Matthew Brockman

So welcome to our GenAI discussion as of November 2024. The aim of this session is really to give some insight into the pace of development of GenAI as it increasingly supports workflow applications and also some of the themes we might target as an investor. And I really want to try and bring out some practical examples of the technology, how it's been deployed in the companies, how your customers are using it.

I'm joined by three visionary leaders who all have very deep experience on the power of AI and how it may influence products and their customers for their years to come. David Carmona, CTO for Strategic Incubations at Microsoft, working directly for the board, responsible for the last five years or more of his 24 years at Microsoft for the relationship with OpenAI, the integration of ChatGPT 2, 3, 4 into the Microsoft suite, the creation of Copilot across Office and other Microsoft systems, and really the person at Microsoft responsible for looking over the horizon and keeping Microsoft at the forefront of new technology.

Avaneesh Marwaha, architect and CEO of Litera. Litera is about a $3 billion EV business in our portfolio, specializing in software workflow tools for lawyers, and thinks very deeply around the products and how they're used to streamline the efficiencies in legal workflows. Finally, but not least, Soeren Brogaard, architect and CEO of Trackunit. Trackunit is worth over a billion dollars providing software and connecting people, machines and processes in the global construction industry. Really a market you might think could be at the tail end of how you might adopt a new technology like AI, but really has been a test case for us about how this can roll out a change in industry. So, welcome.

David Carmona

Thank you.

Soeren Brogaard

Thank you.

Matthew Brockman

David, if I turn to you first and to set the scene. Microsoft, largest provider of workflow software in the world, acknowledged to be right at the epicenter of AI revolution, first tech giant really to properly define AI, to launch Copilot alongside existing suite. Could you highlight one or two of learnings in the last year or two as Copilot has launched? Have you seen how customers have really adopted the product?

David Carmona

Thank you and pleasure to be here. If I think just of one key challenge that every company is having in the last one or two years, is how we move from incubation to scale. Scaling a technology both internally to the entire organization because at the end of the day, you cannot have a bottleneck trying to centralize all the innovation in the company if you want to have every single product to be redefined with AI. So having that at scaling the company, but then also thinking externally how to enable customers at scale as well, that is a completely different muscle that you have to develop that every company has had to develop in the last one or two years.

Matthew Brockman

And that's getting down into the workforce. So it's taking away from a CTO or a CIO or somebody who's an innovator and saying, how do I get this through mass adoption through hundreds or thousands of thousands, I guess, of employees?

David Carmona

Exactly, because we talk a lot about the power of this new generation of AI, but we don't talk enough about the new access that is providing, right? Anybody can now use this AI because the way that you customize the AI is by literally telling it what to do in natural language. So anybody can do it.

Matthew Brockman

And have you been surprised by the pace of innovation? The LLM was launched 18 months ago, start of 2023. This used to be sustained innovation in terms of how people are engaging with it. Now, people are building agents, you are supplying tools to build agents. Is that a surprise to you? Is that as you expected? Just how have you seen the last 18 months?

David Carmona

I mean you can never expect exactly how things are going to happen, but it is something that hasn't happened in the past year. So this is a concept of scaling AI, of having AI. I like the concept of foundational AI even more than generative AI because that was the core of the change. That change has been in the labs for many years and now we are seeing the implications of that.

Matthew Brockman

Where do you think we are on the evolution of this technology? If you compared it to the internet adoption, is there a parallel you'd say as to where we've reached to date?

David Carmona

Yeah, it's an interesting point because if you compare with the internet adoption, it was very linear. It was always increasing. So the number one metric was the adoption worldwide, right? Of internet connectivity. This is different, this looks different because right now AI is at that level of scale. You can use it anywhere. And yet we still feel that we are at the beginning of this wave. There's so much innovation that is going on in the area that it's a combination that brings a very interesting challenge, which is how do I apply AI today while I get ready for the AI in the future? So that's quite an interesting dynamics that we're going to see in every company.

Matthew Brockman

So I guess actually comparing to previous technology wave, almost the rate of adoption has been incredibly rapid, has almost outpaced the development of that in terms of workflow and how it's used, it's almost now there's a catch-up of what might you do with this tool that everybody's got immediate access to.

David Carmona

Exactly. But at the same time we see still this potential for the technology itself to evolve at a very fast pace. So I think it's a very unique technology where you're seeing fast adoption while the technology still adapting a lot and evolving a lot.

Matthew Brockman

Soeren, maybe if I turn to you and obviously in a more vertical application area, does some of that resonate? How do you feel that it's penetrated into your market and people's adoption curves?

Soeren Brogaard

To put things in context, think about Trackunit as the global provider of connectivity to machinery that build our roads and our houses in construction. So today we collect data on a real-time basis for more than 3 million assets throughout the world. We collect more than 2 billion data points on a daily basis and we are in many ways a data factory and that data is consumed by our cloud applications and served back to rental firms, to contractors themselves, but also the machine OEMs. So we are orchestrating this, you could say ocean, this lake of data. And to us, it was a bit of a aha moment when we suddenly were able to start taking our existing analytics insights that comes out in reports to our customers and start putting much more rich context on top of it. We very quickly realized we had to make this useful in our products.

