GenAI and Prompt Engineering at Legal Week

Hi. I'm Andrea D’Ambra. I am a partner at Norton Rose Fulbright in the New York office.

I am Norton Rose Fulbright's US head of technology and US head of eDiscovery and information governance.

I am not the head of IT to be clear. And, basically, what I do is cyber security did privacy, eDiscovery, and information governance in varying amounts on any particular day.

Yeah. Yeah.

Oh, I think it must be legal week.

So this is thirteen for me. I started in 2011. And back then it was called legal tech.

I still call it that if I'm being honest. Yeah.

Yeah. Yeah.

Mhmm. Mhmm. It's much broader than technology now. Right? Yeah.

This year, what I've been really excited about and really wanted to start talking to a lot of the technology providers out there is about AI because everybody's talking about it. But I'm very interested to hear. I've heard a lot about how they're leveraging it in the eDiscovery space, but I haven't actually, I mean, seen it. And by that, I mean, like, generative AI. Right? We've been using predictive AI for years.

TAR models, love that stuff. Norton has been using that for ten years or so, maybe more, like fourteen.

But generative AI is really interesting to me and I'm really interested to see how people are are using it to make the eDiscovery processes much easier, cheaper, faster, better. Right?

I mean I do.

We I feel like we always have been. Florinda Baldridge and David Kessler have been sort of driving that train from the very beginning.

We've always been able to leverage it to make things work faster, better.

Even when people at the time were sort of like, oh, I don't know about this.

We've been fairly successful in convincing our clients that it's the best way to, you know, get through whatever it is we had to get through.

Yes. I'm a huge Midjourney fan and I'm also a huge, ChatGPT fan.

ChatGPT is my friend.

And it's been really interesting to see both of those evolve over the last year. I started using Midjourney probably September twenty twenty two, and it's a very different tool now than it was back then. And ChatGPT, I guess, I started using Ruby in March of last year, maybe a little bit earlier than that.

Yeah. So Midjourney is a generative AI tool, but it's for images. So you will write a prompt, and it will come up with an image, and then you can tweak your prompt.

The actual drafting of the prompt is an art form all of its own. Right? And it's really interesting to see people who are really, really good at it. I am not really, really good at it, but I'm enthusiastic.

So it's just like how I am in regular art. Not good, but enthusiastic.

So to see it develop over the last, I guess, almost year and a half now has been fascinating.

And so I have some experience.

I don't have any skill.

Yeah. I haven't I haven't done any images for my clients yet.

Oh, I wish. No. Just, you know, personally interested unbelievable and interesting and beautiful and trying to get what you have in your head to come out on the screen.

Yep.

So, yes. I think that we all understand that there is great value in being able to, leverage generative AI.

I think law firms like, it's gonna come as a surprise to no one here, that we tend to be more conservative in how we're using it and we're always very worried about, you know, client confidential information and not, you know, creating something that's gonna create risk for the firm, either from a privilege standpoint or from IP standpoint.

We are using it.

I sort of personally, when I use it, I usually use it for things that aren't client related.

But, you know, if I have to come up with a panel description for some, you know, panel I'm gonna be on ChatGPT.

ChatGPT. I put it in there.

I look at it.

I rewrite.

It's like the first draft.

Right? I'm gonna tweak it. Eventually, I'm gonna have it write my bio.

Love it.

Or rewrite my bio because I'm like, oh, I bet I could do way better than I can. But, you know, trying to use it for things like that. But it's interesting. I went to William and Mary Law School, and our law school dean, Benjamin Spencer, amazing, and one of the things he said really early on was that they were not gonna discourage the use of check GPT and other generative AI because the people who can leverage that stuff are going to be the leaders of the future. Right? They're gonna be the ones who are gonna be out there sort of in the forefront.

Oh, absolutely embrace. I think hiding it is probably the worst idea ever from any number of standpoints. So, yeah, I think law firms have to embrace it. They have to understand it, and they have to figure out how to use it in ways that can be validated.

That was one of the problems I saw with TAR originally, right, was that everybody wanted to use it, but so many people didn't understand it and they didn't know how to validate the results, Right? So they were just like, I'm gonna put it in this black box and it's gonna be fine. You can't do that. Right?

You have to have ways to validate it. So that's what I'm really interested to see out there. See, you know, how they're gonna do the validation of, these new tools and, we'll see how it goes.

I have a few demos set up. I'm definitely walking the exhibit halls. Mhmm.

Definitely gonna try and get a few more demos. But I really just wanna understand what's out there. I've heard, like, I've heard tell of things. Right? But it's gotten mixed reviews. I've heard one tool that I won't name, you know, somebody said, well, it's like c plus law student. Right?

I don't know.

First level.

Yeah. First level.

Sure.

That's the excitement. Right? We're just at the beginning of this. Right?

Absolutely.

Couple different ways. I mean, I we at Norwood's Fulbright in our, eDiscovery Information Governance group, We have monthly get togethers with the group, and we all the partners will come in and we'll do training and talk about different things, and we've started doing generative AI training. One of my partners, Steve Rosa, is absolutely brilliant at all this stuff. He was teaching us all these, you know, sort of the back end stuff, which I still do not completely understand about how it works and how it's easy to leverage it in ways that I didn't know about.

Like, you know, you tell it to write something. Right? And you say, you know, write, a paragraph about x. And then instead of being like rewriting the prompt, you can just say, come on, ChatGPT.

Make it not so wordy. And it literally will change it and condense it. Right? And you're like, I can just talk to it like a normal human being.

So I'm, you know, gonna definitely go back and, continue to sort of share that knowledge there. Obviously, I'd love to speak, at different conferences and being able to be here and speak intelligently about how these tools actually work as opposed to sort of the, you know, theoretical, I think is gonna be very helpful.

Particularly because when you're trying to assess risk, it's important to understand the actual as opposed to the theoretical.

Well, thank you.