BDO's Legal Tech TalkHarnessing AI for Smarter Legal Workflows
About this episode
In the last live recording from RelativityFest 2025, Daniel Gold and Eric Derk talk with Angelic Benavidez and Joanne Liu of Hyundai Motor America's class action and IP group. They describe how they use Copilot day to day to summarize and draft, how company training helped them adopt it, how closed-system tools keep product liability documents protected, and why human critical thinking still matters when using Relativity aiR and other AI review tools.
Key takeaways
- AI handles first drafts and summaries so the team can take on more work.
- Company training on Copilot made adoption easier and set clear expectations about hallucinations.
- Closed systems let teams ask questions of their own documents without exposing them outside the company.
- AI-assisted review still depends on internal knowledge, like company lexicons, that only people bring.
- Hallucinations still happen. Pair artificial intelligence with human judgment and verify the output.
Full transcript
Transcript from BDO's published episode, edited for transcription errors. Speaker labels added by Gold Standard Leadership.
Welcome to BDO's Legal Tech Talk podcast. We're joined by judges and legal professionals to discuss emerging trends, regulatory updates, and the latest headlines. We'll provide tips to help your law firms and legal departments make the most out of legal technology.
I'm Daniel Gold, BDO principal and eDiscovery managed services leader.
And I'm Eric Derk, BDO's managing director of legal operations. Let's get started with this episode's exciting discussion.
And welcome back to another episode of BDO's Legal Tech Talk podcast. And, Eric, this is our final podcast recording live here at RelativityFest twenty twenty five here in Chicago. I'm very excited because joining us is not just one, but we have two wonderful people here joining us from Hyundai Motor America. We have Angelic Benavidez and also Joanne Liu. Thank you so much for joining us today.
So first off, so both of you are senior litigation paralegals within the legal department of Hyundai. It's fantastic that we get your perspectives. And so why don't you just take a moment for both of you to tell us a little bit about yourselves?
I'm Angelic Benavidez. I work for Hyundai Motor America. I run the eDiscovery legal holds Relativity database there. I work in the class action IP group, so I'm filing a lot of class action documents, and I also help out PL sometimes. I come from Big Law. I was in employment litigation for a really long time before I went in house, and I love it.
This is Joanne Liu. I am a paralegal eDiscovery manager at Hyundai Motor America and I am part of the class action IP group as well. I also do help with product liability and in my previous life I was also at BakerLaw, at Bird Marella and at Sidley Austin.
I gotta ask you this because as a legal ops person and working with Paris most of my life, we often are, like, the undervalued, unsung heroes. A lot of times we get asked to be doing stuff on a Friday night at seven o'clock and we just do it. Right?
Are you thinking that AI is gonna help you do things more effectively, more efficiently, or are you still kind of in a process of trying it out? Or have you kind of proven out certain and I'm not talking discovery. I'm talking about, like, the operational routines, like building those an estimate, maybe summarizing a note or drafting a memo. So how are you thinking about using AI or how have you already kind of enabled it?
I feel like we're not using it to our fullest ability yet just because it's new. But right now, I'm just trying to, you know, transition us into just getting used to it in our everyday lives, trying to minimize our workload. I see it being used in the future for a case review of documents. I'm pulling terabytes of data, and I want to use it to where we can minimize the amount of documents I'm sending, you know, that has to be for review. That's hopefully my main purpose that I wanna use it for.
Every day, you know, right now, I'm using it to summarize documents, to help me draft documents. I used it to help draft the RFP when we were going to RelativityOne, so I find it very effective.
One of the things that we continue to hear, Eric and I, is that utilizing Gen AI on the one hand is really great, but in the other, it's like, well, you know, I'm not gonna be able to do anything with my time anymore because I'm gonna get all my work done so much faster.
I'll go over to you, Joanne. Is that your experience or you feel like, actually, there's so much more I can do and be more productive as a result of it?
I definitely think that because we always run like a small lean machine team that's with AI, it definitely will help speed up the process so we can take on more.
I think that's the bright side of where people who are embracing and adapting are thinking. I'm not like Dan said. I'm not gonna have, you know, as much to do. I think that what these tools are gonna be able to do for us is act as, like, meta, uber assistance for us.
