Episode Summary
The episode titled "Additive Manufacturing and Workflow Automation — with Authentise and Synera" delves into the integration of artificial intelligence (AI) within product lifecycle management (PLM) software, focusing on how it impacts engineering workflows in additive manufacturing. The guests are Andrew Sartorelli from Synera and Andre Vegner from Authentise. Synera specializes in engineering collaboration tools for simulation and additive manufacturing processes, while Authentise focuses on workflow automation in the additive manufacturing space.
During the discussion, both experts highlight the evolving role of AI within their respective companies. Andrew emphasizes that AI is a tool to solve customer challenges rather than a standalone solution, noting its progression from machine learning in finite element analysis (FEA) to more advanced applications like generative design. Andre shares insights on how AI has moved beyond simple regression analyses and now plays a significant role in areas such as process monitoring and simulation distortion. Both agree that while AI will not replace engineers, it will require them to develop new skills, particularly in interacting with AI tools effectively.
For PLM and engineering professionals, the key takeaway is the importance of embracing AI as part of their workflow automation strategies. They must adapt by developing a deep understanding of how to leverage these technologies for productivity gains and innovation. Additionally, maintaining a proactive and enthusiastic approach towards adopting new tools will be crucial in staying competitive in an ever-evolving technological landscape.
Full Transcript
Michael FinocchiaroOkay, we're live. ⁓ Welcome to the AI Across the PLM product lifecycle. This is Michael Finicchiaro and I'm joined by Andrew Vegner, Andre Vegner, sorry. Then Andrew and Andre, I'm gonna make that mix up. Andre Vegner of Authentize and Andrew Sartarelli of SINERA. Why don't you guys introduce yourself? You wanna start, Andrew, since you were the guy that was on time today?
Andrew SartorelliThanks so much, Michael. Yeah. My name's Andrew Sutterly. I'm the head of product management and software partnerships here at Sonera. I've spent my entire engineering career or entire professional career in engineering software. So I've worked at big companies like Autodesk and Hexagon and small companies like MTOP and Sonera and always focused on doing really fun things for engineers and solving some difficult challenges.
Michael FinocchiaroHa ha ha! Awesome. And you, Andre?
Andre (Authentise)I'm Andre, I've been running Walton Tires for 13 years, I think. We've been doing workflow software, mostly in the additive manufacturing space and have recently branched out into engineering collaboration. Lots of opportunities to collaborate with Andrew in Scenera.
Michael FinocchiaroThat's awesome. ⁓ So on this podcast, we talk a lot about AI and how AI is changing ⁓ the way we develop software and also how engineers get benefit out of the software, right? ⁓ So my first question to you guys, and it's interesting too because... Andre, you're coming in originally from a more manufacturing side, and Andrew, you're looking more from a simulation side, even if your background is also in computational design and so forth. What was your opinion, like, before you started out the latest release that you were working on, ⁓ where were you on the AI thing? I mean, we've been through already, what, three or four revolutions in less than four years, but did you, from the beginning, did you think AI has this massive potential or it's kind of high? We'll see what we can do with it.
Andre (Authentise)Do you want to go, Andrew?
Andrew SartorelliYeah, sure. I think for me, I've always looked at it through the lens of a product manager and what are the problems we're trying to solve? What are the outcomes we're trying to drive? And AI is just one tool in the toolbox to solve customer challenges. So I think it's an era we've always had an eye on it and we've seen the evolution of it from people doing machine learning and reduced order modeling for FEA to now the latest agentic error. And that's an area that we've been. focused on for the past 18 months of bringing agents to the engineering space. So I think myself, I'm always a bit of a, honestly, ⁓ a pessimist, a skeptic, ⁓ but I'm converted. And I think if I'm converted, that says something about the power of AI now coming into different areas. ⁓ So it's a very exciting time, think, in engineering software.
Michael FinocchiaroYou Yeah, me too. I mean, I'm up to, I think when we first talked, was at like 200 startups I had found in this area and I'm at over 320 now. It's just insane, the size. ⁓ Andre, what about you? I think you've been doing this for a while too,
Andre (Authentise)Yeah, mean, and so like the AI story beforehand has always been, you know, before the LLM explosion has always been like, everybody says they do AI to their investors and the clients get a machine learning story. And actually it's just some form of regression analysis. But, you know, now that's become much more accessible to sort of everybody and the limited applications that we've seen in our space, like in process monitoring or simulation distortion. AI's that aren't physics based don't require quite as much. ⁓ We've gone beyond those stories and LLMs have really helped us spread into areas that have been more generic in a way that addressing... reporting infrastructure, identifying events, creating live gunshots, know, all that kind of stuff that is really, really takes time away from the engineers doing their actual job. And so I'm pretty excited because the recent sort of generic improvements in large English models has given us an opportunity to AI, true AI outside of these very limited algorithmic endeavors that existed in the years before. So pretty cool stuff happening. I'm really excited about it.
Michael FinocchiaroAnd with the new product with Whisper, think there's even more of a pure AI play in terms of language,
Andre (Authentise)Yeah, mean, you know what we're realizing and I think everybody's exploring. Let me just put it out. Nobody knows the total truth of how much disruption there's going to be. ⁓ And so we're also in that exploring mode. And if you look at, you know, how we started exploring with LLMs, I think three, two or three years ago, we released a tool called 3D GPT where we talked 12,000. 3D printing out of manufacturing journal articles and we made them accessible via a rag system. That was our first sort of initiative to explore opportunities and we released an AI improvement on top of our engineering collaboration tool about 18, 24 months ago and that is pretty powerful. allows you to investigate the engineering collaboration information that we're collecting, for example, or generally identify events. That was pretty powerful, but what we realized is that there's real inertia around the workflow that engineers go through. There's not a lot of appetite to change the way people do things. If you're going to deliver AI to folks and really... generate that 10x improvement in productivity, then you're have to do that in a way that does it inside of their existing workflow. so, know, our, your Whisper again is an opportunity to do that because Whisper, and it's not released yet, but we're working on it in the background, it's essentially sitting in the background and collecting all of your collaborative data. That's your email, your transcripts, your Slack messages, and is you categorizing that?
