Autodesk, Inc. (ADSK) Earnings Call Transcript & Summary

We've crew from Autodesk here. I get a stand for this session, which is great because I'll be sitting in those chairs for about 6 hours today. So from left to right, we got Sid Haksar, VP of Construction; Mike Haley, SVP of Autodesk Research. And we all know Simon from Investor Relations. I'm going to turn it over to Simon for the most exciting part of the presentation, which is the safe harbor.

It really is. We may make forward-looking statements during the course of this presentation. Please refer to our SEC filings for information on risks and other factors that may cause our actual results to differ materially from these statements. Okay, off you go.

Okay. So maybe we'll start with the forward-looking statements. Anything to share? No. Okay. Well, let's do intros. So Mike and Sid, thank you for joining the conference. I think, Sid, you were here maybe 2, 3 years ago. But just for investors in the room, if you could remind us of your background and kind of the areas that you're focused on at Autodesk.

Sure. Good morning, and glad to be here. So I'm Sid Haksar based in Boston, I lead Strategy and Partnerships for our Construction business. I've been with the company about 11 years now. I joined Autodesk in the mergers and acquisitions group. And then the last 4 years have really gone into the business. So really been at the start of our journey in Construction to where it is today.

Great. And Mike, how about yourself? Give us a quick overview.

Sure. I'm Mike Haley. So I'm, as Tyler said, the Senior Vice President of Research. So I run the Autodesk Research organization, so around about a 300-person international research group with one group in Autodesk that's our mission is to stay sort of 5 to 10 years ahead of the company and the industries we serve and just figure out and kind of prepare the company for what's coming next. As part of that, several years ago, we started our AI Lab. And as a result of that, and we'll talk a bit more about this later, I'm sure, I also direct the company's overall AI work and efforts across the company. I've been a little longer than Sid, 25 years at this year at Autodesk. So had a long career, really focusing on emerging technologies. It's sort of been the theme kind of what I've directed and led across Autodesk and prior companies before Autodesk.

Awesome. Well, great to have you here, and I'm sure a lot of exciting things to talk about. Maybe we could kick off on the AI front because that's the topic of 2025 and certainly been a focus at the conference. So just at a very high level, Mike, how would you articulate Autodesk's AI strategy from here?

So across the entire Design and Make space, it doesn't matter what industry you're in, whether you're in Construction, Manufacturing or Media and Entertainment, there is an enormous amount of laborious sort of tedious work that goes into most things, designing buildings, games, whatever it might be. So there is a massive opportunity to provide automation to those customers to start realizing efficiencies that they've just never been able to realize with traditional software. That's the first thing. Second thing is workflows and the way the data flows with inside these large complex construction projects, for example, are a great example. Sid works a lot with that. There's an opportunity to start accelerating those, making them more approachable, make them more customizable. So that's another one. The third one I would say, which is -- it's one of the ones that I find out, to be honest, the most exciting is design software -- Design and Make software is complex. It's complicated software. Most people that use it find it difficult to learn. It takes a while. And the result is there's probably a lot of people in the world that don't actually get to use the software because it's just so complex. With AI, the barrier to being able to use complex software is being removed. If you can go in and you can sketch a picture or write a prompt or do these kinds of things, suddenly, you're able to use sophisticated software. So it's almost changing the experience paradigm. So those 3 concepts are really at the core of sort of our strategy across the company.

And just to press on that last point, that means there's a lot more potential customers.

Yes, yes. Got it. Okay. And maybe talk about where Autodesk is in terms of product perspective in terms of giving that automation. What have you seen in terms of early traction from -- I know it's early, but...