So first draft, first level of research is gonna bring more data points into our analysis. It's gonna make us potentially smarter, but it's also gonna force the idea of critical thinking. Like, we can't just take it for verbatim. We have to think through it and review it because we're gonna give it to an internal customer like a lawyer.
So, like, tell me a little bit about if you've experienced any, like, I can just summarize this email.
Or have you found that when any situations where you thought it was gonna be really super easy, but it turns out to be actually more complicated because you're getting a more refined dataset?
I would like to point out Hyundai did a really good job of training us with when they adopted Copilot. And I remember the IT guy said, it's here to enhance our jobs, not take away our jobs. So what are we gonna do? How are we gonna adopt it into our infrastructure?
Right. We actually did already have training for Copilot. Our searches are just our own. And therefore, you know, it's also a learning process where they're training us how to create better prompts. They give us, like, examples of templates and, you know, try this type of prompt, see what results you get. And now we're really doing, like, a stress test on known issues, known results.
That's amazing. So first of all, that's fantastic to hear. It delight to me that a company has gone and taken that step to provide the kind of structure and the training around it. I don't hear that all of the time, So I think that's really amazing. So both of you have adopted it, but what about the attorneys? Have you found that the attorneys are as willing as the both of you to adopt it?
Most definitely, but, of course, you know, the hallucinations still exist. And sometimes we do have, like, an argument referencing. Read this paragraph again. You know, try to re prompt it, and, you know, it's still a learning program.
The argument you're having is with Copilot. Yeah. Yeah. That's amazing.
I mean, we you know, you're laughing. Right? But but, like, it's actually fantastic. So, I mean, really, you're sort of treating it as if it's, you know, maybe a member of your team Is the way it sounds.
Angelic, I wanna go back and ask you about something that you said. Because I think a lot of organizations are still going through this threshold crossing about don't use it, don't use it, don't use it. And then all of a sudden, something turns and it's something might be the people are using it anyway. And your company, do you remember what the primary reason for shift was?
I don't remember what the shift was. I remember maybe it was feedback from that training.
And I will say that Hyundai is very open to using new stuff and using Microsoft's new platforms. And if I say, I think this is good, I think we should use it, they listen, and I'm really glad that they bring it on.
Legal teams are struggling to manage resources, maintain compliance, and adopt new technologies.
BDO's legal operations team can help. We enhance efficiency, reduce costs, and support compliance so you can run your legal department like a business. Schedule a meeting with us today at bdo.com.
Are you finding that you're gonna utilize Copilot for other things? Like, there's notebooks within Copilot, there's custom instructions in each of these notebooks and reference files. Do you think that you'll be organizing it and classifying data or projects that you have within legal department based on specific things that you're working on? Maybe you're working on litigation or there's contracts with vendors or anything like that. Do you foresee use cases like that?
Working in, like, the product liability department as well, I imagine myself using, like, Notebook just to organize, like, OSIs, you know, just, like, similar incidents, you know, identify vehicles, what, you know, the claims are, and then just trying to find the similarity and the connection and try to just summarize that way. Have like repeat type of documents to come up.
Perfect use case. Right? It's called a RAG. Well, it's a type of a RAG, retrieval augmented generation.
So you basically put a bunch of documents into a notebook in this case and say, tell me the commonalities. And so people are doing that in the PSY context but also lots of different types of contracts. You know, I will oftentimes use like the Windows word compare stuff And now it's like, well, I just got a document from somebody. It's different than my document, but rather than do the side by side redlining, I'll just have AI tell me the differences.
Yeah. And just to bounce off with that, it's because it is in a closed system.
I can actually try to utilize the translation, and I'm not gonna be able to put it in, like, a Google Translate now. That's, like, another benefit of having, like, a closed system in Copilot.
Either it's, you know, Google or Amazon Translate for that very reason. And so now what you could do is, it almost seems like you can downsize maybe some other tools, other applications because you are leveraging Copilot in a way that does contract analysis. And maybe it's some eDiscovery work. Maybe it's an internal investigation. You're throwing all the documents into Copilot.
