ADSK - Autodesk Inc. (ADSK) Presents at Rosenblatt Securities 3rd Annual Technology Summit (Transcript)

2023-06-07 17:50:25 ET

Autodesk, Inc. (ADSK)

Rosenblatt Securities 3rd Annual Technology Summit

June 7, 2023, 02:00 PM ET

Company Participants

Dan Arlan - IR

Steve Hooper - VP and General Manager of Manufacturing

Conference Call Participants

Blair Abernethy - Rosenblatt Securities

Presentation

Blair Abernethy

Good afternoon, everyone. And thank you for joining us. My name is Blair Abernathy, I'm a software analyst here with Rosenblatt Securities. And this afternoon we have Autodesk presenting. From the company, we have Steve Hooper, who's the Vice President and General Manager of Manufacturing --Manufacturing Solutions, and Dan Arlan, from the Investor Relations side of things. And Dan is going to read a quick, Safe Harbor statement. Dan?

Dan Arlan

Thanks. Keep it short. So 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 the statements. Thanks.

Question-and-Answer Session

Q - Blair Abernethy

Great. Thanks, Dan. So Steve, thanks for joining us. Great to have you. And maybe just for the audience, just to kind of level set things, if you could just give us an overview of what you do for Autodesk. And really just the sort of a year on the manufacturing side and talk about a little bit about your overall product set and vision for this division of Autodesk. I think you're on mute.

Steve Hooper

Yes, I am off to a great start. I think I have heard that by now. But okay, yes, sure, no problem, as most of you I'm sure are familiar, Autodesk operates in a number of verticals too. AC for the built environment, EMS for our media and entertainment solutions. And then I belong to the product design and manufacturing solutions group. My role within that group is that I manage and lead the software definition, engineering groups responsible for all of our authoring applications.

So you can think of these as data creation tools that are used across the product lifecycle process. So everything from concept ideation through to detailed engineering, whether that's mechanical, or electrical, or electronic, across into the simulation portfolio, and then manufacturing. So of course, you can imagine there's a range of solutions across that spectrum that include things like 3D modeling tools, like inventor, our flagship product for professional engineers, all the way across into the manufacturing environment with tools like Netfabb, that are used in additive production processes, FeatureCAM, PowerMill, which are used in the subtractive side of our business. And of course, there's many, many other products in that portfolio.

I would say the other, you know, large part of my responsibility is to help define our industry cloud solution. So, across the company, in general, we're pursuing a cloud platform strategy. So we have Autodesk Platform Services that formed the basis of those solutions. And then we have three vertical industry clouds that sit above those. So former in the construction and built environment industry, then we also have flow in our EMS group, and then Fusion, which is the one I'm responsible for in the manufacturing group.

So over the last, I think, eight years, I would say we've been busy building out our cloud platform solutions, Fusion, really at the tip of the spear in that effort. And part of the reason that all of the data creation tools are within my group is because there's a large amount of product expertise, along with technical IP that we've been refactoring as part of that vertical industry cloud for our design and manufacturing customers. So the first part of that process was obviously to unify the capabilities, the tools the engineers use as part of the process when they're bringing a product to market.

And then the second part of that process was obviously unifying all of the data in the cloud to make sure that we were able to offer some of the transformational opportunities that the cloud presents in terms of workflow automation and productivity. So I guess if you wanted to summarize my role, it's to lead the vision and execution of our industry cloud for manufacturing, as well as helping them make sure that there's a smooth transition for our existing customers. Because we have a pretty substantial base on tools like Inventor, PowerMill, Netfabb and Moldflow.

We want to make sure that they enjoy all the benefits of the cloud, as well as a robust set of offerings that they're using in their existing engagements with Autodesk.

Blair Abernethy

That's great, Steve. And maybe just before we shift over to talk about AI, maybe talk a little bit more about this, the Autodesk Platform Services, just so that people really understand. Is this a product that you're selling out there? Or is it something that they're going to get as they move to the clouds just kind of frame up, how to customer’s sort of get their hands on, how they move from the woods largely been a traditional on-prem base to Netfabb?

