Google, Weil + Gemini Enterprise for Legal

Leading US law firm, Weil, contributed to the development of Google’s new Gemini Enterprise for Legal. Artificial Lawyer spoke to Andrew Simon, Weil’s Chief AI & Innovation Officer, about the firm’s work with Google and especially the new legal offering. We also look at BenchMark, an AI application Weil built internally, and where Google capabilities were used.

Of particular interest is the fact that Simon sees the wider Google ecosystem as a place where the firm can work, deploy a range of LLMs and legal tech tools, and where his team can build new applications – while still ultimately producing a lawyer’s final work product in Word. This goes directly to the centrality strategy idea that AL explored yesterday, see here.

Also of importance is that Weil is not just using Gemini Enterprise for Legal, they actually helped with the development of some of the applications now on offer. Plus, as Simon noted, there will be more applications to come. This is rather like Claude for Legal, which started with a small number of skills, then expanded to providing dozens of them.

Plus, we also look at BenchMark, a tool that Weil built internally – in part by tapping what Google could offer – which helps with judge analytics.

To watch the video inside the page please press Play, or you can go direct to the AL TV Channel.

–

— (Advertisement) —

—

Some of the key points:

• Weil helped shape Gemini Enterprise for Legal
• Weil fed Google two early use cases: parallel research agents and NDA drafting.
• The Google offering is about more than models. Simon framed it as a full stack: models, interfaces, agents, plugins, and supporting infrastructure.
• Weil is not currently part of the Google Workspace ecosystem, but they’re testing it.
• The key idea: lawyers can still work in Microsoft Word while the underlying AI work happens elsewhere.
• Google acts as an integration layer across tools.
• Weil sees value in having one or two main surfaces while still accessing tools like Harvey, Legora, Thomson Reuters, and others through the Google stack.
• The goal is to avoid forcing lawyers to jump between ’45 different places’.
• Weil is building its own internal products on Google infrastructure
• They’ve built two new products on top of the Gemini Enterprise Suite and Google Cloud.
• One example is BenchMark, a system for analyzing judges’ records and helping litigators prepare.
• Google enables a multi-model strategy. Weil is not locked into only Google models (when working not just in Gemini Enterprise). They can use Google models alongside OpenAI, Claude, local models, and others.
• Google’s value is the ability to run all of that on one infrastructure layer.
• The real strategic value is centrality and scale.
• Weil likes Google because it offers the ‘whole vertical slice’: data storage, retrieval, processing, model deployment, and iteration. That makes it easier to build mini-systems, experiment quickly, and scale what clients actually want.

Why it works: the interview’s core message is that Google is becoming not just a model provider, but also reduces friction and centralizes work.

—

Legal Innovators New York Conference – Nov 17 and 18 – Where Legal AI meets the Business of Law – come and join us in New York this November. More info here.

—

AI Transcript

Hey everybody, Richard Tromans here again, Artificial Lawyer TV, doing a special interview following the news this week that Google has launched a dedicated offering for Legal in the shape of Gemini Enterprise for Legal, which is part of a much broader set of projects which they’re working on. Weil is one of the law firms, one of only three or four of the firms they mentioned in their press release this week that has pretty good knowledge of it. So I gave them a call. Andrew, what’s your title?

Andrew Simon (00:50.23)

Chief AI and Innovation Officer at Weil Gotshal & Manges

Richard Tromans (00:56.6)
Fantastic. Andrew very kindly offered to give us a bit of an insight into what’s going on. Plus, it turns out they’ve also been developing some other interesting things, which we’ll come to momentarily. But let’s start with the news of the day. What is Gemini Enterprise for Legal, as far as you’re concerned, and what is your connection to it?

Andrew Simon (01:17.894)
Great question. So, you know, first of all, of course, good morning, Richard. The Gemini Enterprise for Legal thing specifically is the way Google is going to market with its models and what I’ll call the supporting infrastructure for them. So it can be the models, it can be the way you interface with the models, whether through the web or through the anti-gravity at this point, two point zero or IDE application that sits on the desktop.

like many people are familiar with the desktop applications. And then making use of the specialized skills, agents, and plugins that Google has developed in partnership with its launch partners to do productive things with the Gemini suite of models.

Richard Tromans (02:04.279)
So on that point, so has Weil fed into the development of Gemini Enterprise for legal?

