months that you're excited about?
Our signature initiative through JAIC is the Pop-In. We've done three already, and we're going to do another one in May.
Pop-Ins are informal, small group sessions where judges can pop in and just have a conversation. There’s no preparation necessary. You can just sit and listen if you want. But if you have some relevant experience or something you want to share, or a question you want to ask, you can ask it.
We have a dialogue leader for each of the Pop-Ins. The Executive Council of JAIC, Judge Schlegel, Judge Rodriguez, and I led the first three. Judge Thumma is going to lead the next one.
I'm very excited to continue those. We’ll probably do two to three a month for the foreseeable future. We have about 300 judges signed up so far, and we need to have enough Pop-Ins to ensure those groups are small. We want to keep them around 10 to 20 attendees.
"There's another set of skills that will be essential for the lawyers of tomorrow: critical thinking, social awareness, excellent communication skills, and leadership skills like change management and consensus building."
We’re also creating a Contributor group, which are folks who have signed up to say, "I want to do more than just be a member. I want to help and lead." That group is about 40 large right now, and we're meeting in May to talk about next steps and initiatives.
We're also thinking about a resource bank. We've talked about potentially putting out JAIC-created resources, as well as a wider collection of resources. But that just depends on where the Contributor group wants to go.
Your decision in Morgan v. V2X highlights a significant technological gap where high-security, enterprise-grade AI tools are often too expensive for pro se litigants. Do you worry that these types of "safe" AI tools will become a luxury of the elite bar, further widening the access to justice gap?
Yes, I flagged that issue in this case, and I won’t say anymore on the case itself. However, I will say generally that I have serious concerns about the access to justice gap. As firms pour millions of dollars into this technology and self-represented litigants are left to use less sophisticated tools, it could certainly widen the gap.
That said, even consumer-grade tools are pretty powerful. I’ve seen a difference just in the pleadings in my courtroom. I can tell when self-represented litigants are using AI. Sometimes they're not using it very well and I need to caution them. But oftentimes I see it making a positive difference. A complaint that I could not decipher now suddenly is something I can understand. Facts they couldn’t articulate before are now plainly stated, and I can see where they line up with their claims and the law.
So, short answer: yes, I worry about it, but at the same time, I see a real opportunity for this powerful technology to make a difference in how self-represented litigants navigate the system.
In the Morgan decision, you ruled that the identity of an AI tool is not protected work product because the opposing side has a right to know where their data is being sent. How can firms balance the need for transparency with the desire to keep their proprietary workflows confidential?
Conferral has always been an important part of the process. I think it's more important now.
And importantly, it has to be conferral in good faith. It's required by the rules already, but it’s more important than ever. Because if both sides are approaching this in good faith and being honest and transparent about what they’re comfortable with, it will reduce anxiety, increase trust, and promote collaboration.
It's also helpful to the judge because if the parties come to me and say, "We've talked through this. We spent 30 minutes walking through our tools and how they're going to work, and this is the provision that we want in this protective order," I'm not going to get in the way of that. I'm going to say, "Great. You've negotiated that, it doesn’t raise any concerns for me, and if you think it works for your case, let's do it."
But if the parties come to me and say, "We can't agree. We've been talking for two hours and we can't agree.” Well, now you're handing it over to me to craft language. And guess what? I'm probably not in the best position to do that, because I don’t know your case like you do. I'll do it the best I can. But I don't have access to your systems. I don't have insights into what documents you're most concerned about. So, if you want to hand it over to the judge, go for it. But you're much better off conferring and coming to an agreement with the other side and keeping control of that process.
You’ve compared the AI revolution to the industrial revolution, suggesting it’s a similar period that we must and will adjust to. However, for the young lawyers whose tasks are being automated, how do they get the stand-up time and experience they need to become the experts of tomorrow?
I actually just finished teaching a class at DU Law on this exact issue, because I do believe that law students have to think a little bit differently about their careers and the skills they're developing.
To me, preparation for the future is less and less about traditional skills like research and writing. Those are important, and I'm not saying that law schools shouldn't teach those. But there's another set of skills that will be essential for the lawyers of tomorrow: critical thinking, social awareness, excellent communication skills, and leadership skills like change management and consensus building.
They’re also going to need hands-on experience with AI. They'll need to know not just how to prompt a typical consumer-grade AI tool, but also how to leverage and orchestrate with much more sophisticated tools.
The class I taught had three parts to it.
The first was centered on critical thinking. We learned about the technology and how it actually works so that we could think critically about the risks and benefits.
I also challenged them to think critically about the justice system. Because grafting AI on top of existing workflows isn't going do a whole lot. In fact, it may harden existing barriers and widen existing gaps. We have to reimagine the workflows and our systems if we want to make a real impact.
The second component was leadership. I wanted them to think about what they need to do as a young associate who might be leading the way on new technology. How should they socialize it? How should they organize their efforts? How should they think about policies and procedures? Who should they involve? What questions should they be asking vendors? I brought in law firm leaders and in-house attorneys who are doing that work already to help guide the students.
The third component was to roll up our sleeves and actually do some building. The semester culminated in a legal design sprint where they identified an access to justice problem and used AI to build a solution for it. They pushed boundaries and put together some really cool apps in the end. I was so proud of them.
Do you have any advice or tips for this next generation of judges and attorneys who are going to be dealing with generative AI in their daily practice?
To young attorneys, I would say that as you're thinking about your career and where you want to be, go to a place that will invest in your training and is thinking about that training with an eye towards the future.
We can leverage AI tools to actively train legal professionals rather than just automating their tasks. Instead of using a platform that simply says, “Here are the issues,” we should utilize iterative systems that identify problems and pose guiding questions. A platform like that helps sharpen a young legal mind as they work through a case, effectively incorporating technology directly into the professional development process.
I also think there are going to be organizations that are more forward-thinking about so-called soft skills, and I hate to call them that because they're not soft skills anymore, they're practical and necessary skills. But it’s important to discern which firms are thinking about those skills and telling lawyers, "We're going to train you for the age of AI, not just in spite of AI."
To the judges who are thinking about AI use, I would say it's really important to experiment. I don't think we should let fear paralyze us. When new technology comes around, there's often a lot of fear around it. Of course, some of it is well-founded, and I'm not suggesting that there aren't things to worry about when it comes to AI.
But I analogize it to walking across a ledge. The first thing you tell someone who's walking across a ledge is “don't look down.” It's not because you're naive about the fact that they could fall or hurt themselves, it’s because looking down while they’re walking across a ledge is more dangerous. It can be paralyzing, disorienting, and ultimately not very helpful.
Instead of staring down at the peril, we have to look ahead. Yes, it's a difficult walk, but we have to get across.
Get the right training, get the right education, and join a group like JAIC where we're all talking about how to do this responsibly and trying to figure out the best way to get across, together.