Steve Hooper

Yes. Well I mean, it's a complex question -- sounds a simple question, but the complex answer, I guess, there's a little of both. So at the core, the platform services, so Autodesk Platform Services are a series of components, software components that are built on our cloud platform. And so those software components can be used by developers, both our internal development organization and external developers to build Bespoke solutions.

So you can think of fusion former inflows almost like reference applications. Those industry clouds are built with the components that are part of Autodesk platform services. So a customer could and many do use those Autodesk Platform Services to construct their own tailored offerings that they use within their business, or offer to their clients. So in that respect, they are a customer of the platform services as well.

And then, of course, you've got the industry clouds, which are the vertical offerings built on top of those platform services. So from that perspective, like I say, Fusion is the offering that would serve the design and manufacturing market. And you can think of that as like a verticalized extension, industry cloud that sits on top of the platform services, specifically for manufacturing.

But again, when we're building Fusion, we're using core common components. So that if somebody in the manufacturing environment wished to use some of those common components to build a bespoke solution, they could. Now the powerful thing about this strategy is, internally, there's a lot of architectural and component reuse, which is obviously good from a cost basis for the company. But really, the biggest benefit is that we span industries. So not only do we see a convergence between design and make, we also see a convergence between industries.

So if you think about some classical factory owner and operators, they have to manage a building asset. But they also have to manage, simulate, implement, and run or operate the factory lines that sit inside that building. And there's virtually no other vendors out there in the market that have a complete range of solutions that cover that span of industries. What we're able to do in the industry cloud is offer them a complete data model.

So not separate vertical applications that sit on your desktop. And, you know, struggle to connect information together in some cases, but a fully unified stack of components services that run in the cloud. Obviously, if you're using Fusion in combination with something like former, then you have the basis to build a solution, which serves some of these industry segments that span those types of capabilities.

So there's a lot of opportunities there. If you think about somebody like Disney, for example, they span everything from entertainment media services, to theme park construction to the products that go in those theme parks. So having a single set of services that run on a common platform is a real benefit for customers like them.

Blair Abernethy

And how would you tie in inventor to the platform?

Steve Hooper

There's a couple of ways that I would tie that in. Firstly, there's a lot of expertise and core component capabilities, that inventor is able to contribute to our platform offering. So we're able to refactor some of those tools and services. So for example, we have a design automation service that allows people to tap into Autodesk Platform Services, and manipulate modify configure 3d models in the mechanical engineering environment. And that is all based on top of Inventor.

And from an end users point of view, we make sure that any end user with any one of our existing desktop vertical applications also gets access to our industry cloud. So anyone that subscribes to the product design or manufacturing collection, primarily using perhaps Inventor and AutoCAD also gets access to Fusion as part of their offering.

And we ensure that there's a smooth connection, so the data backbone, on which Fusion and Autodesk platform services are built allows you to consume and manage information from Inventor, so if you want to leverage some new forms of technology like generative design, for example, and you're an Inventor user, you can smoothly connect to Fusion. You can bring information in from the inventor environment, use it as the basis for generative design and then pull the information back into Inventor.

Equally, if you're thinking about downstream manufacturing applications, an Inventor user can be designing an inventor with a production engineer later in the process, consuming that Inventor information building machining strategies for production and if the Inventor model changes, obviously, Fusion, the industry cloud maintains that connectivity and updates the data. So you get a smooth contiguous flow of information from our traditional desktop applications through to our industry cloud.

The point I'd make here is, about 20 years ago, the industry was going through a big shift from 2d to 3d. Many vendors that were out there, kind of forced customers through a transition point, where they had to kind of flip the bid, they either had to stay on 2d or move to 3d and migrate all their data almost overnight.

One thing I think we were particularly successful with was not forcing customers through that transition, but providing them a smooth on ramp that kind of reduced the switching costs for them. And that's something we focused on heavily in the transition to cloud because it is another disruptive transition for customers. And we've made sure that existing Autodesk clients have that kind of smooth transition opportunity. So they can start to embrace the cloud, enjoy the benefits of it at a pace that suits them without forcing them through a disruptive change, but still offering some of the productivity benefits that are available there.