Andrew Simon (02:11.907)
We have, and we fed in two things to start. one was and I don’t know if they’ve released this in the initial release. One was about parallel research agents. So when lawyers or any professional really wants to do parallel research lanes on a similar topic, agentically, you know, taking the deep research capabilities many have become familiar with and amplifying them a little. So something to do that out of the box.

And then the second is NDA drafting, which of course NDAs are a very common piece of business, but what’s really exciting under that agentically is drafting documents from simple to over time increasingly complex in an agentic way.

Richard Tromans (02:57.005)
Hmm. Interesting, interesting. So, and and this is another question I think that a lot of people will be asking you is is that a lot of people assume that the vast majority of law firms, certainly the large commercial ones, live in a world of Microsoft Word. And that even though they may be using Google search and maybe they get a few Google Docs sent to them by clients every now and then, they’re not really inside the kind of Google ecosystem. But is Weil part of that ecosystem and has it been before?

Andrew Simon (03:26.883)
It’s a good question. So while is not currently a part of the Google workspace ecosystem, though it is something we’re beginning to test. What’s what AI, what this moment is allowing us to do though, is to decide. Is Microsoft Word the surface? Possibly, very likely, right? Everybody’s on it. But it doesn’t mean that needs to be where all of say the work crunching happens. You know, if you’re interfacing with

A large language model or you know a smaller model or any various types of models to do something productive, the model may or may not need to make use of Microsoft Word, even if you, as the human doing the review, stays in Microsoft Word. And that gives us flexibility to design little systems that work across both, right? So you can imagine a world where something in the Google Cloud Stack is actually putting something together in Google Docs, the lawyer gets a file.

And then simply opens it in Microsoft Word with a lightweight plug-in on the side to interact. So there there’s flexibility there for which pieces of the technology show up when in the Word.

Richard Tromans (04:38.753)
Gotcha. And you know, as people know who who’ve read about this already, and there’s a big article on Artificial Lawyer this week about it. I mean, there’s a bunch of skills within what they’re calling a plug-in, but it’s an enormous plug-in. But one of the key facets is the legal tech partners, Harvey, Legora, Thomson, Reuters, you name it, enormous group. I mean you’ve got consultants, but let’s not get into the consultants, but some people who, you know, who looking at this for the first time might say, Weil, hold on a minute, why do we need Gemini Enterprise for legal?

Why don’t we just bring in all these legal tech companies independently in parallel as we need them? Why do we need this interface?

Andrew Simon (05:16.389)
I think they all play different roles at different parts of a strategy. So if I think about while strategy, we have three buckets. We have enablement, we have building custom applications, you know, to take what, you know, the off-the-shelf tools that we’re all familiar with, you know, take them to the next level, and then we have building advanced capabilities. So if we think about enablement, I think all of the legal tech companies have a very important place to play there.

While they also have a place to play in designing advanced capabilities. So I think buckets one in three for me. And what I mean by that, you know, if you think about Harvey and Lagora specifically, we have lawyers using both, working very well to come up the curve on what is possible in AI, doing it within an environment that is very structured and is helpful to keeping the work on track. Similarly, if we take advantage of

the ability to MCP into those serv some of those services, if we can land on a user interface that people like interacting with and keep reduce the surfaces people have to go to, I think right now there’s a big headache with, which tool I go to do I go to for what, right?

Richard Tromans (06:30.625)
Yeah. And and and s and so many contexts to to move between

Andrew Simon (06:34.685)
So many contexts, exactly. But if we can say, look, here’s the one or two places to go, but you can get the service in Legora that you love. You can get the service in Harvey or get to that vault or that agent you built in Harvey that works really well with, say, the new models Harvey’s developed. Same with Thomson Reuters. You know, they just released a Quen-based model that’s been post-trained, Thomson One, yep, and on on their data. And it’s like, okay, if you want to get to that model.

Richard Tromans (06:55.639)
Thomson one.

Andrew Simon (07:03.349)
Great, you can do that by MCP, but what you don’t have to do is log into forty five different places, which I think is a it’s not a realistic way to ask people to work.

Richard Tromans (07:14.207)
Yeah, so I mean it it’s it connects to something I wrote recently also today, but this week for people watching this tomorrow is the idea of centrality, that everybody, certainly all the larger players, want to be the home of work. This is this is everything comes into this place. This is where it happens. And the output, as you mentioned earlier, may drop out into a Word document. That’s its final destination. But you’re not necessarily working totally within a Word document the entire time. It’s this other layer.

