0:00

Voiceover: You're listening to Augmented Ops, where manufacturing meets innovation. We highlight the transformative ideas and technologies shaping the front lines of operations, helping you stay ahead of the curve in the rapidly evolving world of industrial tech. Here's your host, Natan Linder, CEO and co founder of Tulip. The frontline operations platform. Madi: This time last year, we were also talking about AI Natan: and Madi: how disappointed we were that we were the only ones doing an AI demo. Natan: Are we, are we recording already? Madi: Yeah. Natan: Oh, this is great. This could be a good start. We're in Hanover, high on AI. Madi: Yeah. I think Hanover is high on AI. Natan: Hi Maddie. Madi: Hi, good morning. Natan: Good morning. Every year in Hanover, we podcast from the show. It's been quite a week. Madi: Yeah, honestly, all of the days kind of like blurred together, but it's definitely like an interesting moment in time for manufacturing, especially manufacturing software and kind of the convergence of manufacturing software and manufacturing tech and automation. At least that was one of my big impressions from the show this year. Natan: Yeah. We should like roll the tape and listen to last year episode, but, uh, one of the things I remember we were amazed that finally there's like serious pavilions and halls full of software and that now we're seeing not just more of that, but less of automation. So the trend of how Hanover has transformed over the years to emphasize software is just phenomenal. Madi: And, you know, I'm a little disappointed to hear that because I do think that automation is still a critical part of manufacturing and really segmenting out or having less presence by one side, like whether it's the software side or the automation side, I think that that's not like, The full achievement of having one show fully representing the ecosystem, which I think would be great to work towards. Natan: Yeah, but you know, the flip side was that five, seven years ago, you were hard pressed to find anybody. And like, yesterday, I saw Meta Facebook booth. And we were impressed last year that Snowflake showed up, you know, Facebook is here, like Meta, showing their VR stuff. So it's, it's kind of interesting. And I also saw folks like, uh, ServiceNow that I don't, I don't believe that were here last year. So it's not just more software, it's also more players like, you know, suddenly Cloudera. And people that have some sort of enterprise software, they're coming into operations manufacturing to help solve hard problems, which is okay. But I agree that we can't really have an ecosystem that works without the automation players. And that's actually a good segue into our story here. It was very nice to see not just our ecosystem coming together, these booths. Madi: Yeah, we were in five different booths this year. And honestly, I saw a lot of companies in the wild, a lot of Tulip in the wild, but a lot of companies kind of in a similar suit where they are in each other's booth. So on the one hand it's like personal and it's like a Tulip story, but on the other, it's an evolution of how companies are working together. But we had our own booth across from AWS. We were also in AWS with a lot of friends and competitors and just interesting solutions. And then we're also in Microsoft. I would consider most consulting firms friends. Yeah, Natan: totally. They're all friends. I'm saying it's just like the type of animals in the zoo. You see the smaller company startups, you see the consulting firms next to sensor companies, next to Snowflake. And AWS's booth. Yeah. So it's kind of interesting. Like, some people would think Snowflake and AWS are competitors, but in manufacturing they're working together. Madi: And I think one of the things about the digital halls that we were in that is a little different from the other halls is that kind of solution diversity in one area and kind of seeing the interconnectedness. Like, you don't have, like, A row of companies that are Andonlites, which is also cool, you know, but you're into that stuff, Natan: it's like amazing. It's like Disneyland for Andonlites, it exists. Madi: But, you know, you're seeing the maker of the Andonlite, like in our booth, you know, Banner, next to the maker of the Bench, next to the maker of the HMI, next to the cloud, you know, and that is the representation of ecosystem for manufacturing. Natan: Landing AI that makes AI as a service. That now integrates to Tulip and a bunch of other solutions and makes it possible to put it to good use and work, uh, solving problems. So last year we were saying, Hey, you know, ecosystem needs to happen. Now we're seeing like a meta ecosystem, it's like ecosystem of ecosystems. And they're all kind of nicely meshing. So Madi: yeah, it feels very organic. Natan: I just hope it doesn't confuse the customers too much. I guess that's the flip side of that. Like which ecosystem do you need to be in? I guess the answer is. Madi: I think the answer should be just the manufacturing one, exactly. If that's a question, we still have to be asking ourselves. We have more work to do, but that probably is still true anyways. Yeah. And I think the ecosystem stuff, I mean, we talked about it last year. I think it's going to be something we keep talking about, but it's also foundational to supporting kind of the real evolution or. Value behind Generative AI In a lot of ways. So maybe we can come back to this and get into the, like the word of the moment. I know you love buzzwords. Natan: Yeah. So it feels like AI was everywhere and nowhere almost. That's how it felt. You know, last year we were like, where is it? Everyone is sort of doing it and talking about it, but we haven't seen it. And now It's just every booth, every platform, every product, there's some AI angle story and it's not surprising. Yeah. Because it reflects the general maturity of LLMs with chat interfaces that are penetrating every aspect of our interaction with