Blair Abernethy

Interesting. That's pretty a significant change at again, and yet, but you've been through it, so how to help the customer’s layout the vision and get them there, which can take can take time.

We switch gears a little bit and bringing the topic of AI and let's talk about just sort of where Autodesk is on its journey. What have you been doing in the last couple of years in this area? And maybe we'll start to talk about sort of where you see the products going?

Steve Hooper

Yes, sure. So I think you kind of have to take a step back maybe eight or nine years to when we made our transition to the cloud. So I think at the time, many of our competitors, and even some of our customers thought we were crazy, you know, there was no way people were going to put design assets in the cloud. Here we are today and it's pretty much the industry standard, everyone is busy going through some form of digital transformation. And it usually includes leveraging cloud offerings.

Now, I would say, I think a large portion of the market is mistaken, the disruptive benefits as just purely a move to cloud. And I think the move to cloud is a means to an end, it's not actually the endpoint that delivers the benefits that customers can really expect in the future. So maybe I'll just unpack that a little bit.

The first step for us in our cloud transition was to unify the existing portfolio, if you think of most customer’s business process, there are points of friction, which cause a loss of productivity, and also curb their ability to explore innovation in the market. And some of that is induced by using these kinds of point products that are all proprietary, makes it very difficult to exchange data and information between them. And then there's also a cognitive load for engineers that are moving between applications and products. And many of us experience this in different industries.

And so the first kind of step in moving to the cloud is to unify the portfolio, because if you build a unified stack of tools, that allows engineers to move smoothly through the process, you accomplish two things. Firstly, you reduce the cognitive load. Secondly, you allow data to move smoothly, because it's an open data model.

And that means you can process change bidirectionally, because typically, you need to make a change, you have to open up your design application, make the change to the proprietary data, save it, move across to your manufacturing application, re import the data, update it, save it. And so it induces this kind of waterfall process, which is very ineffective, makes it difficult to manage change, it's slow and time consuming.

If you can move to more of an open data format, and you can think of it as like a hub and spoke the hub is the data, the spokes are all the applications as you move around the digital thread. If you can move to that type of model, obviously, it makes it easier to make the transition as you move through the manufacturing process. That it also means you can process change bidirectionally, you can make a change in the manufacturing process and have that propagate back to design. And that allows engineers to move to more of an agile environment.

So I'd say that's the first step toward moving to the cloud. And it's certainly a benefit in terms of workflow productivity. But I think if you talked in terms of percentage point productivity gain, you may be talking anywhere between sort of 10% to 15%, which is significant, but often not enough for a manufacturer to overcome switching costs to move from whatever applications they're currently using.

So the second benefit, and I think you want to think of these as layers that build up towards the ultimate goal. So the second layer of the strategy is to unify all the data. So once you've unified the tools, so you've created a pipeline for the process, the next thing you want to do is bring the data together in an open data model. And so we've developed what we call a manufacturing data model that is completely open extensible. So there's two things about that that are different the open pieces, there are no licensed SDKs, any customer or any third-party developer can connect to that open data model, and use it to build their own third party applications.

The second bid, is that it's extensible. So our data model covers manufacturing process. So CAM tool pathing additive, it includes simulation, it includes electronics design, as well as mechanical design. But if you're a third-party, and you're building some other application that serves another piece of the overall manufacturing workflow, you can create new data types and our object model. So that makes it extensible and open. That creates an ecosystem with which our partners can build on and contribute to this.

So, first thing is converged the solutions into a contiguous pipeline. Second thing is, bring all the data together in the cloud. Now, once you've done those two things, you've delivered some benefit productivity, you've also made it easier for people to collaborate because you've centralized the data. And obviously, through COVID, that was of a particular benefit. Now we're seeing supply chain disruptions. Of course, centralized cloud data means you can move operations globally with much more ease. And they are big benefits, but they pale into comparison with the last one, which is automation.