Which I guess is what Google is doing. Let’s just let’s just move on slightly beyond that then. So you were saying that you’ve known Google for a while and that you’ve actually built internally some some new products yourself.

Andrew Simon (07:53.37)
That’s right. So we’ve built two new products that are the backbone of which is the the Gemini Enterprise Suite and you know Google Cloud more broadly. And then but it’s not to say other models aren’t in use either, right? One of the great things about partnering with Google is yes, they offer the Gemini models, but they also offer the infrastructure to do things with other models, whether it’s local. You know, you I don’t know if you can hear one of the

fans on one of my machines over here, I have local models running, right? Or whether it’s another Frontier lab who also have great things to offer. And so that’s the beautiful thing about for us part

Richard Tromans (08:32.503)
So so so just to clarify, so you so you don’t have to so with with the Gemini Enterprise suite or ecosystem, you do not have to just use Google models. You can use models from all over.

Andrew Simon (08:42.287)
Correct. As you move up, well, Gemini Enterprise for Legal specifically is Google, but as you go up and say, okay, I’m building on Google Cloud, deploying the front end of the Gemini Enterprise for Legal, I get all the models and the MCPs, but then I can also bring in other models, right? I can use Amazon. I can I can bring in OpenAI and I do both. I have all. Do it all. So we have all three, right?

Richard Tromans (08:58.061)
So you could bring in Claude, you could bring in cla I mean bizarre, you could bring in Claude for Legal Right. Interesting. Interesting. All right. Let’s let’s have a look at the products, which is called, I think one of them is called Benchmark, is that correct?

Andrew Simon (09:13.731)
Yes, absolutely. Benchmark, the the concept of which was, hey, we had a client call us and say, with AI, can’t we just, you know, understand the judge’s record? This particular judge had been on the bench for 15 years. What we’ll demonstrate today is an anonymous totally made up fictitious judge. But in in the instance we’re talking about, this came from how we have an important hearing coming up. How has the judge asked questions about these issues before?

How has this particular judge ruled before? And the idea was if you can go through every part of the record, published decisions, transcripts, anything that is showing up effectively in PACER, you can get a better sense for how the judge approaches an issue, how the judge questions on the issue, the types of arguments and evidence the judge favors versus doesn’t, and it just allows the litigators to prepare a little bit better.

Richard Tromans (10:09.237)
sure, got sure. can you show us a bit more? All right, here we go.

Andrew Simon (10:11.971)
Yep. So my colleague Dalton has logged in. And so when you come in, what you do is you request a benchmark profile, and you here is someone internal to Weil We have not made this part outward facing yet. There’s been questions about that, you know, and it’s something I we explore the market for. So you come in, you put in your judge, you state, you know, your matter. obviously we have the billing code.

And then you know where are you in the case, right? You see the scope and the procedural postures. So it can orient around well, which parts of the record are we curious about? You give some detail, and then you start uploading supporting files. Then you move into well-shaped questions. When you do the questions, you come to the page here and it says, look, here’s the record we are about to launch through the tool. And I’m happy to talk about how.

Tool works. I think I have some videos of me at a Databricks conference talking about parts, but not all of how this works. anyway, so once you’re clear on what you’re about to go through and you can adjust this, you move forward, you do the review and the launch, and then you get this pipeline, right? So obviously, this doesn’t happen in two seconds. The machine has to work its way through all of the data. Once you run the analysis,

You come to a place like this where it’s telling you what’s happening, right? It got the documents, formatting, blah, blah, blah.

doing its work. Dot dot dot dot dot. Obviously we’ve preset this up to move faster. A typical run usually takes on the order of hours.

Richard Tromans (11:55.522)
Yeah. But you you’re going through a lot of data obviously, so that’s fair enough.

Andrew Simon (11:58.926)
Yeah, I mean for a judge that had been on the bench for six months, not that much. For a judge that’s been on the bench for eighteen years, quite a bit, of course.

And then you come here and we get to outputs once it’s done. So what you might have noticed briefly on those is there’s a variety of steps, including some verification checks. And part of that is using different model families to check the work of other models. One of the things I think is important to understand is if you use the same model five times for the same thing, it can be a great way of catching errors, but it also tends to make the same error in the same way.

Right. So models with the same architecture tend to reason similarly, which is why it’s important when you’re doing something like this, and that’s what’s great about Gemini Enterprise for Legal, to be able to bring multiple models to bear with different architectures on them.

Richard Tromans (12:49.865)
sure.