So the scaling discussion I think is exactly right. So to us the scaling is helping our customers to adopt the technology rather than our, you could say, internal organization. So it was actually one of the questions I had when I was listening to you on how does Microsoft see the shift or the balance between AI and GenAI used for internal efficiency purposes and putting it into the products that they deliver and sell to their customers. I mean, are you seeing how is that balance emerging? Because to me, it was always a matter of I got to get in front of my customers and put it into my product for this to really have a profound impact longterm.

David Carmona

It's a great question because we see that changing a lot right now. So it's using the pattern that I think every technology has used in the past, which is initially, we use a new technology to try to do things that we were doing already, but quicker. Productivity, right? And that's the primary focus of AI right now. It's all about, hey, let's transform my company to be more efficient. That's one key aspect of it. But we see that changing and we see companies getting more mature on that usage to start thinking of things that I can do now that I couldn't do before, especially new revenue streams. So how can I use this not only to be more efficient, but to provide services that I couldn't deliver before so I can differentiate with the competition. That I think is going to be the next battleground in the industry.

Soeren Brogaard

Yeah, so to me, that's exactly where I think the focus is today. And here we got to be super quick on defining our assumptions, our hypotheses, and start quickly to learn. And what we've done at Trackunit in this very early phase for the past year is we've been absolute adamant about getting our data quality and our data structure right, and that has been a theme for the better part of four to five years. So the foundation was there. Then we were also adamant about partnering and opening up and not thinking we had to build everything ourself, and also getting a partner or founding customers in very early on and discuss the problem to be solved and get them super close for experimentation and for not just on the product side but also put on the monetization side.

David Carmona

Yeah, exactly.

Soeren Brogaard

And I would love to talk more about that today because the monetization element here is almost a chapter on its own.

Matthew Brockman

I'm going to bring Avaneesh in before we get going. Avaneesh, you've been in the GenAI space for a while, but obviously LLMs coming at the early part of '23, I assume some kind of transformation certainly in terms of how they could be applied. I'd love to hear you speak a bit about what was happening before, how much that's changed, what your sense of it, how fast it's moved within the legal profession, which has created a lot of hype over the last-

Avaneesh Marwah

Yeah, it does. I think the changes that we've experienced in the last 18 months has been really beneficial for this industry. Litera's been engaged with AI for years prior to ChatGPT and OpenAI coming out with a product called Kira that was doing AI due diligence, had its own LLMs built many years ago, been refined, millions of documents have gone through to build the machine and has been probably the only one in the market for a long time doing successful due diligence for transactional matters and was seen as the early entrant into the space. And then as a business, we've also been doing natural language processing in our technology for decades and never moved into machine learning for specific reasons until now, but the industry itself has not been ready to be cloud acceptance until, I would say, since OpenAI came out. We've been pushing the industry for a long time to go to Word online, use Outlook online, but they weren't ready.

But now all the innovation is sitting there, so they have to go there and they have to use Azure. And so that's opened up our ability to now build everything we wanted to build for the last five years. We can now actually put into market very effectively and quickly. But I think until the shift with AI, we couldn't even think about what was possible because firms were so locked down on what they wanted to even do with their data. Now they're open to it, but even then they want restrictions on how the data is used. Who does the processing? Does it leave their four walls? Is it going to your four walls? What model are you using? Is that model changing every day? Is it the same model we use six months ago? So I think they're still learning how they want to use the technology, but for us it's been an accelerator of all of our product management and innovation that we've been thinking about for many years.

Matthew Brockman

How important is the vertical nature of what you're doing, I guess basically on Trackunit and on Litera in terms of specific data or specific use cases or customer comfort with how they use the product? I mean, how have you seen that manifest so far? And I guess again, we're still relatively early in how this is going to play through.

Avaneesh Marwah

I think selfishly it matters being vertically focused because it does help us quite a bit. Everything that we do, it has to be native to how the attorney and partner work today. It doesn't mean you don't evolve it or innovate on it, but I think asking that industry to change holistically what they do to use your technology has low adoption. So we've always found it better to think of their workflow and where can you easily just attach innovation, so that way they get value of everything around them.

So we think of our products today and say, where can we intelligently place GenAI to either reduce waste or increase the accuracy of the result set. To just do GenAI for the sake of saying, hey, we're a firm using GenAI I think is the wrong approach, but to use it in a way that improves attorney-client relationship or improves outcomes for partners makes a lot of sense. So I think for us, I like saying vertically focused is really important because we can deeply think about the problem set and with the outcomes they're trying to get every day. We're not thinking about eight different industries, we just have one and we're honed in on it and we have 15,000 customers that give us really good feedback. So I think it's a good way for us to innovate quickly, iterate with them and find success.