So once you've bought all the client services and the data and centralize them on the cloud, you now no longer need the client desktop to process change, you can start to process change on behalf of the user without them having to be present. Now, that is the key to some of these generative AI opportunities.

So you asked where we're at today, we've explored this extensively already today. So things like generative design, they allow us to process designs in parallel at scale. So we can optimize the design definition, based on a set of criteria, and deliver a customer not one optimized solution, but maybe three or 400 choices that allow them to make tradeoffs between material, mechanical properties, cost manufacturing time, go to market speed.

So that is the type of benefit you can deliver in terms of automation, once you've moved to the cloud, and that doesn't deliver like 10% to 15% productivity gain. In many cases, that can be the difference between 400% or 500% in terms of productivity gain, and that's the type of benefit that will entice a customer to make the switch from what they're currently using today.

Blair Abernethy

Steve, just maybe drill into that a little bit more the customers that -- I mean, this is your core of your generative solutions, really, in Fusion and in Inventor, is that right?

Steve Hooper

No, core of our generative solutions are all based on Fusion. So fusion…

Blair Abernethy

Cloud solution.

Steve Hooper

Yes, because like I say, you need to be able to move not just the 3d geometry from mechanical design and centralize it in the cloud, you need the whole process to be able to generatively design a product, it isn't good enough just to produce a piece of geometry, you need to be able to evaluate that geometry with simulation on the back end.

And you also need to be able to evaluate its manufacturability on the back end, which is why I say you have to connect design, simulation and manufacturing processes together. If you don't build that pipeline, you're unable to do generative design, because all you can do is just build some geometry, you don't know how performant it is, and you don't know how manufacturable it is.

So that's why the first layer of advantage is unification of the actual tool chain. The second layer of advantage is bringing the data together. So you need all the data together to be able to process it.

Now, I'm pretty sure we're probably going to get onto the topic of generative AI. So before I touch that, we've been employing AI for a long time. So, I talked about generative design, it can create 500 results. Well, how do you make sense of 500 results, you don't want to spend all day looking at different potential opportunities. So we use AI to cluster results together look at common results that customers often work with all chooses a solution so we can help present options to customers leveraging AI as a learning model.

So that's one example. On the back side, on the back office side, we actually make significant cost savings because we employ AI to optimize what we call spot instancing. So when we're leveraging AWS assets for compute, there are certain tiers of asset utilization that are more cost effective than others. And we can use AI basically to load Balance what we use on the back end services. So there are many applications for AI that you see in the product that are hidden. But that's a little different to generative AI.

Now, with generative AI, you've got to train a model. And in order to train models, you need large datasets. And in order to get large datasets, you need to be able to move information to the cloud. And so that's why I say the second layer of advantages, data centralization in the cloud so that you can start to deliver those benefits to customers.

Blair Abernethy

Is the -- is your traditional Inventor install base then are they -- is this helping to pull them to the Fusion platform, the ability to do generative there?

Steve Hooper

Yes, absolutely. I think there's a number of like, benefits, what we don't want to do is force a customer to use another solution, this migration is a mistake. What we're trying to do is build benefits that we can deliver only via the cloud, because they're significant, like the few that we've just discussed already, and there's others we'll get into. Once those benefits are in place, we want to make sure that an Inventor user can make use of them without being forced to switch overnight. We hope that what they will see is enough benefit and productivity that over time, they'll make that transition decision on their own.

So what we need to do in order to make that happen is build these very smooth workflow connection points between Inventor and Fusion. So today, if you use Inventor, there are actually capabilities and inventor that connects you directly to Fusion, you don't have to go through the work of migrating data or moving it around, we just connect you straight to fusion.

So for example, if I want to machine a part inside of inventor, I can choose open the part, choose manufacturing Fusion, it will open Fusion, ensure that the data, the design data is there ready and waiting for me to start machining it. Obviously, there are things in Fusion like plastic injection molding simulation, there's discrete event simulation.

All the sorts of forms of generative design that we've discussed that there's a number of key benefits in the industry cloud, that you don't get specifically in the desktop applications, not because we don't want to put them there, we're not hiding functionality. But we can't deliver them on a desktop solution, because we need to have that cloud compute capability to implement them.