Andrew Simon (12:51.045)
All right, so you come here and you get an output, and you see here, here’s an executive summary, right? The Honorable Jane Doe, 87 non-overlapping analysis sessions. So it’s giving you an overview of what it did. This is telling you it had to break everything up into 87 separate sessions, right, with the LLM in order to get through the data and then bring it back all together afterwards. And so you come down.

And it starts to then in the second and third bullets talking about what the judge does. All right. So here we’re talking about inferences of Cyenter and how this judge, the Honorable Jane Doe, fictitious course, thinks about it. It gives you citations to the record. So that way you can go in and you can look for yourself whether it’s citing a transcript or an actual decision and see what it said. As you go through, maybe let’s go down to section two, Dole.

Richard Tromans (13:49.323)
Yeah. Yep. And then w once we’ve just seen a little bit more, let’s we’ll we’ll go back to Google to round off.

Andrew Simon (13:55.334)
Perfect. And so here, you know, you’re getting into the actual substance itself, right? It’s talking in this case on the scienter mechanics, how the judge approaches that, how this judge dealt with it in a prior holding, and and so on and so forth.

Richard Tromans (14:13.229)
Fantastic, fantastic. Really and this was w how much was Google part of well the Google ecosystem part of you being able to make this? Was it connected or this this is a separate project?

Andrew Simon (14:25.647)
That was fantastic. So this was developed on all the models, but the you know, the the Gemini Flash series was very good for coding portions of it, right? So a great way to build some of these products quickly is you use these smaller, more focused, specialized models that are great at coding sections. You use a more robust model to organize all that than a human to manage. And Gemini was crucial.

To getting the product to where it is today by being able to agentically code different modules of it. And then on the legal work itself, using the different Gemini models at different points in that pipeline in conjunction with other Frontier models allows us to get to a result that we believe and I believe is dependable because I know it’s been the way it’s been iterated and how it’s been iterated in a multi model architecture.

Richard Tromans (15:21.133)
And this experience, this this to some degree kind of like helped you in that journey that you’re now going through with the Gemini Enterprise for Legal experience that is now you know sort of rolling out.

Andrew Simon (15:34.937)
Yeah. How’s the experience been?

Richard Tromans (15:38.038)
Well, yes, I think everyone would like to know that. I mean, I think I think one of the things I think people are probably just really curious about is just like how significant is it? I mean, it’s it’s difficult to know right now, and we’ve got a few skills that they’ve outlined, we’ve got the partners listed, we’ve got this sense that it’s got, you know, it’s a very useful ecosystem, you know, we were talking about earlier. But how how useful has it been and how useful will it be for you if if if Weil goes all in on it?

Andrew Simon (16:05.829)
So tremendously useful to me. I think it’s very significant. You know, one of the things I’ve talked about recently is once you have the infrastructure in place for these things, you can scale infinitely. The reason I I enjoy working with Google so much is their total ver they’re they’ve got the whole vertical slice, right? So if I think about developing and storing the data in in clever ways.

And then retrieving data, right? Precedence in clever ways, ways that are being experimented with right now. And then managing those processes so the system becomes self-learning. And then applying models to it, Gemini models, but also others. Google allows us to do all of that. And it allows us to build mini systems like Benchmark. And in we have something that’s public called Privacy Pro and some other things that we haven’t released yet that we’re developing.

It gives us the infrastructure to do it all in one place. And that means we can experiment and test and see what clients like, how they want to be served, just faster, better. And that’s why I think it’s so significant. It’s not just the models, and we’re not locked into one way of doing things. We can be flexible to meet our clients where our clients are going. We don’t have to do it the way a particular company sees it.

Richard Tromans (17:28.503)
Gotcha. And just last couple of points. I mean, people got very excited, me included, about Claude for Legal, because it was the first big LLM giant to really go in fully into legal. I mean, does this displace what Claude for Legal is doing? Or it it becomes like a kind of bigger brother of Claude for Legal? I how does how do how do these two fit together? And what about when OpenAI finally unveils what it’s going to be doing?

Andrew Simon (17:56.186)
Yep, OpenAI has Jason [Boehmig] and I I know Jason, he’s great. I think they will compete. They will compete. But

Richard Tromans (18:02.841)
Hm. Well well all all three or the two will compete to be part of Google in in a kind of relationship.