Soeren Brogaard

In Trackunit, we've been cloud native for the better part of 20 years and it sits within the way we do business. So first of all, I totally agree, unless you are a very large horizontal player, this is the way to go. Gartner has been now predicting that these industrial tailored cloud platforms, 70% of enterprise will be running on a highly tailored verticalized cloud or industrial cloud in the next 15 years or so. And it makes so much sense. And also when we get to GenAI that you actually have a domain specific model that you can translate data into insights and you basically have this opinionated approach to the data. That's really what creates value. That's really what puts you apart. So the more verticalized, the more specialized you will be. That is what our customers really need. They're really data rich you could say, but they're insight poor in many ways. And one of the ways to create deep foundational insights that can change the daily life of an operator in a machine is to be highly verticalized.

Avaneesh Marwah

And I would agree with that. We even have, there's large startups coming out now that are trying to apply GenAI to legal and I think they're hitting the wrong spots. I think end of the day, the funniest thing that lawyers want to do today is find the right answer faster. They're not looking to do big large documents, they're not looking into anything else. They've been spending hours as a first year associate looking for some random clause from an old matter, can they find that within seconds versus six hours? That's the value add for them. So being vertically focused, looking at your data proprietarily and with focused GenAI, I think you can make meaningful impact in those industries.

Matthew Brockman

I'd be interested in your take on that, David. Obviously Microsoft has got billions of users and so much breadth of application. I mean you are almost coming at the other end of the spectrum, right? There's so much you could do and so much scope. How do you hear this and think about it?

David Carmona

Well, the name of the game right now is AI plus vertical expertise. That's the magical combination. AI by itself is not going to do anything. So there's a book that a lot of people are talking about that book now, especially Microsoft, I see everybody reading that book, strongly recommend it. It's called, I think it's called Technology and the Rise of Powers. And it does a deep analysis on the previous three industrial revolutions. And what were the factors behind the definition of the leader of that industry revolutions? One by one, they went through the exercise. And the main conclusion of that book is that the key factor for England to be the one leading the first industrial revolution or the US the second one, or the US again the third one is not the innovation in the first years. The key is how you bring that technology to every industry faster. That's the winner. It's not the first innovation. It's not about scientists, it's about engineers. It's about bringing that innovation and scaling it across your entire industry quicker than anybody else.

And that at a different scale at every organization, it is exactly what is happening now. Everybody's now all about who's able to scale it first and bring it to my domain first to infuse it in my domain, my expertise. That's the key.

Avaneesh Marwah

I think Microsoft's done a good job with this actually. Looking back over the last eight years of being in seat, we've always been a partner of Microsoft. My view is my job as a CEO is to make Microsoft more valuable to law firms, right? Building a good skin on Word and making them use Teams and all those things. But with the advent of OpenAI and their investments there, I feel like they're actually now looking at vertical software and saying, let's group these people together, the experts, let's work with them on saying how can we deliver our technology through you in a way that makes meaningful sense to the end user? And we've had a little bit of that before, but I think with all this new innovation, they're looking at it much more focused of like how do we win legal? How do we win education? How do we win these different sectors? And I think our partnership's actually grown significantly in the last 12 months coming out of all this OpenAI innovation.

Matthew Brockman

Let me pick up on a topic that Soeren raised a few minutes ago around the economics and how this might change, I guess, revenue models for software companies, but also just where the value accrues between maybe technology and labor. Again, any early signs of what you see about the scope of that in, I guess, maybe open up to anybody who wants to comment on how you see that-Soeren Brogaard: So historically we have been built, or our monetization model has been you connect a piece of machinery and you pay for that asset over a period of time. And that has moved to fleet level management and so forth. AI and GenAI and what we are now doing with our product called IrisX, we had to just take a step back and say, okay, how does our customers really want to consume this technology and how do they become comfortable with the monetization model?

And so we had to completely rethink the way that we make our customers comfortable, but also the unit of measure, the unit of measure is no longer a machine, it's actually a operation. And getting our organization, the sales side, comfortable with that customer success, backing all of that into our ERP system and also make it full transparent in the product and the customer's comfortable with it, which is slightly easier because they're becoming more accustomed to this way of buying was a small revolution on its own, had nothing to do with tech and everything to do with the way value is being perceived and consumed.

So we are in the middle of it, guaranteed to change a million times, but that's the beauty of this world. You've got to move fast, experiment and adjust.

David Carmona

Yeah, I agree. And I think another interesting thing about AI that's not exactly the same to what happened in the past with other technologies that in this case it's a full stack of value that is being transformed with AI. So you look at, it's a full economy from the data centers to the chips on top of that to the models. I mean we weren't talking about the model economy in the past, look at that now, right? There's an entire economy now just on models. Then on top of that, things even like agents, then on top of that application. So there's a full stack in there. So I think that when we talk about business model, the first thing that I always discuss with customers is where are you in that stack? Where do you want to provide the value to customers? Do you want to provide the value here at the bottom or do you want to provide the value at the model?

Or usually in vertical applications, do you want to provide the value at the app level? So in that case, the business model is going to change dramatically. The higher you go in the stack, the more value that you can prove to the customer and that likely you can have a conversation with customers that is more about outcomes and not even consumption. What is the outcome that it is creating to you? And coming back to the initial conversation, not outcomes in terms of efficiencies, but outcomes in terms of new revenue streams that I can help you unlock with AI. That's, I think, a very relevant change on the business modeling conversation.