Blair Abernethy

Is there just open up a different angle on the question here. So, I see how you build it from Inventor on an on-prem to Fusion. What about competitors' models? What if I'm building something in SolidWorks, or some other competitive CAD system? Can I access Fusion, is do I have an avenue in there?

Steve Hooper

Yes, absolutely. So SolidWorks data, for example, will read SolidWorks data in associatively. So you could pull some SolidWorks information and you could build a machining strategy. And when I when I say a machining strategy, I mean subtractive milling machines typically take the model, and then you tell the toolpath, how to cut the model. So you could do that on SolidWorks data and the SolidWorks data changes, it'll update inside of Fusion and your machining strategy will update. So we've had this very intentional land and expand strategy with Fusion.

So what we've trying to do is build beachheads, and there are a number of them. So there's machine tool simulation that enables you to take machining strategy from something like Mastercam, for example, and then validate and verify the G-code before you cut it on the machine to avoid any costly mistakes.

We've introduced full multi axis cam capabilities, we've introduced fabrication capabilities for sheet metal nesting, and cutting. We've introduced a whole raft of simulation tools, generative design, we've also introduced a lot of additive design tools, because designing for an additive process is different to a traditional subtractive process.

So you can think about maybe, and that's just the data creation ones. There are also data management tools like our PLM offerings, our manufacturing execution tools. So all of these form benefits that take for example, a SolidWorks user, the SolidWorks user carry on using SolidWorks and add Fusion and get more benefit more than enough benefit to justify the initial investment on Fusion.

So our hope is that somebody who's using SolidWorks doesn't feel like they're being forced off what they use, they will choose Inventor to complement it. And over time, they'll see that Inventor offers more and more capabilities in these different beachheads including mechanical design, and I suspect for many people, they will figure out that it really doesn't make sense to pay maintenance contracts on older software when they can connect the whole process together in fusion.

So it's a way of being able to complement what's out there doing what's right for the customer from the get go, not forcing them through a transition. And for people working in heterogeneous organizations where lots of different tools are being used, we can't force somebody to move to a cloud platform overnight anyway, they wouldn't want to. So it's a very kind of customer friendly way of introducing them to the cloud.

Blair Abernethy

Very interesting.

Steve Hooper

It also means that we're not solely reliant on competitive displacement for market growth, of course, top of mind, but we don't want to be solely dependent on that.

Blair Abernethy

This is an add-on solution for customers that can help speed their path to CAM and help speed the path to doing generative work.

Steve Hooper

It is, but make no mistake, we are intensely focused on winning in the 3d mechanical design. We just want to do it with the customer's best interests at heart. And I genuinely mean that that's why no proprietary data formats that are inaccessible, no charged SDK kits for development is completely open cloud based system with the customer's best interests at heart. But absolutely, our desire is to, to continue to increase our market share, which can only come at the expense of competitors like SolidWorks.

Blair Abernethy

Yes, interesting. So Steve, let's shift a little bit to some of the -- lot of noise and excitement in the market around large language models, and so forth. And can you give us your sense of how Autodesk is looking at this, and what it could do for you potentially, in the manufacturing solution side of things?

Steve Hooper

There is so much to unpack here, and I know we've got limited time, so you'll need to kind of constrain me, otherwise, I'll get carried away. I would say large language models to begin with are only half if that have the potential opportunity. So I'll start with the large language models, the large language models offer a very intuitive way to interpret user interface to interpret user input. So you know, traditional manufacturing tools, and you're an engineer yourself, they can be quite difficult to learn.

And they can be quite constraining in the way you think you've got all these amazing ideas in your head as an engineer, and you have this software that's capable of representing them that in between your mind, and that representation of your idea is a very constrained bottleneck, which is a keyboard and a mouse. And that makes it pretty difficult, especially as you've got to often spend six months learning the language of this 3d modeling tool or package, right.