Michael FinocchiaroI started that workflow.
Andre (Authentise)and understanding the implied permissions so that we can actually leverage that data that already exists but generally gets lost to drive additional improvements in your engineering process and, you know, automatic reporting, things like that, and maybe pass it on to the scenarios of the world or generative design algorithms that are able to then do something with that data. know, ultimately, the vision, and I think, you know, Andrea and I are...
Michael FinocchiaroYeah. ⁓
Andre (Authentise)Andrew and I are both driven by some sort of vision and the vision is quite simple. We want to go from intent directly to the design, the final design ready to be manufactured design. And there's obviously many, many steps that we have to address before we get there, but one is actually capturing intent. And that's something I'd be really interested in is, you know, what can we do with that data today, like automatic reporting?
Michael FinocchiaroRight.
Andre (Authentise)And what can we do with that data tomorrow is hook up that data, that your collaborative data directly with generated design algorithms to drive outcomes and no longer having to, you know, translate manually translate what, what intent is, and then turn that into a CAD file and then store that CAD file and dump the intent data. You know, we want to get away from that. So I think there's a lot of opportunity for AI to work in the background to understand what you're trying to do and then present. you an option to say, hey, this looks like what you're trying to do. by the way, we've already executed on your intent. Here's the design. Is everything OK? Cool. Let's go. So there's a lot of opportunity for that. And we're just at the beginning of that kind of exploration, how AI works in the background in a true agentic fashion. That was a long monologue, but very exciting of what the next steps are. that are coming up. you know, we're constantly evolving, constantly learning what are the problems that are most effective in solving.
Michael FinocchiaroI like that, think that Andrew, seem to agree too. ⁓ I think that ⁓ I was also curious because there's a lot ⁓ of talk on the internet about a vibe coding, a lovable, this European startup doing extremely well from a funding point of view and yet a lot of it, well, the limited experiences I've had have not been absolutely. brilliant in terms of the vibe coding thing. You can use something very simple, but anything that falls out of a very simple spec, you actually need a programmer there. ⁓ That being said, I'm fairly sure that both of you are using AI and your development teams are using AI. I'm just wondering how does that work today? Real development in 2025, AI ⁓ power development with cursor or just an AI, know, chat GPT sitting next to ⁓ code or whatever. How are you guys using AI in day-to-day development?
Andrew SartorelliYeah, I I think you had the right point there at the start. think Vibe coding and these sorts of tools, they work well for a certain class of problems. And, you know, we encourage our developers to pick the right tool for the right problem in a sense. So if it's a problem that's solved before, I think Vibe coding can do a good job of kind of doing some of that low-hanging repetitive tasks or maybe setting up test cases, things like this. But when you want to do something unique and new and novel, You you're not going to vibe code that you're going to get helped by large language models. Of course, and we have developers using large language models to accelerate different aspects, not just generate the code, but help them maybe find a new library to use that they didn't know about to solve a certain class of problem. ⁓ but yeah, I think just maybe to circle back to the whole vibe coding concept and applying it to engineering. think one of the challenges we have in the engineering domain is answers are right or wrong.
Michael FinocchiaroHmm.
Andrew Sartorelliin a very binary way. And when you think about coding, you can get the right outcome that you want visually. Maybe something looks right and it behaves the right way. if the whole system isn't correct, it doesn't work. And this is maybe why we haven't seen vibe coding yet for engineering. ⁓ But certainly I think there's these companies maybe like Scenera and Authentize working on solutions in this space to bring back type of exploration to engineers.
Michael FinocchiaroDeterministic. ⁓
Andre (Authentise)Yeah, a really good point that Andrew, I'm going to come back to that, you know, same order answer what we're doing internally and then talk about the engineering by coding opportunity. you know, internally, the two things that are blowing me away recently. One is that we no longer do wireframes. It goes almost immediately from concept to functional prototype.
Michael FinocchiaroYeah, go ahead, Antri.
Andre (Authentise)Right. that's, that's phenomenal. Right. So, that's, that's an engineer or product manager just like by coding what it should look like so that we can get feedback, real feedback from the customer before we put ⁓ our, our time into coding it. And yes, the developers themselves have oodles of different tools that they're using and we've increased our learning budget so that folks can go and experiment with different tools and come back and share those, those ideas.
Michael FinocchiaroBye. Mm.
Andre (Authentise)We're seeing it in the story point count that developers are doing, who's using AI effectively and who's not. It's very clear that this is a productivity driver. But another thing that I've seen, that I love is at this point, our commercial team, every member of the commercial team, mean non-developers, is vibe coding a tool to make their process faster once a month. These are guys that don't have any engineering background like me. I'm going out and developing a tool that...
