Building robots that survive the warehouse

0:09Gene Demaitre: Welcome to episode 256 of the Robot Report podcast. I'm your host, Mike Oitzman, Senior Editor for the Robot Report, and I'm joined today by Brianna Westling, Associate Editor. Brianna, welcome to the show. Thanks for having me, Mike. I'm excited to get into the news of the week. Yeah, it's good to have you back on the show. We were talking in the green room. We're going to get you on here more often now. And I know, hey, I know you're a big space fan. You cover the NASA News Beat for us, reporting on a lot of NASA news. And hey, there were a couple of big space and astronomy events coming up, including a solar eclipse next week and the launch of the Nancy Grace Roman Space Telescope later this month.

0:52Gene Demaitre: Tell us what's going on there. yeah well and also in maybe less good space news SpaceX crashed a rocket body into the moon earlier this week that's right yeah yeah I think it crashed at like over 5 ,000 miles per hour so there's still astronomers still trying to determine what the impact of that will be so we'll keep an eye on that yeah I was looking at socials and the rest and no one's got a great picture yet But I'm sure in the coming weeks that we'll get some more detail on that. Right. Right. Yeah. I think the U.S. is expecting one of their satellites to go buy it sometime in the next week.

1:33So we'll see. Hopefully nothing too bad. Hopefully the moon's OK.

1:38Gene Demaitre: We knocked it off course. Now it's going to crash. But the Nancy Grace Roman Space Telescope, a lot more exciting. It's a NASA infrared space telescope. Its construction was completed last November, so November 2025. And, you know, it's going to launch later this month, August 30th. It features a wide field instrument on board. And this is a 300.8 megapixel multiband visible and near infrared camera. So the cool thing is that it provides a sharpness of images that are comparable to the Hubble Space Telescope. but it has a much wider degree field of view. So over a hundred times larger and imaging cameras on the Hubble.

2:26So we should get some really exciting images from that in the coming months.

2:32Gene Demaitre: I know it's amazing what they're doing now with the technology. Almost 20 years, I think, since Hubble went up something like that. So yeah, that is amazing. So who is Nancy Grace Roman? Well, Nancy Grace Roman, she passed away in 2018, but she was a pioneering American astronomer who was widely celebrated as the mother of Hubble. So it's very apt for them to name this after her. But she had a central role in planning, designing, and securing funding for the Hubble Space Telescope. She was also NASA's first chief of astronomy. She joined NASA in 1959, just six months after NASA was first created.

3:13So really a very pioneering figure in NASA's history. And she became the agency's first chief of astronomy and the first woman to hold any executive role at NASA. So space pioneer and a pioneer in women in STEM.

3:30Gene Demaitre: That's very cool. Yeah, great, great, great point there. You know, I love the night sky and I'll be out next week at the peak of the meteor, Perseid meteor shower, taking some images. So we'll see what I can get there. It's always fun to go out and shoot the night sky with the camera. So very cool. Thanks for sharing that. And we've got another great show today. We'll start with the news of the week in a bit. And later on the show, we'll be joined by Josh Klor. He's the general manager for North America at NoMagic. And it'll be a conversation where we talk about warehouse automation. Joshua will bring his deep experiences across robotics and AMRs, and he'll share how NoMagic is now deploying and using physical AI, vision language, action models, and talk about how they use real production data to make warehouses more adaptable and reliable.

4:25Gene Demaitre: So stay tuned for that. But first, let's talk about the upcoming RoboBusiness event. We're frantically getting ready for this right now. Steve is super busy getting everything loaded on the website. So if you haven't been there recently, we are uploading all of the speakers and all the sessions now as they come in and get them confirmed. Of course, the call for speakers has ended and we are reviewing that, filling in the last couple of slots there. We've recruited the keynote speakers. We've got a great one starting off the show, Future of Robotics, that features Rodney Brooks, Ken Goldberg, and Jonathan Hurst.