Yes, yes. Well, let me give a little history before that and then I'll get to the traction bit. So one of the things we -- so I mentioned the AI Lab. So in 2018, we established our AI Lab. And the reason we established an AI Lab is not to go and trying to do what Open AI is doing or what Google is doing. We recognize that the nature of AI in our space is fundamentally different. The data is different. It's 3D geometry of the world. It's physics. It's how buildings are created or products are manufactured in a factory. You cannot take a language model and just supply it to that and hope to get useful results. And what we discovered back in 2018 is that there was not enough places doing that elemental work. There were not universities. There were not labs. There were not companies doing that work. So we had to set up our own capability. So that's -- a lot of what we've been focused now on for the last 7, 8 years is really building a level of competency around AI in the sort of unique space, right, of sort of data and these kind of workflows, right? So what you're beginning to see from us, and you saw this -- if you watched last year at Autodesk University, we announced a feature in Fusion called AutoConstrain. I'll go into the details of that later, but it's a thing that just helps you in your design process. But it's an entirely AI-driven using these modern sort of AI technologies. And what's fascinating is we launched that feature -- we announced it and then we launched it in February. So we're at 8, 9 months into sort of launching that. And we've just seen complete takeoff of people just loving this feature. Not only that, we've actually -- it's the first of the really significant AI features. So we're improving this thing constantly. We're getting the feedback now of how customers are using it, what they like, what they don't like about it, when is the AI getting it right, when is the AI getting it wrong, right? And you're bringing that information back in and you're self-improving. And we've just seen radical improvements in the quality of the tool itself, right? So this is the one of the main differences with AI tools. It's not like traditional software where you build a feature, you put that feature out there. And it's the same feature until you roll out the next piece of software. With AI, that thing is constantly getting better. So we've -- I mean, just in the last 8 months, I think we've rolled 3 different versions of that AI already, and it's -- the improvement is radical. So that is then further accelerating customer acceptance of the feature. So we're seeing something like 80% acceptance of AI predictions within that tool. So when it comes and says, "I think you probably want this." 80% of the time, they're going, yes, that looks pretty good. I'm good, which for a tool that's only been out for 7 months and it's our first AI is pretty incredible.

Right. So I guess, it sounds like you're seeing a lot of uptake initial of...

We're seeing a lot of interest, a lot of uptake. And one of the things I always look for with AI is -- I mean, AI is a little bit of a game, right? I mean because these are probabilistic systems. They will make errors at times. We all know that when we use ChatGPT and these kind of things, right? So what you've got to understand is what is the customer acceptance of those errors because there's a balance between how much efficiency is this giving it, giving me in my work and how many errors am I getting and what's the sort of trade-off. So you always got to figure out like what quality do I need to get the AI at in order to hit that sort of minimum threshold of sort of usefulness and what is the expectations of the customer. And you can't go and ask a customer that question. The only way to find this out is to actually build the thing and sort of iterate and experiment. So that is why I'm very excited about it. And what we're seeing now is actually a pretty high acceptance rate because they're getting such a big advantage.

Right. And when customers use the customers' uptake and you see the usage of AI, what is that -- what have you seen in those customers? Are they spending more time in Autodesk? Are they using more consumption? Is there any kind of downstream impact that you're seeing?

Yes. So we're definitely seeing more consumption. So the result is -- so I mean, all of this stuff is all AIs that we are providing now as services. So I mean, this is clearly -- I mean we'll probably end up talking about this later, but I mean, this is directly taking us on to our path of a consumption business. I mean we have consumptive parts of our business today. But relatively speaking, they're not as large as they're going to be in the future. And again, like I said, so what we are seeing -- so I mean, for example, in the average constraint sort of thing, you might find somebody spends maybe a week setting up a bunch of complicated constraints for a complicated design. So with this tool, that's now an hour, right? So when you see that sort of type of acceleration, I mean, the fact is what people start doing is they will actually do more because now they can make more products. They can make more versions of a product. They can try more things out. So we end up almost -- as a company, we end up winning on both sides, right? They're actually using our tools more to do more things, but they're also using a consumption kind of model from us.

Let me just jump in that. So I find you sort of think about AI as a continuum, not a fixed point. So you start at the sort of smaller model end where you're essentially doing feature automation. That's what AutoConstrain is. And one way of verbalizing that is that it's -- essentially it's a small model, which means it's a relatively low compute, which can happen on the desktop. And what that means is that you can deliver that functionality through the product and monetize it like a traditional subscription product, add value, capture it through price. And that's what we're doing with AutoConstrain basically. As you move on to workflow automation, so a particular workflow, you're going on to a bigger model and that compute will likely migrate to the cloud. And you can't put high-compute cloud -- variable cost, cloud compute through a subscription price point. Otherwise, you blow a hole in your margin. So that model will start migrating towards a consumption. As you know, we already have a significant consumption business with EBA and Flex. Those 2 workflows will likely remain human-led for the foreseeable future. But then when you go onto the next automation, which is system automation, where you're essentially doing cross-workflow automation, you get to a point where the complexity and size is beyond the scope of a human brain, and you will start bringing in significant amounts of machine consumption of the data as well. So -- and that will also be a consumption model. So you have that sort of continuum based on the size of compute and also on who is leading the consumption, a human or a machine, and also then the frequency. So because humans have to sleep, machines don't. So machines will work 24 hours a day, 365 days a year. So those are sort of some of the axes that you can think of with AI.

