Colin Rooney 

Hello everybody, welcome. I’m Colin Rooney, partner and head of the Technology and Innovation Group in Arthur Cox. I’m delighted to be here today to talk to you with Michael Egan, who’s a senior director in our Legal Technology Group within the firm. We’re going to talk to you about the various developments in relation to legal technology over the last number of months in particular, with a particular focus on what’s happening in relation to generative AI. So welcome, Michael. We might start by talking about some of the things that are happening in the legal technology space. A lot of developments, speed of development, very fast, increasing products and services available to us. Gen AI obviously playing a key role in relation to those products and services. So what are you seeing on the ground? What are you seeing emerging? What do you think is going to be relevant to our practise and indeed that of our clients over the next number of months?  

Michael Egan 

We’re seeing a lot. It’s cropping up really everywhere in every product that we have. For today, we’ve cherry-picked some of the ones that probably seen some of the successes. And look, one of the big ones is document review. So that would pop up, up and down the firm, really. I mean, cyber incident response, DSARs in our own, in our own group, a huge amount of DSARs and, you know, discoveries and investigations. So the real benefit to it is, and we’ve seen this over the past kind of 7 or 8 months, is the speed. It’s the speed at which you can get the, you know, the critical documents and find them. So to put that into context, we had a pretty urgent matter with a tight enough deadline, which is a great one for Gen AI because that’s the sweet spot for it is when your back is to the wall and you don’t have time to just go through everything and do the keywords and go back and forward. You need answers quickly. So we use the Gen AI product. So 12,000 documents, to put in context, would take four people about two weeks to get through, and that’d be just solely reviewing documents.  

Michael Egan 

We built the prompt. I know there’s absolutely upfront effort in the prompt. I mean, that’s got all the legal people, all the feeding into it. Absolutely critical. The legal tech team are in it. So there is a bit of a lift in that at the beginning. But equally, if you’re mobilising a tech team or a review team tomorrow, you’d have to brief them and review them. So I’m, you know, I think maybe just slightly a bit more in the prompting, but I think it balances out. But once we did that and once we ran the sampling, got the prompt spot on, it took about three hours to do the full document set, which is absolutely phenomenal versus, you know, as I say, two weeks with the team. So I mean from a speed perspective, really, really useful and last week as well, we got a pretty urgent one with a small deadline. We got 1,700 documents, and again overnight, we were able to put them in and surface the key documents for a deadline. So, no problems on that front. The big piece is the accuracy. That’s where the legal judgement comes in, the defensibility comes in.  

Michael Egan 

The legal profession still has to stand over the outputs and validate it, make sure it’s right. And that’s where we’ve been putting a lot of focus because you know, a lot of people can just turn on the technology, but it’s embedding it into the processes where we have to be. That’s our responsibility. But once we ran, we did run samples with, or we put a very junior person on it and we ran the Gen AI model and we had a very senior Of Counsel person go through it. And in that particular instance, the accuracy was higher for the Gen AI. And that won’t be in every case. We did find instances where it missed things and it got caught in our QC process. So that’s working. So yeah, I think that’s where we’re seeing the biggest impacts at the moment. Really positive stuff from what we’ve seen so far.  

Colin Rooney 

So we’ll tie this down in a moment or two to some of the work that we do between our teams in relation to, for for example, sort of, you know, maybe compliance work around data protection, GDPR and the like. But before we get to that point, I mean, this is, as I mentioned, kind of evolving very quickly. There’s a lot of new releases. What kind of things do you think we can expect? I’m not going to ask you to say in the next five years, but maybe over the next, say two to three, maybe leading into six months, like what’s on the horizon that we should be thinking about that’s going to play in in terms of the work we do that are things that are enhancements, if you like, to the tools that already exist?  

Michael Egan 

So yeah, I mean as I said, there’s a huge amount happening and I think the way to look at it is the next two or three months, the big things that are coming really is anything that’s going to help with early insight. So when we’re typically called into action, there’s just a huge amount of documents. There’s lots of emails, teams messages. Our whole forensic team goes off. They do that.  

Colin Rooney 

Huge volume of material. Yeah, of course.  