5:10Gene Demaitre: That should be a great conversation. There'll be an AI keynote panel with Andrew Davidson, Chief AI Officer for Universal Robots, and Michael Vogelsang, Head of AI Foundation Models at Collaborative Robots. And if you haven't been there, go to robobusiness.com. The event is taking place October 20th to 21st. We expect to have over 2 ,000 attendees. And right now you can save money by registering during our early bird discount. So once again, go to robobusiness.com to save a few bucks and get your ticket. Now, you know, just a reminder again that we've got you covered with all the latest news and information in our network of sites that include the robot report and automated warehouse.

6:00Gene Demaitre: So let's get on to the news of the week. And here's a story that I wanted to start with. And it's the fact that Path Robotics and Gray Matter Robotics just signed a$900 million agreement with HII. You know, this is exciting news for Path Robotics. And this deal validates the idea of putting a welder on a quadruped. We had Andy Lunsbury, the CEO of Path Robotics on the podcast a month ago on episode 252. So I would encourage everybody, if you didn't catch that episode, go back and listen to that interview with Andy. But Brianna, what are your thoughts about this idea of putting a welder on a quadruped and using it for this type of application?

6:50I mean, I think it makes a lot of sense. They're talking about building ships, so you probably have some strange parts, strange size parts, putting them in awkward situations where a traditional welder probably wouldn't be able to reach it, but you can just have a quadruped walk right up to it and use the welder. So it's a great idea and it seems to be working well for them. I know that PATH and Grey Matter have both been working with HII, the company who signed this agreement with them for a couple months now. So to me, it seems clear that HII likes the things that they've seen so far. They like, you know, the pilot programs that they've done and they want to do more.

7:34So super validating for something that I think Path is the only company in the robotics industry, as far as I know, who is trying to do this. Other quadruped welders reach out to us if you exist. But as far as I know, they're the only ones.

7:50Gene Demaitre: Yeah, no doubt. I think this is a great application. And as Andy said during his interview on the show, that this is groundbreaking for them in the sense that the ability for the quadruped to get into the difficult spaces on board the ship and actually, you know, help tack down a lot of the parts and do some welding. It's like perfect, as you just said, for what they've pulled together here and all of the technology. For gray matter robotics, their application will be sort of grinding and finishing surfaces. So it looks like, you know, this is really going to help the ship building industry to, you know, come into the modern world here, leveraging robotics to help, you know, assemble and finish the surfaces.

8:40Gene Demaitre: And it'll be one to watch, but wow, 900 million bucks, that's a big deal for both of those organizations. So great to hear about that. Now, here's another story that we reported on earlier this week. It's interesting, and maybe you can share some of the details. Google DeepMind is working with Gemini Robotics deployment to enable full body control, and they've introduced Gemini Robotics 2, which is an updated vision language action model. And everybody's talking about VLA models today and how they help expand robot capabilities for whole body control and what the robot can see. But you want to give us a few other details on this story?

9:24Yeah. So with this release, they released Gemini Robotics 2, the VLA, like you mentioned. And they also released Gemini Robotics ER2, an embodied reasoning vision language model, and Gemini Robotics on Device 2, which is an optimized VLA designed to operate locally without internet connectivity. And the big headline, I think, for this release is the whole body control, which the whole industry is talking about humanoids. It's on everybody's mind. So I think DeepMind sees that and that's what they want to focus on. They're promising, you know, enhanced dexterity, five-fingered hands and things like that, which is, again, these really big, I think, hurdles for the industry right now.

10:09And also multi-robot coordination, which if we're talking about multiple humanoids doing different work in a facility, that's super important. So it really does seem like they're sort of shooting, really aiming at these big issues that we see in the industry. And I think that it's a very like on point release from DeepMind.

10:30Gene Demaitre: Yeah, they've really making a lot of noise recently after, you know, being sort of quiet and silent for more than a year now. I think they've been doing a lot of work and now really talking about all the various companies, lots of partnerships that they've announced. You know, Apptronic comes to mind as well as another company that's working deeply with Google to advance this. And now that they've got a functional releases, as you just described, actually working with robotics companies to begin deploying these in the field. And so, you know, the race is on. Google certainly is well-funded and well-positioned to be a leader in this space.