Michael FinocchiaroMm. Thank
Andre (Authentise)creates customer UTMs that adds those to the emails, or we have a sales signal dashboard that's entirely internal. are a bit like these if this and that tools, no code tools that were all the rage 10 years ago, Zapier and so forth. Now we can go one step further. You what's the thing that we actually want to do? And we can turn that into a Google Apps script or something. And so it's made, made those process automations much more effective. I think I shared some of those with you, Andrew, on a car ride we had recently. there's cool stuff happening there. In terms of Vibe coding for engineers, we're actually starting a new company at the moment. So oftentimes it's incubating a whole series of ideas based on the fact that ⁓ we don't believe engineers are using AI nearly enough. And there's one company in particular I want to call up, Maneuven, that is leveraging AI specifically for reverse engineering. The reverse engineering probably might have heard that the latest Defense Authorization Act in the US has required the US DOW to ⁓ qualify a million parts for advanced manufacturing processes by 2027. So a huge opportunity. At the moment, we're probably... at about 100, maybe 200 parts of the qualified. So a lot of work to do. And to do that, have to really, we have to up the ball game in regards to AI. And we've identified 12 steps in the reverse engineering process where AI can yield substantial productivity.
Michael FinocchiaroMm.
Andre (Authentise)improvements and when we look around to the incumbents, the people that doing reverse engineering as a service, we're seeing nobody use any of them. We're working with Boeing to improve our tools to help them reverse engineer their tools but for Boeing to get an approval to leverage these tools, it takes them months or years. So there's a lot of opportunity. We're not seeing people... ⁓ leverage, you know, utilizer, learn from it, just get going. And so I'm not even talking about actually creating the 3D model from a drawing. I think backflip and the models, the foundation models that being created to drive 3D modeling are still in their infancy and the maturity of those tools is just not there yet. But there's a whole slew of understanding existing drawings, creating bombs. you know, generating manufacturing processes, DFAM, in-process monitoring, inspection, report generation, a whole bunch of different process steps where AI is super important. And so I think that the message out there is if engineers aren't adopting these tools, that organizations are likely to be disrupted by organizations willing to engage very deeply in leveraging these tools.
Andrew Sartorelliyou
Michael FinocchiaroInteresting. And so in the current version of Scenera and Thras and so forth, ⁓ how present is AI, ⁓ either when the customer is actually using a chat or whatever, or behind the scenes you guys are using AI? That would be interesting to know. Where are the touch points between your app and AI and the user?
Andrew SartorelliYeah, maybe from the, the scenario side, mean, we've released, our last release, the, we call a multi-agent system builder. So you can actually construct systems of multiple agents that interact and work together to solve problems. then, you know, that's what the, creative, the, the process is. They see this nice little canvas to construct things. And then the, the people that are consuming it, right? Because generally there's one person that knows how to make the process, but then there's a hundred people that maybe need to interact with the process. They get a chat interface where they get feedback back to them about what's going on. So that's, would say the most direct interface with AI that our customers see, but there's lots of ways that we've seen it applied in SINERA previously, like creating pipelines to feed these traditional machine learning models and things like this with simulation data, CAD data, so that you can create a foundation model at the end of the day, because you need a data set of tens of thousands of data points to feed that model. And then we're working on some other areas where maybe there's going to be AI used, but it won't be so directly apparent to customers. Which is kind of for me as a product manager, the best way to approach this problem is if customers don't even know there's AI there solving the challenge for them. It's just an amazing solution to a problem. we're looking at ways to help people create training data for AI models using AI models. ⁓
Michael FinocchiaroHmm.
Andrew SartorelliIn a way that it doesn't, it feels natural, it feels inside the product. It feels like it's just ⁓ a nice way to interact with things without being, Hey, this is AI. ⁓ You know, I think that's, that's the best way to approach these things.
Andre (Authentise)Really interesting. Sorry, Michael. Yeah, like we have ⁓ three announced versions, ⁓ features that utilize AI. That's ⁓ a event identification tool that basically...
Michael FinocchiaroYeah, because I think so too. Yeah, that's cool. Go ahead, ahead, Antri.
Andre (Authentise)goes through a bit like understands what has occurred if a question has arisen, if a decision has been made and it marks that at the moment, it suggests that event to the user a bit like Microsoft Clippy would have done 20 years ago, but hopefully a bit more effective. And then we have a chat window that. utilizes your engineering ⁓ collaboration information and presents you with answers on something detailed like the material density of a spec sheet that you've uploaded or something more generic like the risks inherent in your engineering project. And actually that latter question ⁓ is an alternative to an MBS process. So like the ROI of just that question is like $150,000, right? Because otherwise you'd have to hire a MVSE engineer to build you a cameo model to identify the risks. And actually we've proven in an Air Force project that we managed that the model is like... twice as effective at identifying risk as a traditional cameo model might be. So huge, huge ROI. And then the third ⁓ area of even larger ROI is a technical data package generator or a report generator leveraging your collaborative data. So we take your collaborative data and you have given us a template of a report and we generate that template. And so you see a repeated theme here. that theme is we'll take manual grunt work and we'll turn that into something that's highly automated and very reliable. And we've done that now. The OCR models, ⁓ AI-based OCR models are 10, 15 times more effective than traditional OCR models. And so we're utilizing them both in our engineering collaboration tools and in our...
Michael FinocchiaroMm.
Andre (Authentise)production management tools that, for example, suppliers, ⁓ material testing labs or testing lab houses can upload their ⁓ PDF or Excel sheet and these AIs automatically extract the test results so you don't have to enter those test results each time. Things like that, again, really effective. We're doing the same with ⁓ inbound reverse engineering projects that I just referred to. The drawings, sometimes they're very old, handwritten drawings, using those, OCR on those is really effective as well. And then there's more we can do with that information, specifically around those drawings, you can automatically create a bomb. and then you can use that bond to suggest new manufacturing recipes given new manufacturing capabilities. So pretty exciting things that we can do to do a short circuit of reverse engineering process, leveraging, starting with the OCR model, using LLMs more generally. three announced, one. ⁓
Michael FinocchiaroMmm.
