Onshape CEO and CAD legend Jon Hirschtick talks about his approach to AI and how manufacturers can extract value.
The general sentiment around the usefulness of artificial intelligence (AI) has seen it’s ups and downs over the last couple of years. After bursting into public consciousness in late 2022, the hype has subsided. As we enter the final stretch of 2024, the current thinking is that AI is in bubble territory and companies should be wary of putting too much stock into its potential.
This is sound advice, no doubt. But as with any burgeoning tech, the early value is often found in the margins, helping companies gain an edge rather than bringing groundbreaking change. This holds true with AI—it won’t change society as we know it, at least not yet. But when applied to the right niche it could have significant impact.
The U.S. manufacturing sector, which was estimated to be worth about $2.5 trillion in 2022 by the National Association of Manufacturers, is one such niche that could reap significant rewards from a thoughtful and intentional approach to investing in AI.
First off, AI in manufacturing isn’t new, it just goes by a different name—machine learning. Secondly, much of the AI-based functionality being developed for manufacturing is done by major companies with well-funded R&D divisions and a strong foothold in the manufacturing market.
Indeed, virtually all of the major design software firms are finding ways to incorporate generative AI and large language models (LLMs) into their products. One such example is Boston-based software developer PTC’s cloud CAD system Onshape.
We caught up with Onshape’s CEO Jon Hirschtick to talk about its entry into the AI playground, how AI can bring value to manufacturers and what he sees in the near future for the still nascent technology. This interview was edited for clarity and conciseness.
Eng.com: How is PTC approaching AI technology?
Jon Hirschtick (JH): PTC is doing a lot with AI. We shipped some AI-based functions in our products and have a bunch of exciting AI-based projects under development, such as Onshape AI Advisor. We’re also active in the research community, where PTC employees are involved in research papers that are public. Our tools are being used by the AI community, perhaps even more than any other tools in our industry, for AI research. That’s another exciting dimension. There’s a ton of work and applications happening, and I’m proud that PTC isn’t getting overly hyped about AI, pushing things out before they’re ready. We’re focused and taking a solid approach.
Eng.com: So, at IMTS, you announced the Onshape AI Advisor, which is essentially an AI CAD assistant. Can you tell me why AI is a good fit for this application?
JH: With the Onshape AI Advisor, we’re using AI to provide expert user advice for fairly complex questions about how to use Onshape. These are the kinds of questions users typically ask another user or contact support or technical services. Instead of that, they can type a plain English question, even one that’s sophisticated, and get an answer about how to use a particular technique. AI is really good at this kind of task. It’s not rocket science anymore—it’s clearly within the wheelhouse of AI. This is very valuable for our users. We have deep professional CAD and PDM capabilities, and this is another layer of assistance.
Eng.com: Just to clarify the Onshape AI Advisor—is it a Q&A tool for users, or are you moving toward AI generating designs or doing the grunt work?
JH: It’s more than just an FAQ. It’s a conversational tool—you can ask questions, and the AI helps you with specific tasks, not just pre-set answers. As for generating designs, we’re working on it, but it’s not quite there yet. We’ve got great demos and research, but generating robust 3D geometry for manufacturing is still too complex. In entertainment, sure, you can generate digital assets, but for industrial applications, it’s not ready yet. There are too many safety concerns and error possibilities right now.
Eng.com: How long ago did you identify AI as something that needed to be developed for your product, your company and your users?
JH: It depends on how you define AI. If you define AI as machine learning, I’d say it goes back more than five years, when we started using machine learning to understand customer satisfaction through behavior. As for generative AI, almost from the moment it became known in research, we were involved in research papers, not necessarily knowing whether it would turn into real products. We had many people in the AI research community saying Onshape is the perfect platform for AI research because we’re the only cloud-native CAD and PDM system. Everything’s available through a REST API, so you don’t have to install a lot of software for large-scale usage, and with the right license agreement, you can access our public data library. Some people want to train on that 15 million human-created CAD and PDM documents. So, the research interest and commercial application started almost from day one. Over the last year or two, we’ve realized we could actually start shipping some of these things.
Eng.com: Absolutely. Your trajectory here matches many others who have integrated AI into their products. Can you talk a bit about the importance of further developing this and creating a baseline of AI capabilities to build on?