Avaneesh Marwah

We're balancing a lot of what is a feature and what's a product in our solution set. We have a very broad set of applications that we provide to law firms. I heard from a firm recently, they're even thinking about how they price their client work going forward. If the client doesn't allow them to use AI on their data set, maybe they'll price a higher hourly rate going forward for their services versus a lower hourly rate if they're okay using AI on their data sets. So even the firms are trying to figure out how they are going to price their services with their clients based on their comfort level of using this technology.

Soeren Brogaard

What we are seeing in our industry is that our customers are quite constrained on just having the tech depth, the understanding that is required to take full benefit of our solution. And when I look across my 200 top customers, I stop counting at 300 open positions in tech and IT and GenAI for that matter or just data scientists. So there's also a big ask for someone like Trackunit to make this tech stack ingenious, make it simple, make it something that can be consumed very, very easily by non-tech people. And that's actually one of the big promises of GenAI is that we can make this a lot more simple and is our obligation almost to take the deep tech out of it and make it so damn simple. I mean the ChatGPT is 4, 5.0 now can pass the bar and the level of intelligence that sits there at the fingertips. How do we translate that into a context rich new reality for all customers in our vertical market? That's where we are spending the most of our time. Get the tech out of the way and humanize and make it all ingenious.

Avaneesh Marwah

Are your biggest customers asking for their data set back now? So we have some customers saying, can we have greater visibility into what you hold so we can have our own teams also working on data scientists and stuff. We love what you provide, but if we had access to it as well, we could build our own applications on top of our own data. So now we're investing a lot in our product teams as you look at our API architecture and say, can we make it more robust to allow for more data to go in and out versus just us doing all the work? So that's a new thing we never thought about is how much investment we now need to make on the pipeline of data moving back and forth.

Soeren Brogaard

We see the exact same thing. The largest of our customers, they want to take our product that they love and have been using for years where we have a general data lake and we have a lot of benchmarking capabilities, but they also want to take that stack and almost create like a Delta lake and then start inserting their own data streams, build proprietary applications just like we are and then pass it back out into the industry. So that's a clear trend.

David Carmona

You are both hitting something that I think is super interesting as well. Behind all of that, there's a huge effort that we have to do in every company on upskilling, training. It's critical for the [inaudible 00:22:39], it's not that it's my book, I promise that this book is not mine, but this book, when it goes deeper, it actually identifies upskilling, skilling, as the primary component to enable that infusion into every industry. And it's true, it happens in every company.

So this is not a technology, and I always paint this as an inverted pyramid where at the bottom of the pyramid, the peak of it is just the data scientists, the deep technical people. But that's a small portion of it. And yes, it is important and yes, customers will want to have that muscle many times so they can go deeper, but the critical one, it happens as you go up that pyramid. You have the professionals. So talk about the lawyers, so how much skilling you have to provide in there for your product to be successful. It is critical. We have to upskill an entire generation of professionals to use this technology if we want this to be successful.

And then on top of that, now you go even at the societal level, it's the AI basic understanding to be used even as a consumer. We had to do this other times in the internet, we will have to do again and it's going to take a lot of effort and energy and alignment across the entire society.

Avaneesh Marwah

I was blindsided by that a bit, just talking to a client recently we're showing them technology and they actually asked, how are you going to help us put this out? We like it, but how are we're going to get 10,000 users using this? Are you going to hold our hand? Are you going to be here every day teaching us? So now we have to go back and think from our budget perspective, how do you actually afford this training that we now have to provide to get adoption of our technology? That's we think revolutionary, we think is high value, but it's going to require a lot of manpower to get it out in the street.

Matthew Brockman

I'd love to talk a little bit around that on your own product work within your own business. I'm conscious in Litera, you've got a bunch of people building product that's been in the market for years. Their job is to continue to improve and deal with the old bugs, and now you want to bring this innovation to wave in and just the classic innovator's dilemma, that what made you successful isn't what you need to be building from here. How have you tried to tackle that challenge?

Avaneesh Marwah

I think we have three models working. So one that we did pretty quickly was we had an external SCRUM team that worked with the HG team here and their data scientists to build a product really quickly to bring back to market around data extraction, around deal points in a transaction. And that was our first proof point that we're onto something and we had customers and we changed our go-to-market model. We had a beta program and then an EAP program we never had before and we proved that we could do this rapidly because in a business like ours, rapid doesn't happen anymore. With that success, we started building our own innovation then internally. So instead of having external SCRUM teams, we now have an internal innovation team that is able to look at our existing products set, not new products, but our existing products, how do we innovate internally on those things to bring GenAI to it?

And the third piece is going to be acquisitions. I think we've been very cautious about watching the market and saying, we don't need to be early to the game. Let's let people figure out where it's going to apply. Let us ask the market where they think the needs are. Let's learn and listen and then hit the ground hard with our answer. And so, one way to hit the ground hard is innovate internally, but equally it's to go acquire the talent and the products that are hidden in businesses and bring them to market faster.