Andre (Authentise)One in production that's not been announced separately yet, and then three more use cases coming in the next three months. So quite a lot of work being done behind the scenes.
Michael FinocchiaroVery cool. Last week's scenario gave a really cool conference about the future of engineering, and I was privileged enough to present there. When I'm looking at the big three, the XEMAS, DASO, PTC, I'm seeing sort of three classes of AI usage. There's the first generation, which was just the chat GPT wrapper, which is usually sitting on top of documentation or knowledge bases, potentially a vault. Right? And the second one is what Andrew you referred to as more the agentic approach where I've got agents and probably agents very, very focused on one piece of expertise in order to try to keep it as deterministic as possible because if you give them too much leverage, they're going to start hallucinating. ⁓ And then I think the third one, and I don't, and maybe Andre you've referred to that or, Andrew, I'm wondering where you're playing to that is this industrial metaverse, right? You've got these new standards of OpenUSD where you can export models into, ⁓ well, I think the biggest commercial solution is, of course, the omniverse from Nvidia. And you can do these simulations in real time with machines. ⁓ And then that data is a closed loop, right? I can put that, whatever I learned, back into the model, improve the designs. And that seems to me... pretty much that AI factory that Jensen Wong announced at the GTCs this year when he said every physical factory will have an AI factory. I think that Industrial Methodist is sort of that AI factory that we're all collectively building, Sonera and Authentize and all these other awesome startups while we're waiting for the big three to kind of catch the same train into AI land, right? So I don't know how you guys want to react to that, but I just, it was one of the thoughts I was having as you were talking.
Andrew SartorelliYeah, I I think as Andrea said, all of us in the, think engineering software space are building towards this common vision of where we kind of see the direction of the future, but it's going to take us time to get there. And there's a lot of stepping stones along the way. You brought up ⁓ open standards, open, open USD. I think standards are a great way for us as an industry to collaborate and work together and make sure that, you know, my biggest pet peeve. is I have to redo what I've already done before in the past, you know, and additive manufacturing, there's orientation ⁓ problems that you have to solve. it's the same algorithm more or less ⁓ everyone does, right? But everyone needs to have that. And I think standards are a great way that we can make sure that André is working on the problems that his company is best suited for, Saner is working the problems we're best suited for. And then in the end, we can bring all of these together.
Michael FinocchiaroIt's annoying.
Andrew Sartorelliand create a really compelling solution for engineers at the end of the day. And so I mean, for agents and the agent agrar, there's things like MCP. I'm sure Andrea is working on things there. now we're working on a lot there as well. A2A so that agents from different companies can start to work together and collaborate together. I think that's really going to bring about, you know, this new era of open ecosystem, hopefully, in the engineering software space, because at the end of the day, know, engineers, they work with products from one of the big three or all of the big three and a list of other vendors. And we just need everyone to kind of collaborate, which has been a challenge in the past. But I think this new AI agentic era is going to force that collaboration, which is pretty, exciting. When you think about the types of challenges we're going to be able to solve for folks. And so if we want to have this digital twin of the product at the end of the day, well, you're going to need to do every, all the hard work beforehand. to get to that point of having that ideal end state. so you also said one thing I want to pick up on, which is this determinism. know, again, this is the challenge with engineers. Your solution is right or wrong. can't be, you know, right 85 % of the time and wrong 15%. That's not a working solution for engineers. And I think that's, you know, one of the challenges we're solving at Scenera is trying to make that, add that determinism to this, this AI error so that every time you use your agentic system. you get the same outcomes ⁓ because that's, as engineers, we've come to expect from our tools.
Michael FinocchiaroAbsolutely. Yeah, it's funny because I was having another webinar where we were talking about determinism and manufacturing as someone pointed out. Well, actually manufacturing is also probabilistic. It's possible the mechanic makes a mistake and it's possible the machine breaks down and therefore you don't get that stuffy reason. So it's kind of interesting to see it from that point of We'd like to see the idealized engineering view of the universe which everything is deterministic, but in reality it's not. mean, Andre, you've done a lot more in that ⁓ industrial space of machines. What's your feeling on that?
Andre (Authentise)Yeah, I mean, one of the areas I have been bearish on is kind of digital twin solution that omniverse kind of solution set is pushing towards. Primarily because I believe that what we need in manufacturing and engineering is more agility. ⁓ And every time we try to encode an existing operation in a digital environment, we reduce agility. We don't add to it. ⁓ So I'm less bullish on ⁓ that specific opportunity. I do think that capturing ⁓ what happens in real life is important because we want to add traceability, we want to use that data for additional ⁓ algorithmic work before and after. But to just encode something, and we've seen it in factories, it predates AI models, idea that we record what's actually, visual record of what's happening on the shop floor doesn't help. a great deal in my view. it's something I've been struggling to understand about personally. ⁓ And so what we're doing really is going way down with our customers and just understanding what are your specific that the material test ⁓ upload, we just we understood from engineers on the shop floor that they're spending a lot of time just copy pasting test results and they, you know, try to match. And that's not what they went to university for for 10 years. And so that's kind of where we're at. And I think I've always believed that listening deeply to folks on the shop floor, than having these kind of, other than this kind of one great vision of reducing the time from idea to part, really spending a lot of time like grassroots and taking from there is the most important thing.