JH: I think it’s important that we explore what AI can do, and start shipping these products to customers because it’s new tech. I think AI is critical. I think our users must feel like product developers did when plastics or carbon fiber came along. It’s not just a better way of doing things; it’s a whole new set of tools that make you redefine problems. It allows you to approach problems differently. And so, the baseline is important not only for study but for releasing products. Just like with the first plastic product, you can’t know what it’s really like until you use it. We need to build reps, understand how to deliver and leverage the cloud-native solutions of Onshape.
Eng.com: The way you describe the Onshape AI Advisor brings to mind the concept of tribal knowledge within an organization. For years, we’ve talked about how companies lose access to knowledge due to retirement or attrition. This seems like a repository for that knowledge. It learns what people are doing and stores it for future users, correct?
JH: Yes, but it’s only a partial step. The AI assistant can give access to knowledge that a user might have taken with them when leaving the organization, so it helps there. But we’re far from tackling the whole problem. In the future, we might be able to develop a model specific to a company’s use of Onshape. Right now, the AI assistant only looks at general Onshape user knowledge—it doesn’t look at your specific data. But eventually, we could create a model around your company’s practices and provide insights like, “Here’s how experts in your company apply this technique.” So, it’s a partial step in that direction, but there’s a lot more we can do. We’re starting with Onshape AI, and there are other PTC products, like ServiceMax, which are also capturing expert knowledge.
Eng.com: It sounds like what you’re describing is essentially custom AI agents for a customer. Is that something on the horizon?
JH: Possibly. We’re not announcing anything yet, but if you ask me to speculate, I’d say it could definitely be part of our vision. PTC has a real leg up here, given our cloud-native infrastructure and the highly secure services we operate. We’ve been doing this for a while with Onshape and our PDM systems. So yes, we could imagine something like a custom AI agent for a company, where the AI looks at the best users in the company and helps new users align with those best practices. That’s something we could do in the future, based on the data we collect. It could be something like “What’s the best way to apply this technique in our company?” AI could inform decisions based on data, helping save time and improving efficiency.
Eng.com: AI agents are slowly becoming a thing for companies, but you need sensors and other devises to collect data for the agent. Is there a way around that that?
JH: With Onshape, we don’t need to go through that manual process of collecting data. Our system captures every single action as a transaction—if you drill a hole, undo it, or modify a feature, that’s all tracked. We have more data than any other system about a user’s activity, so we don’t need to go out and collect data manually. This gives us a huge advantage in training AI applications. In the future, users might even be able to combine data from their channels, emails, and other sources, and create a composite picture of what’s happening. We’re working on ways to give more value without relying on the kind of manual collection you mentioned.
Eng.com: Another PTC product named Kepware is a software layer that collects and aggregates shop floor data. Is this something you’re planning on incorporating into your AI products in the future?
JH: I can’t announce anything specific at the moment, but Kepware is well-positioned for this and there’s definitely potential there. Kepware handles sensor data, and we handle digital data from our systems. Combining those sources could be very powerful. The point is that PTC is uniquely positioned with both Onshape and Kepware, which span the digital thread spectrum. We’re also working with ServiceMax, which collects a lot of field service data, and they’re looking into AI for capturing service expertise as well.
Eng.com: Moving on to AI and digital transformation—what have you learned about AI’s application in manufacturing while attending IMTS 2024?
JH: At IMTS, I learned that we’re still in the early days of AI in product development and manufacturing. The promise is huge, but most organizations aren’t using a lot of AI yet. That being said, the number of projects people are working on is incredible. Everyone’s still figuring out which use cases work best, balancing doability, usability and value. I saw companies using AI in some exciting ways, like summarizing manufacturing data or configuring industrial products, but the technology isn’t fully ready for more complex tasks yet. I think the next few years will be about figuring out what works and refining those applications.
Eng.com: Do you think smaller companies are adopting AI faster than larger ones?
JH: Absolutely. The smaller companies tend to be more agile, and AI doesn’t necessarily require a huge investment in hardware or infrastructure. With Onshape and the AI assistant, you don’t need big machines or complicated installations. Smaller companies can jump in and take advantage of the latest tools without needing millions in capital. They’re moving faster in many cases, though not always.