So it's been different. We've been trying to protect our cost margins along the way because you could go crazy on this and burn all your EBITDA on innovation, but we're being very strategic where we can and I think external SCRUM teams is a great way to protect EBITDA and then acquisition is a great way to bring in the ideas without spending a buttload of OpEx on it. So we've been balanced in the approach, but I think we're going to see as we go into next year, a lot more investments internally showing up to whether it's rolling out the tech and training or innovating, I think we'll see a lot more of that cost-

Soeren Brogaard

There's also a bit of are you protecting your position? Are you actually playing defense or offense? I think that plays into your strategic choices. We are clearly playing offense in Trackunit and fundamentally I think we are dealing with a massive generation shift as well. We don't have enough people in construction, there's not enough of young people that comes into this industry and our promise is to actually being able to get someone who is one year in the job, the expertise of someone who's been sitting behind his computer for the better part of 20 years and managing a large scale fleet. That expertise of 20 years of hardcore knowledge on how to do this, which is super complex, put that in the hands of someone who just came in and is one year in the job. And to do that, you got to just accept the fact that you cannot send Deloitte in or 200 consultants to make this work. You got to be so simple as picking up your web browser and have a strong agent or a co-pilot as your new best buddy that gets you quickly to results.

David Carmona

It's funny because I hear you both and exactly the same concepts apply to Microsoft. It's exactly the same. AI assisting an individual to do something. We're seeing AI also as a catalysis for Teams to work better. And it is funny because the same problem we have in software development, so we have huge needs of talent for software development and what we see when you look at this, we have done a lot of analysis internal, external on the improvement on both quality and productivity for software developers using tools like GitHub, Copilot, and what you identify is not only the productivity that you create, but it's also the knowledge transfer that you can do for the most senior software developers to the more junior software developers. So the highest improvements is always in the junior developers.

If I may add one additional step on top of that that we will see also happening is AI from being applied to an individual to being applied to a team, now we'll see to being applied to an entire organization, to an enterprise. That is a concept that we're still not grasping and I think it's going to be a huge thing coming next is how do I look at the entire enterprise as an entity that can be optimized with AI beyond individual processes or beyond individual teams or individuals? That's a huge opportunity that we will also have.

Soeren Brogaard

I mean this all zooms in on this concept of pace of learning that the future competitiveness really is not about knowing it all but learn it all and it's how fast you can learn it. How fast can you enable your organization to deliver differentiation and deliver value to your customers. And that flywheel, I mean back to what's really different now, the clock speed has just increased really, and when we get to talk about how does this look five years from now, I would say, well, how does it look next year and how do we get our organizations and our teams and our investors comfortable with this pace where this constant experimentation in the good old days, maybe it was a cycle of three or four years, now it's a cycle of three, six, 12 month.

And there's a high degree of unpredictable models, not so much from a tech point of view, but more from a business outcome point of view that is going to be harder and harder to describe the ideal business model that will make investors comfortable. And we got to step into a world where we allow for, I mean you are famous for saying, or I think you stole the quote, get the adults out of the room. There's a little bit of that going on where and how do we create comfort across the different ranks of investors and executives and teams to get to that new way of running our organizations because it's almost impossible to predict what's going to happen.

Avaneesh Marwah

Having the breadth that we have from drafting documents, transaction management, data warehousing, providing analysis, on any given day, a startup can show up and try to compete on any one of those factors with us. Even if we're innovating doing things, it's easy for someone to get funding now to go build something and all of a sudden they're competing with us in an area that we didn't think we had risk for a while, and now we have to go and divert resources to just put that to bed and go forward.

There is a little bit of this whack-a-mole situation a little bit for us where we're innovating, we're being aggressive, we're on the offensive a lot, but we do find ourselves in some areas like, oh man, we [inaudible 00:31:18] be defensive and catch up and just so we can keep doing innovation. It's very much a quarter by quarter as I'm thinking about the [inaudible 00:31:26] going forward, they're going to be variable in nature. You're going to hear competitors and the next day, they're gone. Or they've been around for a while. So I think that is going to be a tricky water to navigate for a bit for some of us.

David Carmona

Yeah. Learning from us, because I feel also very identified with those, at the end of the day is like two different speeds that you need or two different muscles that you need in any organization. One learning that we had is how they require completely different, not only processes or approaches but even cultures to accomplish those two different speeds. The way that we end balancing those two is realizing that hey, we need to have those two in parallel. So we need to have AI from today infused into every single product in Microsoft. That's the only way that you're going to scale. You have to create that motion with the technology that is available today, scale it. You are already late if you are not scaling the technology that you have today. You are behind if you are not doing that. If you are waiting for something to happen, you are late. You need to do it now for sure.

But then at the same time realizing that the technology is changing so quickly that you need to have a muscle preparing for that. And in our case, that's the department that I work on, a ring-fenced department, strategic incubations that is really focused on incubating those new business opportunities that are going to be enabled with AI. That's a muscle that you also need to keep because at the end of the day, if you just focus on making your existing business better, you're missing another big opportunity which is, hey, maybe my business can be transformed with this as well. And you need to do both in parallel.