So you already know a language, it's the language you were brought up with. It's the native tongue that you learned in whichever country you were born. So these loud, large language models, they offer the opportunity to expand the adoption of tools and reduce the barrier to entry. So for most customers, they have to go through like a four-year degree program, or a vocational program of the same length to gain the industry expertise. And then on top of that, they have to learn this sophisticated secondary language of the computer aided design tool.

I think large language models are part of the solution in knocking down that barrier. So I'll also say that with these generative AI tools, everyone's worried it's going to put everyone out of work. Trust me, it's not going to put people out of work. What it's going to do is democratize the industry.

Now I know it, I've seen it happen in other industries, I've seen parallels. So what it will do is it will empower more people to enter the manufacturing market, and it will probably decentralize it. So you might see less consolidated centers of excellence. And you might see more distributed startups, you're going to see a lot of people enter the manufacturing market that couldn't before because of some of these barriers to entry. These large language models are part of the solution and pulling down those barriers to entry, because it makes it very intuitive to start using technology.

It makes it intuitive on the number of points. I think the first one would be data entry. So you build a prompt, it's in natural language, it makes it very easy for the system to interpret that prompt and give you a response. It also makes it easy for you to learn. So if you've got questions about how the application works, you can ask natural language questions and you're not worried about a forum where you've got to wait for somebody to respond or product support.

So it can augment product support and training to a massive degree. And then you can also create macros for you. So quite a big application for large language models is actually in software development. So you can use a large language model to actually create scripts for you one of the reasons the AutoCAD became popular at the beginning that its inception was a programming language that it used called Auto LISP, which is a form of LISP.

Now, with a natural language, with the LMM, you can basically build scripts through a natural language prompt. So again, that brings down the barrier to productivity for many customers. So those large language models, they offer an opportunity to kind of knock down the barriers to interfacing with the computing system on many different levels. Now, I say…

Blair Abernethy

It's interesting, the script issue, you could you not even run these against your existing customized flows and find problems, find errors?

Steve Hooper

Absolutely. I mean, the people use it even when we use it, people use it for things like C++, all the time, it doesn't matter what computing language you're working in, most of these large language models now can interrogate programming language, as a series of classes that perhaps you've written, and they can actually look for optimizations.

But also, one thing that many developers are terrible at is actually commenting their code so people can maintain it. And again, a large language model can come in, interpret the classes, the methods, the variables, and actually comment in for you. So lots of productivity tools there.

Now the reason I say it's only half the answer is that we're not dealing with tax based output, we're dealing with pretty sophisticated product designs that are multidisciplinary, things like electronics, and 3d modeling and 2d documentation. So our belief is that you don't just need a large language model, you need what we would call a large product model. And this is pretty significant, because in order to build a large product model, you need to have data in the cloud on mass. So that is where we see this huge opportunity.

Those vendors, and there were very few of them, and were the leader, very few made the transition to the cloud early enough. There are very few products out there that have grown at the rate that Fusions grown out. And you've seen that through our earnings calls over the years. That's built a huge pool of data from students, commercial users, hobbyists. This is data that can be tapped into to build something like a large product model.

So I want you to think of a large product model very much like a large language model. So we take data and we prep it and structure it. And then what we do is we encode it, and I won't go into all the technicalities, but you can encode 3d data in the same way that you can encode a language in a language model.

Now, when you encode that data, then you can start to train models to look for patterns, which then creates some of those huge productivity gains that we've seen. So in the same way…

Blair Abernethy

Just for people to understand this, Steve, so you're saying, just to put it into a real life context. If I show the model 1500 examples of a drive train in an automobile. It begins to recognize the various components of a drive train, what works well, how it works, why it works, kind of thing that can help in my new design.

Steve Hooper

Yes, exactly. Now think about, there's a lot of non-value added tasks that happen when you're designing a 3d system. So in your mind, how that transmission is going to work, you know, how the gearing system works, you know, how the transmission shaft is going to work, you have to sit there and then create it using the language, which is sketch based, and you have to turn it in 3d models, then you have to assemble everything. And there are artificial things, artificial constructs that we've created and software to interpret how that product should behave. So you have to create joints between all these different components.