11:12Gene Demaitre: when we talk about major foundational models and VLAs. And so, yeah, we've got this article on the site. If folks haven't heard the details, you can go learn about this. Links to all the places, I'm sure, to go download and check it out. So thanks for sharing that. And the last story that we're going to talk about today is one that I reported on this week. I thought this was interesting and it was great to sort of research this. The University of Florida has opened a new robotics lab dedicated to industrialized construction. And really the story here is not so much robots in the field as it is robots in the factory as construction becomes more modular, right?

11:59Gene Demaitre: So using robots to actually construct walls, pre-construct walls. When we talk about pre-constructed housing, the ability for them to build the walls, prefab them in the factory, throw them on trucks, take them out to the site, and assemble them real quickly in the field. So this helps in so many ways. It helps from a quality perspective, the concept of industrialized construction. It helps from speed. It helps in repeatability. It addresses some of the labor issues. So what the University of Florida has done, they have gotten some investment, a$1 million donation for equipment renovations and support from Autodesk and some other investors as well.

12:47Gene Demaitre: And this new lab is called the Autodesk Design and Make Laboratory. It'll be housed inside the Bruno E. and Maritza F. Ramos co-laboratory on the campus, which I think is a part of the overall, you know, construction and architecture department there. So the idea will be that they'll fund this lab and they're opening up this fall some new degree programs in industrialized construction as well. So for students who have an interest in this, they can go far, do research using robots to assemble, you know, from CAD drawings, from architecture drawings to actually interpret what that means to build the walls and how you might deploy robots in this construction process.

13:39Gene Demaitre: This is a nascent market that, you know, we've reported on several companies over the last year. In fact, I think we had a RBR 50 award winner a couple of times over the last couple of years for robots and construction as well here. So ongoing research, ongoing opportunity to train engineers for this space and ongoing investment. So it's great to see this and, you know, we'll continue to follow the construction robotics market as well as new innovations come together. So very cool story. Again, you can find this on the Robot Report. If you want to learn more, if you have a student or you know a student who might be interested in this area, point them towards the University of Florida as a potential university to go to finish their degree.

14:27Gene Demaitre: So go check that out. And just a reminder that we've got you covered with the latest news and information on our network sites that include the Robot Report and Automated Warehouse. All right, let's get on to our featured interview this week. Joining us today is Josh Klorum. He's General Manager at NoMagic. And, Brenna, you want to set this up a little bit for us? Yeah, so this conversation, it's really about the shift from traditional warehouse infrastructure to more flexible edge automation and what it really takes to move from pilot to production. So NoMagic is focused on bringing physical AI into warehouses through each picking, recovery workflows, and service forward automation that stays reliable in production.

15:14That reliable is like a really key word. The conversation covers the future of warehouse automation, including why lights out operations are possible in some areas, but still likely to need humans for maintenance and exception handling. So let's listen.

15:34Gene Demaitre: Well, Josh, welcome to the podcast. We're curious to learn today about yourself and no magic. So why don't you start there? Tell us a little bit about who you are and what NoMagic is doing for the industry. Sure, yeah. Thanks for having me. My name is Josh Klor, managing NoMagic here in North America. I've been in robotics from the beginning of my career, so started with studying engineering at Georgia Tech. Actually integrated a lot of camera systems, robots early in my career from the engineering side, and then was the third employee in the U.S. for mobile industrial robots, Mir, and and really was there at the beginning of this new AMR expansion into the market.

16:19It didn't really exist until Mir and a few others started that trend and it's become huge today. Then I went over to Mugen and helped Mugen get started in North America here as the VP of Sales and marketing. And then most recently at NoMagic here to help grow AI into the warehouse through each picking primarily is what we're focused on. So NoMagic is really a robotics technology company that is heavily invested in the latest and greatest research and development of AI within the warehouse. Really this term physical AI is something that we embrace and something that we are truly bringing to reality.

17:05And yeah, happy to dig in with you guys and tell you a little bit more.