Andrew Simon (18:07.587)
No, all all all three are going to compete with each other, but ultimately these models have to run on some infrastructure. The data that gets fed to them has to live somewhere. And there is, in my view, room for all of them for some, you know, one of the reasons I articulated you need different models and different architectures and different training sets in order to do a multi-model strategy. Well, there’s room for all of them, but as I was saying, Google allows you to use all of

Richard Tromans (18:35.137)
Hmm. Yeah. That’s a key point. That’s a clincher, isn’t it? Because obviously Anthropic and OpenAI are not sort of like, you know, enthusiastically encouraging you to use other models. They really do want you to use their models, right? But what also about Microsoft? I mean, in some ways that’s the elephant in the room, really. You might say the underexploited elephant, because, you know, Microsoft captured the legal world with Word and the whole, you know, the whole suite there.

Andrew Simon (18:37.185)
And that that’s just fantastic.

Andrew Simon (19:00.399)
Did did. I remember Corell by the way a long time ago, right? I’m sure some some do.

Richard Tromans (19:04.941)
Yeah.

Yeah. Well yeah, I they’ve been there right from the start. So and yeah, the the in terms I mean they’ve been in and out of legal you know a few times. They’ve had various little projects over the years, as people have told me, and then obviously the latest one is the legal agents. Does does what Google’s doing kind of create some friction there? I mean, do you see that you know, like people will start to go, well, we’ve got Google Docs and we’ve got Google this and Google that, and now we’ve got the Gemini suite for legal. You know what? I’m gonna start asking my clients if they’ll accept Google Docs.

You know, I mean is that are we gonna go down that road?

Andrew Simon (19:39.65)
I I think

I can’t speak, of course, for how Google sees its strategy, but that seems like an obvious path that would get walked, right, over time. I I I still maintain what I said earlier. I think there’s space for both of them. I don’t know if either Google or Microsoft would like that answer either, but the reality is you can develop things to happen in Google Docs or LibreOffice. Right? LibreOffice is fully open source, it’s been around for a long time.

And then you can still have the lawyer’s interface inside of Microsoft. the struggle there would be where does the compute sit? Who where are you using the data centers, right? If you’re using Google Cloud to do the AI work, and that becomes where more and more things, whatever those things are for a given organization happen, then that necessarily takes the share of the expanding market away from other cloud competitors.

Richard Tromans (20:40.331)
Yeah, yeah. And that’s it again again it goes back to this point of centrality, doesn’t it? It’s just like, you know, if the information is coming in, if the information is held, the compute is there, the partners are being fed in and then you’ve got these other models which you can bring in, you’re even doing your coding there now. Y you start to live in this environment.

Andrew Simon (20:59.065)
Yeah, that’s that’s a great you start to live in the environment. And the other great thing about partnering with Google is the access to the forward deployed engineers, right? We found that’s actually been a really good reciprocal learning. you know, in working with our team. We have a very small app dev team, by the way. It’s it’s I do some own my own some coding, and then we have three people. Dalton leads it, he he’s on here in the background. And they we’ve learned things from them and they’ve learned things from us.

Richard Tromans (21:08.501)
Mm, you know.

Andrew Simon (21:28.643)
And so it’s been really a nice symbiosis. And aside from the technology, the interaction with those professionals has been enriching.

Richard Tromans (21:36.289)
Last question. I think some people may have been surprised that even though it was a very, very chunky release, that c Gemini Enterprise for Legal has relatively few applications ready made to use. Like you mentioned the NDA one, which while fed into. But there isn’t there isn’t a val I mean, is it gonna be a bit like ’cause Claude for Legal I know also started with a very, very small group. Now it’s got I think more than sixty or so different skills, as they’re called. do you think that will happen with with this? You know, Google scenarios as well?

Andrew Simon (21:52.602)
Yeah.

Andrew Simon (22:05.029)
yeah, more coming. I mean, I can tell you even from what we’re working on with Google, more coming. And was having a conversation with one of the Google executives leading the program yesterday and we were talking about the okay, hey, you know, we have this little bit of a pipeline, where do we want to prioritize and not? And so for sure more coming.

Richard Tromans (22:09.674)
Okay, right.

Richard Tromans (22:21.496)
Huh. And also some of the it kind of felt like some of them were tilted towards the in house world. It kind of felt that way. Not so much for these sort of big, you know, New York, you know, transactional powerhouses as it were, you know.

Andrew Simon (22:34.563)
I think that’s right to start. and you know, where we want to contribute is remember philosophically, I want to go where our clients are going. And so helping meet them there is incredibly valuable because we can’t get to the efficiency and the change folks like to talk about without building trust and confidence in the technologies and the systems we’re using to do the work. And so I think

Richard Tromans (22:36.44)
Hmm.