Michael Finocchiaroyou I agree. It makes me think of another question that we were pondering on the future of PLM Webcast. I had a professor on, Patrick Hilberg, really good PLM guy. And he posed this question that I thought was really good and maybe you guys are both on either side of this thing too. He was saying like, you know, these problems that Boeing had with the 737 MAX, right? They were engineering problems. So. in a perfect world, would PLM have solved those problems or not? And it's an open question. it's a I'm just throwing at you guys, I'm just saying how you would react, because I think it's an interesting one. mean, if we're in the world to make it a better place and to make better products, and yet one of these big products falls out of sky multiple times despite having done all the engineering, you know, what do you guys think? Is there is AI? like the bridge to make sure that those errors don't happen because we can test it any times more with a machine than we can with a human being. I don't know. What do you think? You seem to be thinking about it hard there, Andrew.
Andrew SartorelliYeah, I mean, I think at the end of the day, we'd like to think about as engineers, think there's only engineering problems, but they're really business problems, right? you know, engineers, we work for companies that need to make money. ⁓ And we can't spend all of our time just doing engineering stuff in a sense, right? And so I think AI is going to bring an opportunity to speed up development processes. This is exactly what we see with European automotive OEMs. They're under this pressure to design vehicles at the speed that China is developing vehicles, which is every two to three years for a new vehicle class, right? And if you think about a European OEM, it's five to seven years. ⁓ The challenge is the same though. They need to engineer a car that meets all of the standards, all of the requirements in a much shorter period of time. And we know that this is not going to happen by just adding more people to the process, more human engineers to the system. We need to find a new way of engineering, designing, yeah, I think agents and AI are that solution to that problem that, you know, I always say, if you're a European company, try to schedule a meeting in June, July, and August with five different people. And the first meeting you're going to come up with is in September, right? It's just how it works. And we need to find solutions to those problems. If we want to develop products at the speed we need to. with the quality that we need to make sure that incidents like you described, Michael, don't happen in the future. Because as engineers, yes, we are responsible for the outcomes of our work and what happens to the end users.
Andre (Authentise)I'm not as familiar with the Boeing root cause analysis but one root cause analysis that does come to mind is the NASA climate orbiter crashed in the 90s. Exactly, that would have easily been called. ⁓ I truly believe that there's a lot of information that we currently just discard that
Michael FinocchiaroYeah, with the metric conversion thing,
Andre (Authentise)that is inside of the collaborative practice between engineers and the more complex systems we're going to get, the more people are going to have to collaborate on them. And it's really important that we leverage that information. And what really pains me at the moment is that we're reverse engineering things, and we're doing that basically having to start at ⁓ the beginning again, because all we have is legacy drawings. We don't know why decisions were made, why tolerances were added, why materials were selected, and all that jazz. what we're doing is we're recording it in yet another 3D model that doesn't record any of those in that intent. So we're making the same mistakes over and over again, despite the fact that we have these AI tools available, able to address some of those challenges by taking semi-structured or unstructured data and turning it into structured information. That's kind of the value that we have in these tools. And it's insane to me that we're still making the same mistakes.
Michael FinocchiaroExactly.
Andre (Authentise)know, the forms like this, where we hopefully can shake people awake and make sure that they leverage these tools and start using them is an important step and everybody has to start listening.
Michael FinocchiaroAgreed. And so I guess I usually wrap up the first part with a question about, we've now talked about AI, your opinion of AI before you started your current release and how you're using it development. Now we're heading towards the end of 2025. Who knows what AI is going to come up with? mean, just in the last two weeks, you've had the e-commerce platform from OpenAI. Oh, and then we're going to have adult content in OpenAI, which is like...
Andre (Authentise)Thank ⁓
Michael FinocchiaroJust another low, right? Who knows what's going to happen? How has your position changed on it? Is it fundamentally different or it's the same wait and see? We'll use what's the best ⁓ in class and we'll hold off on some of other stuff. Where do you guys sit on it today?
Andrew SartorelliI think I've done a complete turn myself. would say, I think AI and particularly large language models have a place in engineering. They've established a place in engineering by solving fundamental challenges that folks have, whether they're, you know, the low hanging fruit you're talking about, Michael, of just making documentation more accessible. Cause who wants to read through a 300 or 3000 page manual to understand where there was this one comment about something. Yeah. Having, ⁓ you know, agents accelerate, saw accelerate engineering processes from weeks to days and hours. ⁓ it's here to stay. It's going to solve, ⁓ complex challenges. And I think what we're going to see in the future is, know, today we have, ⁓ language models that are, you know, working on a tech spaces. ⁓ I've talked to some very smart folks, ⁓ on my team about AI and they say, Vision language models are where it comes next so that we can start to understand and interpret 3D models because at the end of the day, engineers, we don't work with text. We work with physical objects described in three dimensions. And so I think when we start to see those classes of ⁓ solutions come out of the research area and into commercial software, we're going to start to see some really interesting things done. know, at Scenario, we're always building towards this vision of, you know, Jarvis from the Iron Man movies.
Michael FinocchiaroMm. Yeah.
Andrew SartorelliAnd I think when we start to see vision language models, we're going to start to see that really become a reality.
Michael FinocchiaroThat's interesting because when I spoke with the Mauer at Leo AI last week, he said he had they've already created an LMM, a large mechanical model. So he's already put like all these mechanical engineering textbooks and best practices and all this stuff into a model, which then if you're inside a 3D part and there's a fill it, it knows what a fill it in a chamfer is and it can make a difference and say, no, actually, it can't be done there because the machine can't get in there kind of. It's pretty interesting that we're getting there. It's not quite fluent in 3D, but it's getting better at mechanics. How about you, Andre? How's your thoughts on AI changed over the last 18 to 20 months?