enterprise software. Every product that is implemented in operations, whether it's backend or frontline has some data and that data has context. Yeah. So you can start talking to that data. That's the main use case. The main unlock was like, okay, we can fundamentally do more things with the data, Madi: a couple of observations. I think that to your point, a lot of the use cases are non unique, right? They're like. Yeah. Data related and the how is a little different, but in a lot of ways the how is also the same. But there were two things that I kept hearing in conversations with customers. And one was, there's still a lot of people who are still using paper and Excel. So whether it's like true or not, the perception that. Using generative AI is just like not going to happen anytime soon. Like they're, they're not even in a stage to start thinking about or talking about, like, don't say the word, the G word that came up a lot. And the other is manufacturers have a ton of data, but for. Generative AI and co pilots, which seems to be like the word we're all kind of converging on, to be useful. The data has to be structured and a lot of the data is disparate and not structured. And so getting the insights from the data There's like a lot of work to do before then, which is where I think a lot of the demos felt really shallow to manufacturers because it's like, this is cool, but like the road to this is long. You can ask, uh, Natan: I saw on demo, it's like, you can ask the machine about its OEE. Which is very cute, I think. Like, you know, anthropomorphize, anthrop whatever, how do Madi: you say that? Natan: Anthropomorphize. You got me. That one. The machine. Which is actually a very important kind of concept in user experience, you know. Dress up our Roombas and give it names and give feelings towards Madi: it. In the hotel breakfast area where we're in, they have these little robots. Natan: Yeah. Madi: That go around and pick up your dishes. Yeah. And they have faces and yeah. So very cute. inside that. I Natan: think that's, uh, Hans over there. . . Anyway, you ask this like, hay machine, what's your OEE? And I was like. We just spent a decade in Industry 4. 0 land, making sure the machine is connected and there's a dashboard and like all this like UNS business and why the hell do you need to chat to the machine? Ask it, what's the OE? And it replies the OE. So it's like we went from kind of nice, well designed in a web browser, beautiful dashboard to like, Back to text. And it looks like a terminal from the eighties, but it's in the chat interface. It's like, Oh yeah, that's what I can ask though. It's great. I can ask it though. And then I look at this, it was like, actually, you know, LLMs are pretty bad. They can't invent calculations. So I'm asking like, what are you doing here? And they're like, Oh, we have a calculator that is not in the LLM and we pass it on. It was like, so basically this is a chatbot. And they're like, yeah, this is a chatbot. That's to me, shallow Gen AI or like almost fake Gen AI. Madi: Yeah. Natan: But on a more serious note, it is true that organizations. Any organization cannot ignore the need to have an AI strategy that encompasses your systems, your processes, and most importantly, your people. Yeah. Who's going to use this? And like, there's been a whole digital transformation revolution that kind of skipped operation. By the way, we need the first class frontline operation platform for those people, blah, blah, blah. And like, are we going to have the same constituency not have access to AI? Like the rest of the engineers, knowledge workers out there, of course, rhetorically the answer is no. And that's sort of where we focus our AI efforts, right? So it's like, what frontline copilot means to us. Madi: For like, everyone is going to have some offering that summarizes information and can maybe give you some data points, do some translation. It's like, what is an immediate way that we can take those capabilities, make it super easy to set up and use and also operations first. So like AI for the operator, AI for the process engineer, AI for your plant manager while building towards being generative. And I think that that's kind of where we are. Natan: Yeah, so let's, let's break it down for a second. So a very important thing, I don't think it's lost on anyone. It's like, we're going to be living in a multi modality, like multiple LLMs talking to each other, exchanging data. And so the, the main thing you're talking about the structure and that structure provides a lot of context to give the AI something useful to learn and. What I think actually matters, which is like, how does it put to use in continuous improvement? Not just like, Oh, it's cool. Like the example I gave you, it needs to be supporting how humans make decisions and how they actually take actions to solve problems, which is a human thing. Who makes those decisions to say, Oh, we have like 10 risks. How are we going to decide what to do? So of course they're going to take data, but there's going to be some intuition, gut, whatever. And. Then this thing called analysis and you know, all the demos that we were showing in the breakdown, like, Hey, put the AI for the operator. What does it mean? It means that now we have this widget, you know, we live in an environment where we let non software engineers, but very much engineers create no code, low code application, all that. And they just drag a widget and they decide where to put the co pilot and on what data to train it. Where to invoke it and how to pass that information. And they're not bound to some other engineers who decided, well, here's how you're going to use your co pilot. No, they decide where to put the co pilot. This is an example of like bringing co pilots to people who use it. Madi: I think in a lot of ways to the, not to like quibble about language, but it's one of the We Natan: like quibbling about language. Madi: One of the things that I am liking a