Michael Egan 

Yeah, massive so I mean the challenge is it can take a week or two for our legal teams to go back and forward where our clients understand what’s in the data set, do interviews, and so on, right? Where the Gen AI is going to be inserted in here is for early insight. So, it’s going to be almost like a chat experience. You’ll put in the data into our secure environment here in the firm. Our tech team will set all that up, and then our lawyers are going to be able to go in and ask questions a bit like you would on, you know, any of the LLM models that are out there at the moment. So you could say, who are the key people, what emails are you seeing, what dates? And that’s going to feed back information to our team. So if we have millions of documents and we’re going off and we’re trying to establish keywords kind of blindly a little bit at the moment, like they are informed, but we’ll be able to enhance those and make them far better to be able to target the data and find. So we’ll be able to go back to our clients, get that targeting much better.  

Michael Egan 

So, we’ve played around with some of the early models, excited about that. There is a limitation I have found with the platforms at the moment. They can only take maybe 10,000 or 20,000 documents to do it. So, you know, we get terabytes in here. So we could have 200,000 or 300,000 documents, but we’ve managed to find workflows within our other tech stack to do it.  

Colin Rooney 

Of course.  

Michael Egan  

You know, I’m really looking forward to when you can cast that over the full net and that’s going to be really, really powerful.  

Colin Rooney 

One of the most exciting things that we’ve done over the last while is the collaboration that we’ve had in terms of looking at things that we’ve been doing, work that we do on a very frequent basis, like for example, data protection impact assessments or data subject access requests. And I think trying to see how we apply the technology to those processes, which are often document intensive, they’re time intensive, they can be large volumes of material, they can be complex in the sense that, you know, they’re not necessarily litigation, but they can look a little bit like mini discoveries. Or in the case of the impact assessment, it can look very much like a detailed analysis that needs to be undertaken with lots of different inputs. And I take your point entirely in terms of there are some limitations at the moment. Like we talk a lot about redaction sometimes, the challenges around redaction and the technology is not, you can see it’s moving in that direction, but it’s not quite there yet. But I mean, one of the projects that we have had, I think a lot of success with so far is the data protection impact assessment, which just for those who are not aware, is a requirement in certain circumstances where you’re rolling out a new solution or a new service that you need to do an analysis in terms of the impact on an individual’s privacy, and the use of their personal data and how that’s managed.  

And we’ve used some of the intellectual property that we have within the firm, the expertise that we’ve built up over a long period of time, mixed in with the expertise that yourself, Michael, and your team bring, to work up a reasonably compelling solution in relation to this. Do you want to elaborate a little on that?  

Michael Egan 

Yeah, I mean, we’ve been working for quite some time on this one. And as you said, look, Gen AI is a team sport and what I mean by that is you need your knowledge lawyers who you know, they are constantly tracking legislation. You need, obviously, our lawyers who are using that. They have the commercial acumen, they have everything that, you know, and they’re very close to the client on it. You need the technical team, which, you know, we have in our practise group as well, really good tech team to design the workflow. And then you need the product itself. So you need to bring all of that together. So that’s what we’ve been doing and that’s easier said than done. And yes, the DPIAs we’ve a solution at the moment that captures all of those skill sets working together and it’s going to be bespoke to our clients. And look, where, as you said, DPIAs arise is when there’s a high risk to people’s personal data. You can keep me honest on this, and it’s often brought out when there’s a new tech product being brought out, and there might be tracking or profiling or sensitive data, or when some of our clients putting in CCTV they have to do it. So yeah, these are a statutory requirement so they have to be done. Our conversations with our DPOs and our clients are telling us that they do have a lot of GDPR requirements. And this is one of the ones that they’re saying is very time intensive because, I mean a DPIA, although the DPO signs off on it, I understand the goal that has to go around different parts of the business, there’s policy documents, there’s a lot goes into it. And people start with a blank page every time, and it just takes a lot of time. I think the other thing we’re hearing is it can it’s quite expensive, like it can be expensive to get someone to do it from scratch or even asking a law firm to do it from scratch. So we’ve taken all of that on board and we’ve looked at the whole process and we’ve looked at the the documents that we would typically use, like the DPC guidance.  