17:10Gene Demaitre: Yeah, I think that's exactly where we want to talk about today on the show. And taking a look at the warehouse over the past decade, a lot of decisions about investing in heavy infrastructure, things like ASRS shuttles, conveyors to move product around, to bring it out for picking and packing. And, you know, in your view, you know, how has the focus shifted now towards, you know, the connective tissue, things like robotic picking, packing and sorting? Yeah, that's a great question. You know, I think it was, I saw some time ago, I think it was Eric Nieves from Plus One speaking about the center of the warehouse, the automated center of the warehouse and that that was really the focus over the last really 20 some odd years in warehousing.

18:04You know, everyone was going to get some shuttle system or ASRS or mini load to help automate the storage and retrieval aspect, but not so much focused on the edge. And I think that's changed over the last decade where we see a lot more things like decanting, depalletizing, going into these systems at the case level, from the pallet to the case level. You see a lot of systems starting to come to bear for the mixed palletizing world, especially in that grocery market. Some through big integrators like Vitron and SSI Schaefer, some with smaller technology companies that are doing more flexible mixed-paced palletizing.

18:44But you also see what we're doing at NoMagic here, which is the each level, break pack, split case, each piece picking. There's a lot of names for it, but we're talking about picking items in the warehouse, primarily supporting e-commerce or the high variety retail areas. And yeah, I think you see a lot of growth in that space over the last decade. A lot of companies focused on it, both on the induction and dissortation side, but also directly connected to that goods to person station. Now it's a goods to robot station. And the technology has progressed very rapidly over the last decade. First, you had this wave of vision hardware that created more data points for us to learn from and how we actually can find the right thing to pick or the right place to pick something.

19:35And then you see this wave of reinforcement learning and machine learning where you're training a data set and a robot is really kind of learning from that data set on how to pick these items. And now we're trending more towards even newer technology like vision language action models and world models. And I think you see this incoming wave of humanoid and two-armed systems that are going to do more nimble and dexterous tasks like receiving. So there's a lot of new things happening and a lot of things that have happened over the last decade that support a huge growth in this area of robots on the edge of your warehouse doing a lot of different tasks that had to be manual in the past.

20:27Well, Josh, now Magic has talked about that this is all in service to the idea of the always-on warehouse. And can you explain to us what you mean about that? I know we touched on this in our webinar on automated warehouse back in June. But how is the always-on warehouse changing the stakes for autonomy? And what are your customers' expectations? Yeah, I think one of the challenges with automating the warehouse from the beginning has been that it is a lot of times designed as kind of a batch flow system as opposed to a continuous flow system. And with robotics, we really like to do the same thing over and over again if we can and keep the throughput high and keep the availability high and continue to do our thing.

21:20We don't need breaks. We don't need to stop for any reasons except for maybe some planned maintenance. Right. So I think that is something that we really lean into, both in the way that we structure our contracts and the way that we work with our customers. We want to move towards that 24-7 continuous flow with the warehouse to help optimize the use of that space and the use of those resources. So really, we can look at it as time to trying to also optimize the bigger system that was a big investment, the ASRS or whatever sortation system that you have. You want to keep that thing running as often and as much as you can.

21:57So the robots can support this journey from batch flow to continuous flow. And one way that we support that with our offering is through this human in the loop approach. So we have our happy path where we're doing vision kind of guided movements and picking and placing things and it's moving very fast and efficiently. And then inevitably something will happen and the robot can't recover within that happy path. Of course, it can do retries and that kind of thing within the happy path. And we have two levels of autonomy after this, right? We have the vision language action model where we are observing with AI and trying to figure out, okay, is this something that I can recover from with this real-time input to the system and seeing the problem and in real time solving that problem as opposed to, you know, making a motion plan and picking and placing real-time movement of the robot grabbing something and solving an input to the system as the issue, right?

22:55And then the next level is that human in the loop, right? So jumping in with a human to say, hey, robot, we can see what this problem is. Let's solve it this way. So really in support of that always on warehouse, we want to make sure that we're giving you a service as opposed to just a system to put in place.