Now, you can learn from a data set, right that you can learn that a shaft is more than likely to go through the hole that has approximately the same diameter. And so you can start to automate some of the processes for customers. And I would think of it very simply as like, when you're using Microsoft Word today, you start typing a sentence, and it's intelligent enough to be able to infer how that sentence should probably end. And if you hit the tab key, it just finishes the sentence where it's like an autocomplete.

Think of the opportunities there for us to be able to do that in 3d modeling. There are so many non-value added tasks, and that's in the design phase. Once you've built a design, everything downstream is complete non-value add, it is just you having to prep the data for other people to interpret it. So you have to create a drawing so that you can hand the drawing to engineering in the shop floor for assembly so that people can interpret the design. That drawing is pretty much rules based on a set of international standards.

If you have enough data, you can build a large drawing model and then you can start to interpret 3d models and figure out how the drawing should be laid out and dimensioned on behalf of the customer. So, I spoke to a customer recently who make large structures, still work structures, they estimated about 72% of their engineering time and overhead was allocated to the documentation, not the design of those structures. And their view was that if they could automate just half of that, it will be enough for them to make a decision to switch applications overnight.

Another downstream application is machining, 3d printing. Now, it takes a production engineer hour to take a 3d model. And then in further machining strategies to actually produce that 3d model on the machining center. I want to give you the analogy of many of you probably taking notes right now, how would you feel if at the end of this call, and you taken all your notes, it was going to take you as much time again, to program the printer in your home office to print the notes that you just spent 45 minutes taking. That would be complete insanity.

And then on top of that, before you do the printing, you would have to teach your computer how to talk to the printer. Now that might sound completely stupid. That's exactly what manufacturers do today, they have to create what's called a post processor to tell it how to drive the machine tool to you and I call a printer driver. And then they have to interpret what they've built in terms of a 3d model as to how to print it. And that's why I say I've seen parallels like this before desktop publishing revolution, exactly the same thing.

It used to require all these different specialist applications and people to set up a large centralized infrastructure to amortize costs around mass production, desktop publishing came along, yes, it put a few people out of work initially, in the first couple of years. Ultimately, it created an order of magnitude more jobs, because so many people now entered the publishing industry is smaller startup or independent organizations.

And that is exactly what you'll see happen here in the manufacturing environment, you'll end up with the equivalent of what you see is what you get that we've already produced drivers for machining centers. So we can plug straight into a five axis milling machine, like a hose, for example, vary the controller, learn how to speak to it without the user being involved. And ultimately, we're currently working on taking a 3d model using one of these large models to be able to automatically produce the machining instructions, so that you can literally push it to production in seconds rather than days.

And so that's the 400% or 500% productivity gain that I'm talking about. And that's why these layers of advantage are important, converge the tool chain, create the pipeline, common language, centralize the data. And then the last layer of advantage, the important one, is automation through these generative learning algorithms, that allows you to deliver 400% or 500% productivity, and that's what will create the benefit customers are looking for to justify a switching cost.

Blair Abernethy

If they are saying Steve, the question is, on the large language model training, is it -- I guess I especially see two aspects of it. One being Autodesk training Fusion, to be better understanding software program. But the other would be if I'm a large company, like a large automotive company, I've got a lot of my own data on how I do things. If I'm a German car manufacturer, for example. Is there a way, or will you will there be a way for your large customers to also create their own bespoke models?

Steve Hooper

We've got to figure that out. But I think our stance at the moment is we are much more interested in -- and again, this is all kind of being we're working on figuring this out right now. But our stance right now is that we really want to be an Autodesk has always been good at this is the volume mass market. So what we don't want to do is build hundreds of exclusive independent large models that are proprietary. Remember, our strategy is open extensible to everyone.

And I think if you see the most successful technologies out there, that's the approach that they took. It wasn't exclusivity for the high end. Now, I'm not, I wouldn't rule that out, I'm sure there are creative solutions to solve that problem. And this is all relatively new. But I think our stance would be we want to build something for the mass market that really democratizes this technology.

The reason I say that, is take a look at ChatGPT, I know everyone tired of talking about ChatGPT. But the thing that it did really well is it didn't build a specialist high end solution for a select group of experts. And that's why it's the world's fastest growing software product in human history.