So it seems like you've got now, like, a much bigger team, but it's all just built within Copilot. You started touching on this before. So there's these studies that are out about how people are utilizing Gen AI and the fact that, well, if you utilize Gen AI too much, then you start losing critical thinking skills. Right?
Eric asked, you know, when have I used it? I'm trying my hardest to okay. This is what I have to do. I'm going to use AI for this scenario. I'm not, you know, too I don't know this topic well enough, so maybe I can use AI, get some facts, write it myself. So it's a learning curve, definitely.
That's amazing. And so now you guys have just adopted RelativityOne not too long ago, And so now you're gonna start seeing, I'm assuming, in addition, obviously, utilizing Copilot, utilizing some of the RelativityOne air functionality as well. Is that something that you guys are looking forward to?
One hundred percent. We just reached out to our Relativity representative to train us and, you know, so we can start using it.
What are your thoughts about how Relativity has created this thing called aiR Assist, and you're able to basically have a copilot conversation, but with your documents inside of Relativity? What are your thoughts around that?
I think law firms are gonna push back, honestly. That was the whole reason why I wanted to switch to RelativityOne. I was like, if Hyundai doesn't use this chance to get AI in our eDiscovery, we're gonna lose it, and we're gonna be behind.
However, you know, despite the air and, like, the AI, there are documents that only a human can detect. Because after discovery, just reading one off, you know, an email that has, like, potentially a code word. And all of a sudden, because it's not part of your keyword, your agreed list of keyword searches, but there was a human factor that recognized the pattern and now you have your hot dogs Completely different from your original set. And that is not something you can train AI to do because the input is what the human knows and the human won't know until they go in and dig in as well. So by limiting that, sometimes, you know, we still need that human aspect, the human eye, even though we have, like, the results of, you know, the search terms.
I love that. It is literally applying the human intelligence with the artificial intelligence.
Correct. And that I think is a hundred percent where my biggest concern with AI is it's not the application of AI. It's not being fearful of it. It's letting go of the critical thinking.
Yeah. Exactly. And I mean, I think that when you talk about internal lexicons, it's not something that you can just push out to an algorithm. You have to be able to discern what that code word might look like in Enron. That was one of the big breakthrough, you know, we'd heard about the word raptor. Well, we didn't really go in there looking for raptor, but we eventually uncovered it and Atenix helped us do that. But the idea is that without a critical thinking process and pure reliance upon a machine, you're gonna miss stuff.
Well, can you imagine if we had Gen AI in the Enron days, how much faster that investor would have been. Yeah. Yeah. Angelic, you were actually on a panel yesterday. Tell us a little bit about the panel and maybe what was the big key takeaway that you wanted people to learn?
Yeah. I was on a panel yesterday, and it was discussing the transition from Relativity server to RelativityOne. I mean, my big takeaway is I didn't know that I thought I was the only one that was having these issues, you know, making the transition. I didn't know that so many companies were trying to also make this switch and needed help making the switch, and I kinda feel like Hyundai was a little guinea pig in that sort of way.
And it was hard. It was difficult, but I gave him the pointers. I told him the benefits. And, again, don't be scared of it.
Mention AI, and they will say whoever you report to is like, okay. Yes. Because it's all about, you know, it's all about AI right now and everybody wants it in the pile.
Angelic, Joanne, thank you so much for joining us here today. It was wonderful having you here.
Thank you.
Thank you.
Thanks for joining us on BDO's Legal Tech Talk podcast.
If you're enjoying our podcast, don't forget to rate, review, and subscribe for more episodes.
On www.bdo.com, you'll find all of our episodes, transcripts, and ways to continue the conversation.
Until next time, this has been another episode of BDO's Legal Tech Talk.
Daniel Gold is an attorney and a Principal at BDO USA, where he leads eDiscovery managed services. He hosts BDO's Legal Tech Talk, co-hosts the NeLI Pod and founded Gold Standard Leadership.
More about DanielGet the next one in your inbox
Daniel Gold writes weekly on leadership, legal technology and eDiscovery.
Subscribe on Substack