Right. And how should we think about Autodesk's road map and what customers need to do to kind of progress along that continuum?

Yes. So I mean, this is -- change doesn't happen quickly in most of our industries. So there has to be a bridge, right? We can't throw the switch and absolutely every piece of software suddenly becomes a magical AI thing because, well, first off, the technology is not ready yet for that, but nor are our customers. So the nice thing of starting with the sort of all kind of first stage, the small AI, like what Simon is talking about, the sort of automations that I've been talking about, too, those are not fundamentally changing the work. Like even this AutoConstrain thing, I meant, you're still constraining a drawing, you're still making a drawing. The overall process is still the same. You've just kind of shrunk the whole thing and you've taken out a lot of tedious work. So our strategy has been to really start to introduce those kind of tools all across our portfolio. That begins to get our customers used to using AI. It starts to build trust in AI as well because let's face it, AI is a new thing. There's a lot of things going on in the world where people are, "If I'm going to trust this thing, what does it mean to use AI?" So now they're starting to use it, starting to use it in these narrow contexts that don't overall -- now what you will see us do, again, like Simon said, as we move more to the system side and start thinking of workflow and then systems, you will see us gradually begin to expand the aperture, right? And the nature of those tools will become more transformative on people's workflows. But that's still to come. We've got to create a bridge from where we are today of traditional software.

Mike, perhaps as we were talking about last night at dinner, give us sort of things that you might be able to come in the future of how you input stuff and how it works.

Yes, yes, yes. So I mean again, like I said, I mean, the idea of how people use software is really significant. And I mean, I think what you're going to see in the future is abilities for people to -- well, first off, leveraging your historical data, I mean, this has been a classic problem in manufacturing and construction and actually in media and entertainment as well for that matter. However, everybody starts a new project and all the historical information is archived away somewhere and it's just not used. With AI, you actually have the ability to constantly mine all of that data. So you now start a new project. It's mined all your previous budgeting as well, that looks a heck of a lot like the building you built 7 years ago. Do you want to bring in a bunch of the material information, cost estimates, subs you use, bring it all in, right? So I mean, that's one example of the sort of transformative effect of how they're sort of being used. But also, like I said, on the user interface side, the way people actually physically interact with the software, the way teams are brought together is going to fundamentally change. So the ability to just sit down and use natural language with your software, have a user interface that is dynamically produced, that is giving you exactly the features you need for the job you're doing today. You don't have to go and spend a month learning Maya, which is an incredibly complicated software package to do a very simple little animation, which you could knock together in a few hours. So again, it's fundamentally changing the accessibility of the software, which, to Simon's point, actually, I think, is very exciting because that actually expands our market as well. So there's again -- so that's why I say there has to be a bridge because you can't go straight to this future vision of like it's with bang, all the software looks great. It's going to be these new kind of interaction paradigms. You've got to find a path of value for our customers to get there.

Mike, if I may just add with Construction, we kind of break it down into a very simple. Our business is very, very simple. And you've got to talk to the end user in very simple ways. So we think of it really 3 ways: augment, automate and analyze, right? So when you talk about automate, this is the agents that we're building to do very mundane manual tasks. When you think about augment, this is where we can use generative AI capabilities, whether it is drafting up an e-mail or having a spec assistant read something and give you all the information that you need. And then you can review that versus you having to look at a specification manual and extract the information. And then the final piece, which is analyze, which is really around predictive analytics, to what Mike said, we generate so much information. So how can we start to drive correlation with certain events that are happening. So I think AI just unlocks this opportunity for us to really ultimately drive productivity out in the field, but then also mitigate a ton of risk in our business.