Soeren Brogaard

This is a super interesting discussion because in a company like Trackunit, we have 450 people, right? That's a rounding error in the Microsoft context. So I am obsessed by trying to get this incubation, this experimentation DNA into my core development team on a continuous basis. Clearly there will always be people who look after the core foundation, the stack and RevOps and DevOps and whatnot, but it's in the size of a company like Trackunit, we got to insert this DNA and this mindset into the everyday dealing with our products and our people. And it also goes for HD and it goes for the board as well. So I need people on my board who can challenge me on the next generation tech and I think it puts investors like HD a little bit also challenging you guys, Matthew, on what can we expect from our current investor and what can you bring to the table, which I think you're doing great with your shared services, but in the boardroom, what can we expect from future investors?

David Carmona

Yep. And I think that that can work as long as you are successful doing one thing, it's creating the motivation for them to do that. I'll tell you something, if you have a business unit that is being incentive on one particular metric like revenue on that P&L, they will take any innovation that you throw at them that they will use it to improve that metric. And it's very difficult. If you don't rethink the way that that business unit is incentive for them to create new businesses, that can in some cases even challenge their existing business. So it's against that initial approach, so very tricky to do.

Avaneesh Marwah

That's a good point. We actually, if you only pick revenue as your metric, now going forward, I think it's going to be an issue. There's a KPI just put in the business it was to track how many implementations of this protocol foundation we have. Foundation is the data lake that we're trying to give to firms. A hundred firms have it. I think it should be 500 in five years, 200 in 18 months. I want to track that as the KPI. I don't care how much you paid for it, let's just get it into the business because it'll have material long-term value for Litera if that is sitting there. It is the box, it is the brain, it just has to get in place and then we're long-term selling after that.

But it's counter to bookings because we may want to have a deal with some firms saying, hey, it's yours, let's just get it in the door because it'll give us long-term sell-through of cross-sell of other products. So if it was just booking and just revenue, that product may not move as quickly, but it has to get out. It has to move so fast. In this day and age, we have to be the one that's sitting there with the brain. Whatever it takes to get there is the ultimate prize.

David Carmona

And it's scary. Scary, right? Because for some moments, I still remember I was in Microsoft where we have to do the huge transition for us from selling, I mean, let's say boxes, software in boxes to sell the cloud. I mean the primary barrier that we had at that time was not the technology, it was actually the culture. So you have a seller who's being their entire lives selling boxes that they know that the monetization of a box in the short-term, let's be clear, much better than the consumption-based model. The effort that you need to do to change that, it goes in that direction. So I still remember that we had this specific metric that was very scary to see. No, we're going to incentivize the field sellers by focusing on the cloud

And a cloud product that you sell is going to count much more than a box product that you sell. And it's scary because you also have to look at the revenue.

Avaneesh Marwah

Right. And the margins, right? You got to look at how this plays out over time. It requires good forecasting and abilities, but to your point, it requires an investor who is okay with also short-term variances for long-term gains, right? If you're investing only for 18 months, probably not the right decision, but if you're a long-term investor, it works.

Matthew Brockman

Sorry, Avaneesh. I was going to say it is something we spend a lot of time worrying about, which is our industry is managing money. To meet a lot of very financial people and you give them financial metrics, which is basically bookings, churn, retention, bang, bang, bang. You drill it into somebody when they're 25 and hopefully they absorbed it forever. And now we're saying, can you also think about product-market fit? And you can think about customer adoption and customer satisfaction and usage? Stuff that is in their consciousness, but it's not been drilled in. It's been the number one thing. So we are rapidly trying to get people to think about that and evolve and this kind of discussion, these kind of forums are so valuable for even our people to hear what's really driving right now.

Avaneesh Marwah

Right. This thing about the book I want to present to a future buyer of this business. What are the metrics I think they're going to care about is going to be usage adoption. What's around the data that we've procured for firms? These are all things that are not bookings and EBITDA and churn, but they're long-term metrics for success. These are all things that you can now invest in long-term and I think we haven't thought about that historically, but now it's top of mind.

Soeren Brogaard

I also think businesses in the future is going to be valued on, one, that tech stack is can you actually apply modern technology against the company's raw infrastructure? Do you end up with a lot of tech data and so forth? You can't really break free. But also the quality of data and access to data and your position within the ecosystem. I mean they're actually able to share and monetize a point, a data point, whatever it is, multiple times because the industry trusts you, you have the right security layers, you have the right mentality and you're a trusted entity. So that mode of access to unlocking future tech and create new value props I think is going to be equally valued as ARR, churn, growth and ultimately where can you take the company in five years. Because it's a foundational element that we haven't, I think catered too much to.

Matthew Brockman

David, pace of change, pace of the underlying technology, but obviously the models are evolving, the way the models have been deployed, agents evolving. Does that keep going at the same pace? Do you see it accelerating from here? Do you think the major real technology innovations already happened.