Michael Egan 

We’ve looked at the legislation work really closely in-house to pull all that together and all of our IP, our precedent documents as well and that was a really big lift, and the only credited team for that. And on top of that, then we designed it so it can take in our client’s documents, their policy documents. Like there’s various things that you can put in there, and that’s case specific, I would say. And we have it set up now that we’re able to if they’re running a DPIA, they’re able to give us the structured information that we need. We’ve designed a very structured form to capture what we need, quality inputs, you know, for quality outputs. And we’re putting it in and the results are phenomenal. I mean just the time savings like that could take days. I mean internally, if a client is to do it, it could take weeks because you have to wait to go around the houses and people are busy. But once we now put it into our Gen AI platform, we’re able to do the DPIA within minutes and then it’s into our, obviously, the human is in the loop still, our lawyers then are able to take it from there. But what we’re seeing, and I understand the feedback from the associates, is that they only need to make minor refinements to get it to where it needs to be. And so yeah, it’s a really, really exciting product.  

Colin Rooney 

Yeah, it’s been a very interesting process. I mean, I think it talks to a couple of points that we touched on at the beginning of the discussion. Certainly the value in terms of, working to establish a process, following that process, making sure you work with individuals who can leverage the technology, know, have experience and are comfortable with the technology. At the same time, making sure that you have very good inputs, which I think is the benefit of the expertise that we have within our team. And of course, that oversight as well. That’s the extremely important role for associates and lawyers to play in relation to the material, that’s for the output for the client. But what we have, I think, is a solution which looks at the process, has very good inputs and produces a very strong, robust output, factoring in, you know, other DPIAs, but also factoring in guidance and case law and other material which is relevant in terms of that process. And I think the thing that’s certainly struck me in terms of this collaboration, and I know you’ve been doing with other senior lawyers in our firm, working closely with Rob Corbet, for example, one of the things that I’ve noticed is that this will lend itself very neatly, I think, to things like data breach processes, where it can be a very similar process time and time again. The subject matter can be different, but the nature of what we’re doing and how we’re doing can be, to a certain extent, similar across matters. And I think this is the sort of expertise which we can really leverage for the benefit of our clients.  

Michael Egan 

And to add, the real beauty of it is it’s scalable. You know, this is all the key messages here. This is a repeatable process. People typically do it from scratch.  

Colin Rooney 

Yeah.  

Michael Egan 

And this is very scalable and it can be customised. And once you do the upfront lift, you can tap into that every time. And I mean, the other big thing as well is it’s going to be privileged because as we say, the end product goes through our legal team.  

Colin Rooney 

Of course. Really important. It’s a vital consideration, the privilege piece. 

Michael Egan 

And we need to capture all the core fundamentals by bringing in the new. So yeah, I mean, and we’re moving on now to things like LAs and other areas, like that’s in the not so distant future that’ll be released too. So yeah, we’ve other products in the mix.  

Colin Rooney 

So we’ve talked Michael a lot about sort of the importance of the accuracy and the importance of, consistency and the various different inputs. Just maybe want to close out by saying these are really powerful tools. They’re going to be transformative, but there are some challenges that we should be aware of and that I think the client should be cognisant of, particularly in the early days of adoption.  

Michael Egan 

Yeah, look the big challenges really are embedding into the process. I am envious of the legal tech vendors because you know, they come out with the products and, you know, there’s a lot of hype around it but ultimately we have to embed it into the processes for our clients. The regulations are catching up, the regulators are catching up, the other side are catching up, and there are rules of engagement you need to abide by. And ultimately, it’s our lawyers and our tech team that are going to bring this to life, but they have to be able to do it so you can stand over it. So I think for me that is the biggest challenge with Gen AI but, you know, if you approach it and you take your time and you’re sensible about it, you can use it.  

Colin Rooney 

Thanks, Michael. I think that’s a great point and a good place for us to pause for now. So thank you very much for taking the time to listen to us. I hope you found that to be informative. And if you’d like to get in touch, please contact us. You have our contact details. We’re also available at arthurcox.com/technology.