Andrew Simon (23:02.763)
doing things that are incre that are helpful to the in-house side and working with them to say, here’s why this is working or why it’s not, right? That feedback is very valuable too, is in my view the right way to start because otherwise you’ll you’ll never unlock the changes I think many of us are you know keen to talk about.

Richard Tromans (23:23.436)
No, totally. And and if this is to become a really successful market ecosystem, then the clients have to join in, don’t they? Your clients have to join in. If you’re going, Hey, the Google thing’s great and they’re all like, Well, we don’t wanna use it. It’s not gonna work, isn’t it? You you want everybody together.

Andrew Simon (23:35.925)
Or I’m so correct, or I’m so glad you used it, but could someone just redo it for me just to triple check? That’s then where did the efficiency go that everyone’s asking about?

Richard Tromans (23:45.283)
Mm. Exactly. You’ve you’ve all got to play you’ve you you want to play on the same, you know, game surface. Absolutely, absolutely. So very just la last thing, the where do you s I mean, I know it’s very early days and it’s very hard to make predictions in this market, but very, very roughly, where do you think you’ll be with Google Enterprise in the next twelve months or so?

Andrew Simon (23:50.437)
That’s right. That’s right. That’s right.

Andrew Simon (24:04.759)
I think what you will see out of Weil specifically is a variety of what I’m calling mini-systems like Benchmark that operate as useful and productive experiments for advanced capabilities, doing things that are really meaningfully different looking than how legal practice was done. In parallel to reducing the number of surfaces our lawyers have to interact with.

in order to get to the service they want. right now I think one of the biggest complaints, I can’t tell which is the biggest, is hey, I don’t know what tool to open to do what for which client. That’s just horrifying.

Richard Tromans (24:42.344)
Yeah. And well the problem is but they they often all do the same thing, don’t they? I mean, there’s I can think of twelve companies off the top of my head that all do contract review.

Andrew Simon (24:51.865)
That’s that’s exactly right. And the the interesting thing is clients all have different perspectives of which tool you’d like they’d like you to use. But I’m finding that if you can just give them some very clear, consistent answers and say, I promise, here’s what’s going to get used, here’s where that data sits, and it’s not there’s 15 options. it’s just a simpler way to have the discussion, right? It’s just people can make decisions and

Richard Tromans (25:16.906)
Fantastic. And very, very, very last thing, ’cause I just keep on thinking of new things to ask you. But the would some of the things that you’ve built for your own firms use eventually become applications for everybody sitting inside Gemini for legal?

Andrew Simon (25:30.639)
I can’t comment yes or no, but I can say under discussion.

Richard Tromans (25:34.336)
Okay. Okay. That’ll be interesting. And and obviously I’m sure the other law firms out there who are getting, you know, involved in this project are probably having the same discussions as well. Fantastic. Really interesting. Really, really interesting. I mean, you know, the field of legal AI, it’s it’s it’s going at an extraordinary rate, isn’t it?

Andrew Simon (25:42.35)
I would imagine.

Andrew Simon (25:52.323)
It really is. I I am very excited. And I think, you know, we didn’t we didn’t get into this, but if we think about where things go beyond large language models, as we people get comfortable in this industry doing things that aren’t just on a frontier language model, on smaller language models, on models that aren’t architected at all, like LLM, right? There’s you know, there’s interesting there’s interesting research in the liquid neural network.

And world models that I think artificial lawyers

Richard Tromans (26:23.891)
And digital twins as well, that’s another popular topic. Yeah.

Andrew Simon (26:26.307)
Digital twins, exactly. And that’s been very interesting, you know, when I was before I joined Weill working with some big tech companies on that and say, Okay, how are we going to do physical physics models to simulate factories before we build, say, a semiconductor fabrication? I think though, as those things apply, maybe not physics models, but as those things apply to the practice of law, you will begin to see just another wave of innovation. and a lot of pressure on

Are you sure what it output was correct? Are we sure it used the right preference?

Richard Tromans (26:58.508)
Yeah, well that’s that’s that’s the elephant in the room. Yeah, yeah, exactly. Well thank you, Andrew. We will have to cover accuracy another time, and digital twins and all the other things. But thank you very much. Very exciting and thank you very much for sharing with us. Thank you.

Andrew Simon (27:13.529)
Thanks for having me, Richard.