Yes. And I'd love to sort of bring in the competition angle, too, as it relates to AI. I mean, certainly, a lot of software companies are under pressure just around some of the existential and disruptive concerns around AI. I think vertical software, design software, particularly where the Autodesk of the world feels a lot more insulated from that. But I mean, you talked about kind of this idea of simplifying the user experience and gaining more users. What makes you confident that, that simplification isn't going to invite more competition over time and that you can hold on to what is a really dominant competitive position?

Yes, yes. I'm sure there's going to be more competition over time. But I actually think the competition is probably going to come from different places. I think our traditional competitors, what we're seeing in the work that we're doing, so the lab that I mentioned that we established. I mean we are the world's -- I can tell you right now, we are the world's leading publisher of scientific information in the space of AI for design. I mean we have published hundreds of papers in this area, which just shows out our leadership. If you go to the average conference, we're just all over the place. That has given us a technology advantage, which none of our competitors, frankly, right now seem to have. Our position in the market as well gives us the ability to not just develop that technology, but now to bring it to our customers to start that loop that we talked about before, sort of self-improvement, understanding how customers use it. And then, of course, you've got to have data to do this. Our customers entrust us with their data. So these systems are built using their data for them, right? So I mean, there's so many bits that are difficult to replicate. If you want to make AI at scale in the industries we are at, if you don't have data, if you don't have customers and you don't have the skills of how to build that unique type of AI, it's very difficult to enter the space. So right now, we've got a lot of start-ups. There's a lot of start-ups doing kind of interesting things. They have some skills, they don't have data and they don't have customers. Very, very, very difficult for them to get to any sort of meaningful kind of scale kind of quickly, right? Two, you have our traditional competitors out there. Most of them have been pretty slow off the market sort of building the kind of skills. We're beginning to see them do it now. They're now beginning to build up AI Labs and things. But we've been at this now for 8 years. So we've got quite an advantage over a lot of them. And then finally, you get the big players who get the hyperscalers out there. And for them, I think you were sort of alluding to this, Tyler, I mean, this is the nature of the vertical business. It's -- there's some sort of safety in being a vertical business. But I will tell you also, over the years, I've done a lot of work with the hyperscalers. Their understanding of our industries is negligible because their entire play is to be a platform. They always want to make everything incredibly horizontal. But by definition, we're a vertical. Our customers are looking for specific solutions to specific problems. So it's very hard for them to actually understand all the nuances. And by the way, I will just tell you, the nature of data in the design space is so much more complex than almost any other industry out there. The nature of the data that makes up this building we're sitting in right now is incredible. So just the ability to actually grok that and understand it and be able to build products around that is a really unique capability that typically only sits within companies like ours.

Mike, how long do you think it would take a competitor to get to where we are today?

I think if you had a well-funded competitor that had all the bits and pieces, the data and everything, it would still take them probably 4, 5 years to get to where we are right now.

And then talk about the rate of change as well that we're achieving now.

Yes. So what's kind of interesting, with AI, there's an entirely new platform and approach to this, right? So it's not that we just all of a sudden built AI and it's all the same people doing the same thing that they were doing before. It's just AI stuff. With AI, the entire product life cycle, the entire platform technology, everything that's behind the scenes that's going into building this is different. So not only are we building AIs right now, we're building different infrastructure. We're building different operating models and processes, and that's what we're refining. And that's -- this is sort of like the tip of the iceberg kind of thing, right? What most people are seeing is that tip. What they're not seeing is the large iceberg under the water because what that is, this is the accelerated computing initiatives that we've put in place. We've got a compute infrastructure, for example. So we run on Amazon, whereas in Amazon, we largely put all of our stuff in the Amazon Web Services Cloud. Amazon offers their own AI layers, their own machine learning stuff. We've built our own on top of them. We don't use Amazon stuff because our stuff is 70% more efficient than theirs. But we figured that out. We figured that out over the last 4 years because we want to do this stuff at massive scale, right? We figured out our own development practices. We have model operations groups that, like I said, with my example around AutoConstrain, are constantly working to improve the models, figure out how our customers responding to that thing, how do we make the model better. So these -- it's kind of the non-sexy stuff, if you want, sort of behind the scenes, but this is the stuff that really counts. This is what makes you good at doing these things. And this is the -- to my answer to you, Simon, this is what really creates the difference. It's not just that we know some magical formula to how to do this. We've done the work. We've hit all the bumps along the road. We figured out what it needed to be smoothed over.