David Carmona

Yeah, I think I will go back to this full stack that we see on AI, right? There's so much to do in each of the layers of that stack. And yes, of course at the bottom of that layer and even higher than that, like models, we're going to see still a lot of improvement in there. We talk now the Moore law that it was very focused on the computational, how that is starting to translate into the AI exponential growth. And the unit there is going to be for sure token. By what? By dollar. So we have to do a lot on increasing the power, decreasing the requirements of those models and at the same time make them more available, make them more affordable. So that's for sure at the bottom of the layer. Yes, we haven't reached the ceiling there by a long shot, but as interesting that that layer is, it's more interesting the layers on top of that because there's even more potential in there.

We're just scratching the surface on the possibilities of AI when you apply it to the business. That's the huge exponential cure that we're going to see in the next years. And yeah, we can talk at that level that application patterns will change. We're seeing now this huge evolution into agents that are not only assisting humans but also acting on behalf of humans with their direction. That is a significant paradigm shift.

And we'll see more than that. We're starting to see these long-term running models that are able to reason over time and be even more proactive on the goals that you are asking the model to achieve. Those are significant paradigm shifts that for sure we will see. But then when you apply that to the industry, that implementation, that is where we're still at the very beginning of it and we'll see a lot in that direction that we were saying before, moving from productivity to more about co-reasoning where you can do things that you couldn't do before because you augment your capabilities as an individual, to even going proactive into an entire organization, entire enterprise AI that can help you as an entire organization in that context, right?

Matthew Brockman

The legal industry in five or 10 years time, how much evolution do you see or how much change do you see in a model which has been very white collar, selling my hours? Do you see that being fundamentally restructured or do you see that being still the core but with a much higher level of productivity

Avaneesh Marwah

I think it's the latter. I think we, during every evolution that the industry goes through, there's always been this view of alternate billing, fixed fee. We'll do work like that and I think the industry defends against it pretty well and maintains their by hour model. But I think what we'll see is we'll have more associates doing higher value work. There are still first, second year associates spend their time at the copy machine scanning documents to send to a client. We still see associates spending manual hours doing big changes in documents that could be done with simple AI tools today. So I think all that gets taken off the table to some degree and they can do more high value work, probably more work on business development, finding new clients, finding new matters to work on. There's a significant untapped market of legal work that just never gets done or talked about, that I think this does that.

I view access to justice getting better with AI. If you think about down market and just the consumer side of legal, landlord-tenant disputes, think about parking tickets, think about DUI. That whole market of legal, which we don't really think about, I think gets changed the most with access to justice because now you can have the price of labor come down hopefully, and people can actually hire a lawyer to represent them. The amount of cases in the US at least that are [inaudible 00:43:18] by attorneys and people represent themselves against a corporation, it's pretty high. So if we can make an impact up market with technology, then it also has value down market, I think it's a win. White collar legal work will always exist, right? There's always M&A work happening, there's always transactional work happening. But I think where we're going to see applications of GenAI it's not going to be in the creation of a document.

I think that's an area we're all talking about, but I can't see Skadden or Linklaters or Norton Rose saying your first draft of an SPA is done by a machine. It's just not palatable to the buyer. But I can use GenAI to find the right clause to change or negotiate. I can predict outcomes of this SPA maybe faster. I can predictively do comparisons and there's things you can do along the way. So I would hope in five or 10 years we see the white collar firms being a lot more innovative in how they figure out the future of work and how they're going to do things. But I do think the change that we're going to see is actually down market. And I think the B2C legal market as we were, GenAI should have the most material impact. And I'm hopeful that it's access to justice is what we see as the outcome of this is more people just getting protected and having their freedoms recognized because they can now hire a lawyer.

Matthew Brockman

And if I ask you the same question, Soeren, construction industry, again, crystal ball 10 years out.

Soeren Brogaard

Yeah, so $13 trillion industry, second or third largest in the world. On the contractor side, a contractor running on single digit margins, the least digitized industry in the world measured on 10, 15 different parameters. So there's a lot of work to be done here. Clearly tech is one of the answers and fundamentally I think we will see a shift towards the use of what we call composite AI where we have the insights of machines, workflows, processes, budgets in the construction cycle being inserted in a way where you can literally act as a construction professional with 25 years of experience the first year out of college. So the knowledge that is available to improve margins, improve productivity is going to be available to everyone. I think the way that people are interfacing with our software today will also fundamentally change. It will be you speak to your language, you interact with your language as an agent or a co-pilot or on some other form where it's proactively infusing insights that will make an impact and fundamentally change this margin erosion that we have.

Matthew Brockman

I'm going to wrap this up, but maybe two final questions. One is pitfalls you see, things you've experimented with in the last year or two that you said, if I had to give a bit of advice to a fellow CEO or a fellow executive running one of these, I'll open the floor, what would you say I would've done more of this or would've done less of this or one of my key takeaways was this?

Soeren Brogaard

I don't think there's something radically different that is doubling down on certain elements such as being super close to your founding and most loyal customers in the very early stage and also finding those that are thinking about how to do things differently and get them super close. That's got to be super close to customers, the problem and reality and avoid being in this abstract think tank and trying to outsmart people through thinking. You got to get in, participate in the actions and get dirty, which really isn't that big of a change except from that the pace is a lot faster.