Andre (Authentise)I'm seeing a lot of data, but basically not. But the opportunity set that we identified three years ago when LLMs first started, or four years ago when LLMs first started really making waves are still the same that we have today. It's been clear to me and I wrote, think, a Forbes piece about this years ago that these models aren't ⁓ geared towards accuracy, so we can't expect them to replace 3D modeling for mechanical design because that requires accuracy. Straight out, people are working on those foundation models, but we're still a way away. Let's not fight to do something that these systems weren't created to do, but leverage them where we know them to be good by using them to turn unstructured data into structure. ⁓ It's been about understanding that, is, know, fundamentals of that haven't shifted. What has surprised me a little bit is that the product is not reflected in productivity gains as fast as I would have expected it to. Like, societally, we are behind the curve. And if I look at why that is, I think there's fear in organizations and also maybe software companies like Scenera and Authentice haven't been good enough at delivering.
Michael FinocchiaroMm.
Andre (Authentise)those tools to people that really make sense to them, that drive the value home. So a lot more work to be done, I think, in delivering the productivity gains of value to users. once we do that successfully, once it's not like 97, I just saw a survey result that said that 97 % of people
Michael FinocchiaroMm-hmm.
Andre (Authentise)have used AI but all of them are less happy now than they were before. We want to do. So yeah, I think there's still a lot of work to do in terms of making these tools accessible to people and making them useful. But the fundamentals of where I think these tools are going to be useful hasn't shifted. It's just that it's taking longer, as with every industrial process, it takes longer to deliver that gain.
Michael Finocchiaroyou We have to wait for Jarvis a little bit longer. ⁓ So when you guys ⁓ go into take your products to market and you go into a customer, I like to think of digital maturity on sort of ⁓ a spectrum of one to five. One is like they're still using email and they're still using Excel. Five being, ⁓ know, agentic, adaptive, automated digital twins, which is basically science fiction still.
Andre (Authentise)Yeah, always.
Michael Finocchiaro⁓ Although many science fiction writers I've read thought we'd be there by 2025, but we're still a good 10, 15 years out at least, I think. Where do you assess, like are you able to assess going in? are they, do you feel like in the industries you serve, the customers are more between one and two or between two and three? Are there some outliers that might be as high as three and a half? I mean, just to get an idea.
Andrew SartorelliI mean, I think it really runs the spectrum, right? You see folks maybe at the leading edge of digitalization, maybe typically smaller companies that are newer, right? They don't have the existing processes that they've had to digitize. start...
Michael Finocchiaroand all the technical debt, right? They have less technical debt.
Andrew SartorelliExactly. As you think about, let's say a company that's transitioning from ⁓ one CAD version to the next CAD version, that is a massive undertaking to go through that transition and convert things. Whereas if you're just starting from scratch, you can pick whatever CAD product you want, whatever PLM product you want and really establish processes there. So ⁓ we see it all. It's an era. I don't think anyone's at this five state yet, this ideal end state. But the nice thing is we're all working towards it. You know, we've been working with companies in a collaborative manner where we bring a number of our customers together and take best practices of what works, what doesn't work. You know, I think I always joke it's, 2025 and most engineers still work in Excel every day. And I think in 2035 or 2040, it's still going to be the same thing. Engineers are going to have Excel documents all over the place. Why is that? It's, it's easy to use. Everyone understands it. Maybe our desire should be, can we get people using specialized products that solve problems in an easier way? Hopefully. ⁓ I think part of it too comes down to IT challenges of, you know, as an engineer, what's easier for you to do open up an Excel document or create a SQL database and your IT infrastructure with the right permissions and all of this kind of stuff. No, the engineers is going to pick Excel and store everything in Excel. So I think we as software companies have the challenge of bringing our software to, to our engineers where they are. with the tool set that they have or the knowledge they have and the capabilities they have ⁓ and help them then bring them further up into this digitalization process.
Michael FinocchiaroAndre, same observation or you got a different one?
Andre (Authentise)Yeah, I mean, was gonna, I was gonna say, I think we might need to work on that one to five definition because people are going to use email and an Excel forever, right? Like that's completely agree with Andrew.
Michael FinocchiaroOkay, but just take eBOM to mBOM conversion. That should have been inside a PLM system 15 years ago and it's still in Excel. So that's what I mean. The majority of engineering transformations when you go from one department to another, is it done in an Excel spreadsheet sent by email or is it done in enterprise system where it should be so that everything's recorded, all the design intent that you were talking about, that that's recorded somewhere. It's not just...
Andre (Authentise)Yeah. Yeah,
Michael Finocchiarolike you were doing in Whisper, you're gonna get those chats out of Teams where all that information is being exchanged and not actually landing in the system of record anywhere,
Andre (Authentise)Yeah, I I think we have an awesome opportunity to make life easier for people and not force them to change their process. Like, let them continue to exchange information on Teams and, you know, via Excel sheets. Like, learning a new tool, getting that adopted throughout the whole organization is a massive undertaking. Never happens. Like never happens. Like why? It doesn't even happen for an organization that's 30 people strong. it's like, why would we try and force somebody else to do that? So let's try and step away from this idea that, I mean, that was like the major twist for me. ⁓ I've always thought that we're going to end up turning intent into, you know, designs using generative design systems. But, you know, five years ago, I thought that would have to be done via... like a structured JSON script and a standard, right? Now I don't believe that. think there will be a standard, but we'll be auto converting it into that JSON script from unstructured data. And that won't be a problem. I don't, I don't actually, I do think there's a problem in AI adoption, full stop. but I would be interested in working with YouTube and identifying those like one to five, you know, what, does that actually mean to folks? And like, Andrew, we got
Michael FinocchiaroInteresting.