lot more about the word assistant when we're talking about generative AI versus co pilot is when you think about a co pilot, it's like more partnership. Natan: Yep. Madi: I think that we're elevating the AI there more than the person using it, at least at its maturity right now, where we really want the domain expertise, like the person who knows what to ask and how to use the data to be the captain here. And like, we Natan: want the person who has the data and the knowledge to set up the production line with the, you know, some people call it software. We just call it frontline operation workflows. Madi: And so there's like types of work that are like, not very valuable for people to do. It's a poor use of like, you know, the horsepower behind a person. And like going through a binder to find an answer to a question is like a good example of something that's low value. Natan: That's going to completely disappear. We had so many people saying like, look, we have all those binders. We need to bring them in. We need to have the manuals for every little thing that can happen on a production line with the equipment or the materials available. And of course now that they can do it digitally and they can have LLMs find information for you pretty quickly. But, you know, the biggest waste in operation is time. And another important principle of this lean is like, respect the people. So if we really respect the people on the shop floor, if we give them back time, there's nothing, nothing more. And like, where, how does Gen. AI come into that? It's like, If they need to get the data to make a decision, and they can do it literally in a few minutes, what would take them a few hours to like, get the stuff, get to Excel, think about it, run the, whatever regression, the interaction modalities, like make me a chart that does this and this and that, and then click a button, and it generates that. And really what it did, it turned natural language into SQL statement, and then in Tulip Analytics they see it, and then now they're in the builder, they can refine it, they click a button, they can go back to this modality, start again, click another button, and it's in the application, they hit a connector function, it's in their Power BI. That's not saving one or two hours. This could save days, weeks, months. Madi: Yeah, it's like taking smart people who are thinking critically, know the outcome they want to get to, and know what they're asking, and they're doing Natan: it themselves. And this is not different than the BI revolution we've seen like 15 years ago with Tableau. This is not about like, hey, do we think this is cool? This is about remaining competitive. So like organizations that are not going to do that, And there are going to have waste in the form of time and will have less value to give to their customers and therefore they will be less competitive. Madi: I think this is like picking at the big question coming out of this. Like you're talking about, well, what does this mean for businesses who are being very risk averse about even thinking about this right now? But I'd frame it even bigger. Like what are the risks? What comes next? Natan: We did not touch on that because, you know, we started from AI, but the other big thing we've done this Hanover essay is launch ComposableMES, which was very exciting. We had quite a splash. What is ComposableMES, Madi: Maddy? That is the multi million dollar, billion dollar question. And I would say it is, one, like the way that Our customers have been deploying Tulip for as long as Tulip's existed, but two, I think it's a really helpful mental model and quick start tool for seeing what a production system could look like outside of like the confines of your preexisting Natan: siloed software. Madi: Yeah. So Natan: if Madi: we're like defining it, I'd say there's like two definitions of composability, right? So there's the IT definition, which is definitely more about how the solution is architected and how the component pieces work together. And then there's how you think about your solution, which I'd say our definition when we talk about composable MES, and that's much more operations centric and problem solving centric. So you take the problem you're trying to solve. And if it's like, I want to understand everything that's happening across a process or production and then each of those pieces kind of ties together so that you can configure them to match your unique needs and the real value for this composable MES app suite is. It's a baseline of 21 applications across inventory, quality and production management that enables people to like get started with that system on a human readable common data model. So it's like you can get as small as one app as big as all 21 apps and they all already work together. So I think it helps to show what is possible. Based off of what we've learned from how people deploy Tulip, and the hardest thing to do is to start writing. Natan: You know what's also amazing about it? You hit a button and you download it. Madi: Yeah, the power of the internet. Natan: And people actually do that. Yeah. We just launched it and we see several dozens of downloads. Madi: Yeah. Natan: Just in the show and it's pretty cool to see. But we have to get to the part of the show where we do our traditional jargon rant. So I will lead that. Okay. Because the thing I've heard is like, people started confusing composability and modularity. And I was like, Modularity is like, take the smartest group of engineers in the room, whatever module they build for you, that's what you're getting. Composability does not assume what you need. It may give you a starting point, it may give you a good example, it may give you, this is the other John Golden Rat, it's like, out of the box. What is the box? Madi: The box is a place that keeps you locked up. Do we need a box? We need to break the box. This is the, the quiet part hearing like a lot of people talk about like composable solutions