Then also assembly, advisors, and board and investors that understands the new pace, the new clock speed of tech and be comfortable with that and accept the fact that we need to put higher risk into our budgets. I know that sounds scary, but if you want to play offense, it's the only way you will end up taking 10 or 15 bets and you might only be able to monetize two or three. How do we make future investors comfortable with that? And then it's finally making our people at ease, not catering to change management, not catering to pace of learning, not elevating the entire organization up to the challenge, but thinking that it's only down to the 10 smartest people in the room to figure this out. You create a divide in the organization and ultimately it will catch up to you.

Avaneesh Marwah

I think there's probably a handful, I think, but the ones I think about is I think the business could have done a better job of using the HG, the scientists seem to actually build more product. I think we did one successful opportunity and it has proved itself so well, why wouldn't we do more together? It's a shared service that it's highly valuable to us. Number two, I think you hit it perfectly well. We were slow. If you look back historically, all the product development we've done, we didn't keep our customers close. Now we do. Had we done it a long time ago, maybe we would've done stuff faster, but it's a lesson learned. Like the beta EAP program that exists for everybody else, we should have done that a long time ago. Now we do it. We're about 12 months in providing great success to actually launch a product GA with five customers using it already and they can brag about it. So those are two I wish we just did more of earlier.

David Carmona

I always say three things. And I try to remember them all the time. This actually they appear in my book, not in the one that I was talking before, but the three things that-

Matthew Brockman

So you'll put promotion chance there, David.

David Carmona

That was super smooth, but I'm not saying the name. So the three things that I always say is first you need to think before you act. Strategy. And when you define that strategy, you have to think short-term and long-term. And the things that you're going to do short-term and long-term, not only they're going to be different, you have to measure them in a different way because a common pitfall that I see all the time is just thinking about the long-term and then that dies because you don't get the short-erm results or focus in the short-term and then you are missing an even bigger transformation that is going to happen, important to have those balance.

The second thing that I say is culture. This is not something that you're going to do, creating a center of excellence in your company. This is something that you need to infuse to your entire organization. And that requires up-skilling, that requires thinking how you're going to incentivize those people, and that requires not only the operational part of it, which are the processes, but even how you actually organize the team and how you are preparing them and empowering them every single person in your organization to embrace AI.

And the third thing that we haven't talked a lot today, but I say that is hugely important, is decentralize the execution, centralize the governance. Responsible AI, there are risks associated with AI that we also have to be on top of. And the only way that you can do that is by enabling and empowering every guy in the organization to execute, but centralize the governance so you have the control on what you are doing. Because if you don't do that, you are going to get in trouble.

Matthew Brockman

There's some very interesting recurring themes around, I guess, quality of products, quality of data, proxy of seated customer, organizational culture, investor culture. I mean, there's a bunch of these themes that just keep coming in terms of the way that you're going to ultimately be able to innovate fast enough to capture the value. My last question for you, David, just there's a lot of people, we are going to have a load of people in our audience at the HG and more broadly who are using Microsoft every day. In 10 years, the Microsoft suite of products, what's that going to feel like to use?

David Carmona

We're seeing this evolution in AI that is from being very focused on the individual productivity to being more focused on the core reasoning. So augmenting your reasoning capabilities to then we're going to see with this proactivity in the AI guided by humans, how we're going to enter in a phase where AI will be able to proactively reason on top of the knowledge that you have in your organization. To be proactive and not only assisting you on that reasoning that you're doing with it. So that's one interesting angle. The other one that I would say is the technology itself. So we are seeing how, not only from the model point of view, but also from the application pattern point of view, we're seeing huge application patterns emerging that are going to change the way that we interact with the Copilot in our case. So we do see a Copilot acting on behalf of humans, not only in the digital world, but then let's be on top of robotics in industries like yours where we're going to start to see AI acting on behalf of humans in the physical world.

That is going to be another interesting conjunction of LLMs and robotics coming together that we'll see also happening. And then the last one that I need to mention because that's my favorite incubation in my team is quantum computing. Let's not forget about quantum computing, not as an evolution of AI, but more as a compliment to it to a new set of problems that you can't just do even with AI today. So the level of scaling accuracy that quantum computing will have on very complex processes, like the processes in nature will transform every industry from now to 10 years for sure. We're already seeing this huge shift from a noisy quantum computing to this new era of reliable quantum computing. We're seeing, in Microsoft we're making a lot of announcements on how we are starting to create this reliable quantum computers that will be able to be used in specific problems at a level that is not even possible with any super compute capability today or AI.

Matthew Brockman

Listen, thank you very much. It's an exciting time to be a technology investor and to have exposure to a bunch of technology that serves these end markets. So listen, Soeren, David, Avaneesh, thank you so much for your time and we hope this has been enjoyable and I suspect if we were back here in 12 months, we'll be talking about a whole bunch of new things as well. So thank you.

David Carmona

Thank you.

Soeren Brogaard

Thanks for your time.