Andre (Authentise)know, steering committees of customers that are working with us to identify what that means to them. But it's still only anecdotal. would be actually quite pretty interesting to define that. ⁓ I am on like I've had several conversations with large companies in the last six months that are only just starting at an organizational wide level to understand the use cases of AI in their company. like tens of thousands of engineers working there and still did how clearly, like surely their engineers are already doing stuff, but like at the top level, they're only just waking up, which is astounds me. Like, and so, yeah, I think that's one level of that one to five is, know, like how agile are you at promoting the use, even if it's not as like a centralized use case, but like how are you?
Michael FinocchiaroYeah.
Andre (Authentise)How successful are you at advocating for the use of AIs and then capturing where that AI is generating value for your organization?
Michael FinocchiaroInteresting. It's sort of, when you're saying that, it makes me think that, I guess that in Cheyenne, we haven't had an iPhone moment, right? Because before, if you remember, I had a paper calendar and I had a Walkman or a CD player and suddenly I had this iPhone thing and everything was on one device. And suddenly that was my life and everybody's life is on this damn iPhone. PLM and engineering in general hasn't had that moment. We're still using Excel, like you said, and we have to because it's easy. Occasionally, we'll have awesome tools like Scenera and Authentize, and we could do some pieces of our job in it, but most of the people are still going to be using the Nokia and the Motorola flip phone and not necessarily on an iPhone or a Blackberry. Well, talking about companies that didn't see the wall coming. ⁓
Andre (Authentise)or even a blackberry.
Michael FinocchiaroSo, but what happens though when you bring in Authentize and Scenera, when people see the power of these tools, you can get to a certain, maybe I should have been more clear, I was trying to talk more about data maturity, you the fact that you've recognized that data in silos is a bad thing and you need to have data custodians and data owners and data stewards in order to make sure that the flow of data across departments is smooth. and governed and so that you can use AI in order to gain data. Maybe I should have said that for me. That's sort of what I meant on the one to five also, it's just like in terms of data maturity. ⁓ But when you come with your solutions, is there an aha moment or is there maybe a slow ripple effect of the companies realizing, my God, if I gave better data to authenticizer scenario, I would have better results and more powerful results and I'd be able to go faster and more efficiently.
Andrew SartorelliYeah, I think that that's this idea of an aha moment. We see it all the time now. So engineers start to use scenario to automate some of their processes. Then they bring in the agentic layer and it's like, my gosh, there's these things we've talked about, you know, as science fiction that we can start to achieve now. And we often find that the next challenge they have to solve is this data challenge because, you know, as engineers, data is scattered all over the place in different sources and
Michael FinocchiaroReally?
Andrew SartorelliNow there's like, actually, I would say real value in consolidating that information and making it accessible to tools like Sonera or tools like Authentize because I think before it was organization for the sake of organization without so much value being brought there. And now we see that there's the 10X, the 100X value proposition coming from having your data connected in a way that wasn't before. And maybe one thing I wanted to get back to you of why we haven't had these, these bigger iPhone moments and things like this. ⁓ you know, think for our industry engineering, typically processes are so unique to each company. So, you know, you take two automotive OEMs, they probably have the same tool stack more or less. They have the same department names, but the way that they engineer cars is complete, maybe completely different from one another, right?
Michael Finocchiaroyou
Andrew Sartorelli⁓ And how do you create a system that works with these two different processes? Right. And this is why always there's a degree of customization and consulting required to roll out systems. And maybe now, know, Andrea is talking about working with ⁓ unstructured data and that's the advantage of AI, right? You can work with these different data pools and bring it into the right format. Maybe there is that now that opportunity that we can create solutions that work with all of these different processes in a way that doesn't create so much overhead for folks to implement.
Michael FinocchiaroI like that. Andre, you want to add to that?
Andre (Authentise)Well, I mean, like, I haven't really seen... ⁓ I haven't seen a change in the urgency of centralized data repositories. ⁓ Having taught at Singularity University for a decade, I kept on hearing that reference way before LLM came in. It was acknowledged at this CIO or the VP level that some sort of centralized data infrastructure was critical. ⁓ I haven't seen a step change of folks saying this is either easier or more important now than it was before. So there's definitely, and that's kind of what Wisp is trying to address is that making it easier to create the central repositories without forcing the user to use different tools. But the proof point is still out there. So yeah, ⁓ I think that the aha moments and Andrew alluded to this are still stuck at the individual engineering level. Like it's not at the organizational level yet. And it's really tough for folks to learn like from those individual aha moments and create a consolidated view of what the world should look like for the whole enterprise. It's possible that the productivity gains that we're making with AI in engineering ⁓ will be limited to small and more agile organizations. And that wouldn't surprise me because we have historically seen that, you know, the large enterprises are beholden to one CAD or PLM vendor.
Michael FinocchiaroMm.
Andre (Authentise)signed a 10 year agreement with the SOE and others have done similar things. Maybe that's where it's at. Maybe those gains aren't going to filter down anytime soon to large enterprises that are making our cars and as a result create more opportunity for enterprise disruption because the hypothesis that we're pursuing at the moment is that none of these organizations are very good at. like the reverse engineering incubation that we're doing. They're never going to get there. So let's do it and show them how to get there. ⁓ But I don't know. Andrew, have you seen that urgency? I'm not trying to undermine your point. I'm just like, I haven't seen it. I want to see it, but I haven't seen it.