versus, you know, a lot of the legacy solutions is. You got to configure those things. And then once they're fully configured, they're really hard to change. So it's not out of the box. Natan: It's into the box. Madi: Yeah. If we're comparing it to Legos or blocks, there's some blocks in there and that's all you get. And they don't interact with any other blocks. And if you decide you need another one, you're in trouble. Natan: Yeah. That's the fallacy of like MES requires no configuration because it's out of the box and it's just false. Yeah. The configuration Madi: word is funny because it's like out of the box, but you have to configure. And then if it's not in a box and configurable, then it's custom. And all of these things are just like coded, loaded words to try to keep people from actually having solutions that match their business requirements. Natan: Okay. That was a brand. Yeah, we did the brand. We got that down. Check. Check. But now there's like an interesting integration because you were asking, what's next? Yes. I think it's where composability meets generative. Madi: Yes. Natan: And I don't think we're gonna shock a lot of people. Composable architectures lend themselves well to generative. To generate things. Yes. Artifacts. Yeah, all types of content. All types of content. And what that does to the people who use composable platforms is they give them superpowers to make things. And for us, this translates to almost every first class artifact that we have in our system. For example, what is the application? It's a format. It's structured JSON format with a lot of smart things we do around it. But guess what? We can generate that. What is our analytics? It's SQL queries, our table structures. We can generate those things. So with that, people can jump and say, Hey, I need a new table based on this material table, but it has fields X, Y, and Z that are special to us. And they need to relate to table ABC. That is over there and feed it information from connector. Madi: And you can apply that to all of the content, I think. And then Natan: all the content and what is our automations? That's finite state machine as described in JSON as well. You can ask it to like build me an automation that every time ovens go above some reference temperature, Send three emails and stop the PLC that controls the oven and no one needs to do anything other than like use natural language except the human needs to verify it. And I think our accuracy levels will increase the more we learn and have more context over time. Again, back to what we discussed earlier, it's like giving operation waste elimination in the form of time. Madi: Yeah. It's Natan: just like time to do stuff. And it's true for custom widgets and it will be true for taking data that comes externally So, you know, we meet a ton of customers who say like, hey, you know, we invested a decade or more sometimes creating thousands of processes and they may even be in a regulated environment. They may even live in some QMS and you don't want to discard them. You want to reference them, but you want to turn them into say interactive forms. Madi: Yeah. Natan: We're not that far from a day you just ask Tulip to do that for you. So stay tuned. So that's on the Gen AI. But the other thing we should be looking for is like. We talked about Composable MES, but why stop there? Madi: Right. I mean It's Sensitive AI becoming your assistant as you compose solutions to all kinds of problems. And really, like, you want to talk about boxes and breaking out of them. Yeah, Natan: so many boxes. Madi: All the three letter acronyms that we know and love, the WIMS and LIMS and Natan: And CMMS. And CMMS. That's a four. Madi: That one's four. I was avoiding that one, Natan: it breaks Madi: my analogy, but all of those guys, I feel like they're all up for being re imagined and giving, you know, the people who have to interact with all of those like individual systems, again, more ownership of how they address their business requirements and a lot more capability to make solutions that work for them. Natan: So we're in some sort of a tidal wave or a race still with a lot of high noise to signal ratio, but the future will be composed and generative. And I guess we'll have to do this again next year and see what people composed and generated. Madi: Yeah, hopefully some very impressive and interesting things to go over then, but I'm excited for what's happening right now. Natan: It's super exciting. It's super exciting. I'm glad we're not alone anymore. I'm glad the ecosystems are opening up and you know, this is actually a good ending point because we kind of talked about just about composability as a theme and generative as a theme, but they also come together and we have done that. You know, our announcement with Rockwell, which is pretty cool, is taking composable platform like ours. With its front line co pilot capabilities and marrying it with cutting edge, state of the art, HMI environment like optics from Rockwell, that basically give anyone who builds machines the ability to have a great, beautiful, HMI is very compliant with great two way communication, all that kind of stuff. But no one can say that you can't combine those two things and create really smart machines. And this is not new to us. We've done this with DMG Mori in the past. And I guess we'll share a bunch of links here in this episode in the comments to see it. So all this change and all this excitement, it's the stuff that is, it's here and now. It's here and now. Madi: Yeah. Well, until next time. Natan: Until next time. Thanks, Maddie. Madi: Thank you. Voiceover: Thank you for listening to this episode of the Augmented Ops Podcast from Tulip Interfaces. We hope you found this week's episode informative and inspiring. You can find the show on LinkedIn and YouTube or at tulip. co slash podcast. If you enjoyed this episode, please leave us a rating or review on iTunes or wherever you listen to your podcasts until next time.