If they gone the other way and built something exclusive for a high end enterprise, it wouldn't have enjoyed that kind of explosive growth. And that's our view. That's where we see the opportunity. And if you think about Fusions philosophy from the beginning, it's been to democratizes technology across the industry, because we believe that scale will deliver the higher revenues.

And by the way, on the large language models, I talked about this large product model. Again, we don't believe it's exclusive, we think you basically what you would do is combine a large language model, there are people out there in the market with the expertise in that what we're seeking to do is build a competitive differentiation that sustainable around the large product model, because the data is unique to us, and it's in large volume.

So we combine these together, because what's unique in our industry is that we need precise output. If you look at something like ChatGPT text, or you look at something like mid journey for art. What you get as a result of your prompt is imprecise, it's subjective, it's interpretive by the large model. What we need in the engineering disciplines is precision output. Now, if you had to create precision output, you'd end up having to build a text based prompt that would take you longer to write than it would have done if you just modeled the thing in the first place. So what we believe is we'll need like a multimodal input. We'll need the ability to take a sketch, maybe a source model, some text description, some specifications from a catalog, input that, and from that we'll be able to interpret a precise output.

Blair Abernethy

Interesting. And I'm going to ask the question, when do you sort of see this stuff coming available, you don't have to -- asking for a confirm date. But just give us a sense of, how long is this going to take?

Steve Hooper

No problem. My teams are always like, it'll be done when it's done. No, it won't need a specific date. So I to put it in context. And what I love is I think probably it was Tesla that did this first was the kind of I can't remember exactly where the source was. But these kind of five steps to autonomous vehicles autonomous driving. And what I love is that we back chain from it. So you kind of saying at the beginning, we're going to start with things like adaptive cruise control and parallel park, then we'll be able to summon a car, then we'll be able to lane assist and autonomously drive on the freeway, and then ultimately, we'll get to city driving.

And I think it's through those sequential steps that will actually get to what I'm talking about, which is full autonomous definition of a complete multidisciplinary product. But to get there, you need to go through some gates. So the first set of gates are what I would call automation, productivity benefits. So we're expecting to be able to get prototypes of those up and running within the next six to 12 months.

So we're expecting to bring those to market in a way that delivers significant benefits today to people that are already using these applications, and their big benefits. So I also think there's a kind of culture shock for folks, you need to be able to augment existing tools with the new technology and gradually expand them, which is what I think's clever about these five steps to autonomy.

So beyond that, then we'd look at kind of multidisciplinary systems, then we'd start to get to things like generative machining, augmented design. And eventually you would get to this ability to create like a systems based definition of a product offering, that's its optimal place.

Again, I do not see that putting a single person out of work, I think it will create jobs right now bureau of labor statistics for the North America estimates by 2030, there'll be a gap of employment in manufacturing of 2.1 million unfilled jobs in America. So the argument that it might somehow put people out of work is ridiculous.

What it'll do is it will democratize it. As we onshore more manufacturing, it'll create the ability for more people to enter the market with less domain expertise. So that will create more opportunities. And if nothing else, it will help plug that gap. And that's the only way you can sustain GDP growth. There isn't the human population growth to do it otherwise.

Blair Abernethy

Yes. Excellent. Well, that's a great spot to end our 45-minute chat on, on Autodesk. Steve I really appreciate your insights and looks like you've given us a really clear vision of where the where the products and the business can go.

Steve Hooper

This is really exciting. There has been no bigger disruption in the manufacturing industry since the industrial revolution. And I think everyone can see it. So just remember, you had to get the data to the cloud, you had to unify the tool chain if you didn't do those two things, you can't play in this market.

Blair Abernethy

You can't get there. Yes, yes. Excellent. Okay, thank you. And thank you, Dan, for helping to set this up. Really appreciate it. Thanks. Take care. Bye.

For further details see:

Autodesk, Inc. (ADSK) Presents at Rosenblatt Securities 3rd Annual Technology Summit (Transcript)