Michael FinocchiaroYeah. No, it's a face, I thought.
Andrew SartorelliI think we've definitely seen a change in the past year in the urgency to adopt AI. I the automotive industry we all know in Europe is going through some challenging times. And when you're going through challenging times, it's a great opportunity to change and adopt new processes. And so you're right, Andrea, we haven't seen the organizational company level aha moments yet. think, I think 2027 is probably that year where we see
Michael FinocchiaroHmm. Yeah.
Andrew Sartorellireally big news from some companies about how they've had organizational transformation ⁓ because of AI and the engineering domain. ⁓ Today we're in the, I would say the individual seeing those aha moments and success stories and maybe pilot projects, the involvement of the CAO, the CTO, these sorts of things. ⁓ And we're starting to see that, okay, I've done it for this one use case. What if we apply it to our entire product line now? What does it take to apply to our entire product line? How do we go broad? How do we get bigger? And I think for scenario, that's what we'll see more and more in 2026 and 2027. I think we're going to see some really exciting success stories of, you know, bringing the, again, the engineering time down from years and weeks or months to days and hours.
Michael FinocchiaroYeah, a of mine, we were talking to, a friend of mine was saying it's almost like, you you've got an idea of a product in your head and you want it to be on the desk like right away. But I don't know, maybe by 2027, we'll have AGI and Fusion and Quantum. I mean, maybe it'll just be the trifecta, who knows? So we got a couple of things as a kind of closing question, because I do...
Andre (Authentise)But you're a patient man, Andrew.
Michael FinocchiaroI noticed looking at my audience demographics, do tend to get some younger engineers and probably there's a lot of, well, absolutely there is even on people our age, there's a lot of anxiety, right? Like AI is gonna steal my job, right? ⁓ However, I think from what you guys said and what I've taught when I've talked to other startups and folks is that ⁓ AI should really be creating more opportunities actually, because AI is taking away the more mundane repetitive tasks and allowing you to be more creative. So my question to you too, and sort of a closing is like, what advice would you give to people that are ⁓ either in transition or starting out a career in order to find the right path that keeps them valuable as a human contributor and not easily replaceable by an agent or whatever?
Andrew SartorelliYeah, I mean, you're right. AI agents, they're going to change the way that engineers work. But we've, we've been through these transitions before, right? And we haven't ended up with less engineers with more engineers. We went through the transition from, from paper drawings to, a 2D drafting on computers with tools like AutoCAD. You know, still tons of architects out there in the world. We've made the transition to 3D, grew an order of magnitude again with the amount of engineers, designers needed. and I think AI is not going be any different. We're still going to need engineers, but we're going to need engineers with different skillsets. Right. And so the architects that didn't transition to, ⁓ AutoCAD, ⁓ you know, they survived for a while, but probably not the longest, the same for, for the AI error. I think, you know, we're, seeing engineers need a new set of skills, right. To be a prompt engineer, right. To understand how to interact with other people, how to get the right information out of them, how to give that, have them. work in the way that you need is something that I think engineers are going to need in the future, like all of us, right? We're all using AI in our jobs. As a product manager, my job is also probably, you know, it is changing dramatically from what it was even just two or three years ago. So I think it's an exciting time for all of us. And, you know, we're going to see a lot of cool new things also possible that, you know, we never could have imagined, you know, a decade ago.
Andre (Authentise)Yeah, in order for that truly to be more opportunity, we're going to need to see productivity gains dramatically ⁓ increase. And I'm not seeing that, frankly, full stop. The only way we get more opportunities if we're more efficient as organizations and therefore are able to do more and have bigger gains to be able to employ more people, but it's not filtering through yet. I think that we really need to start turning these ideas of improvements into actual bottom line improvements. That's a pretty urgent thing to do. ⁓ Well, you know, that singularity we kept on talking about, we keep on talking about is... Decade, a decade, a generation or two generations ago, you had a job for life. Now you don't have a job for life anymore. You have a career that changes every 10 years, right? You used to be a microfinance investor and now I run a software company in engineering, manufacturing, Things change. And I think that folks need to get ready for that. ⁓
Michael Finocchiaroyou
Andre (Authentise)I see it across my company and we have engineers and commercial people alike. The single biggest determining factor on whether you're going to be successful or not is if you turn up every day like it's your first day and you bring energy to the table. it doesn't like skills on that.
Michael FinocchiaroHmm.
Andre (Authentise)on that power flick. It is that easy. You've got to really force yourself to be there and it's easier to do if you love your job. But the moment you switch off, you've already lost. The moment you express fear at adopting new tools, you've already lost. So you've got to turn up with real energy every day and that's how you're going to be successful no matter what new technology comes along.
Michael FinocchiaroAwesome. Well, it was a fantastic conversation. I appreciate you guys both taking an hour out of your busy days to just chat with me and then talk a little bit about AI. Did you guys have a good time? Thank you.
Andre (Authentise)Love it. Thanks.
Andrew SartorelliAlways great to connect.
Michael FinocchiaroWell, I hope you have a good time. This is Fino once again. I'll be back with a couple of French startups pretty soon, NUS and ⁓ SP3D. And in the meantime, thank you. Thanks to Andre and Andrew, and we'll see you next time. Cheers. And then of course, the recording will be on the YouTube and LinkedIn as usual. One second.
Andre (Authentise)Thanks a lot.
Andrew SartorelliSo.