23:13Gene Demaitre: Yeah, I think you maybe preempted the question I was going to ask you, but I'll maybe ask for a little clarification in the context that, you know, in your experience, and I'm glad you started with your rich experience across the industry here, that historically many robot pilots failed to convert to operational scale. We've seen that, you know, happen in several things. I'm sure you've experienced that various stages of your career with clients and projects. What is NoMagic doing differently to help ensure that customers are successful moving from a pilot to production? Yeah, Mike, I've for sure seen a lot of these situations.

23:50And going back to the early AMR days, regardless of the vendor, there was this localization problem, right? As things move around after you leave, after you've installed it, okay, we moved some things around. Now we need to take that into account, into the map and be able to do this. And this was a challenging thing back in those days when AMRs were just being adopted. Sure, to some extent, it's still a challenge today, right? So seeing those situations, and I think, you know, how do we solve for these types of problems? It's being very close to the system and understanding how it's operating on a very, very regular basis.

24:28So not only do we have the human in the loop, but we're also reporting on all of these metrics consistently within our service team to where they know that, you know, hey, there's this potential for this, you know, on site issue. And we're proactively going out to the customer and saying, you know, you need to change this or you need to do that. Or, hey, we're going to come and change this or we're going to do that. So always kind of creating this regular plan of attack to keep the system with high availability. That's the name of the game, right? To keep the system available for whenever you do want to run the overall system there.

25:07So when we're talking about this production data, how does that help improve both the physical AI and then ultimately the capabilities of the robot arms in these warehousing and logistics applications? Because I have the sense of a continuous improvement loop. And how do you make sure that you're going all the way from what's happening on the floor to that physical AI and then back to the robots? Yeah, it's a great question. And it's a core tenet for us here at NoMagic that our real moat is this kind of production data, right? There's a lot of companies that are out there today that are doing, you know, they're building their models with simulation and they're building their models with not production data in the real world from real use cases.

25:57And we see this as a shortcoming of the overall market, that it will be a big challenge for those types of companies to solve the long tail of edge cases that you see when you get a real world system in place. So we call it mastery first. AI is really what we're promoting, that we're going deep into use cases and trying to fully solve them for all of the edge cases that you might see. And then that data for us is very valuable. It helps us to fully understand, you know, what happens in the real world when you try to apply robotics to these use cases. And we can train off of this to create more capable overall systems, right, to build a system that has these edge cases in mind.

26:46Now, the use cases might be different or the edge cases might be different for product type or workflow. But a lot of these learnings can be generalized to help support a system that can be more available within that happy path. So you're not always having to go to the human in the loop.

27:04Gene Demaitre: I want to come back. You hinted at this earlier, so I want to drill down now a little bit more on the concept of how you're deploying vision language action models. You know, this has been a concept that's been talked about for the last four or five years. And now I think making it into production systems and, you know, as you've already hinted at, you know, real world warehouses are chaotic. You know, there's things are different every time you look at things. Certainly, if you're doing each's picking, the chaos is in, you know, what the disorder of the box you're pulling bin, you're pulling things out of to singulate them for custom order, those kind of things.

27:44Gene Demaitre: So give us a little bit better detail now on NoMagic and how you're using, you know, visual language action models to allow the robots to adapt to what's happening in real time. Yeah, so we're using VLA's primarily as a secondary workflow at the moment. So we see this as a great way to take what we're doing in our labs and R &D and to bring this into production is to not make it the primary objective for the robot, but to make it a way, a means for recovery. So it's actually already deployed in production. And the main use case that we're going after at the moment is where the robot has packed an item into the tote or the box, and it didn't land where it was supposed to.

28:32Maybe it's angled in a weird way or maybe it's above the lip of the tote so it can't go back into the system that way or it can't go into the boxing, the lidding machine if it's a box that way. So we have censoring technology with AI that's constantly looking for these types of scenarios and flagging the system to switch to VLA mode, essentially. And then the robot can drive down and solve with that input that, hey, this is the problem that we have. Go and try to figure out where it is and how to solve it. And in real time, instead of building a motion plan to solve it and where to place it, a packing plan, it is just taking that language input using the vision that it can see and creating a step-by-step action plan on its own in that loop to solve that specific issue.

29:26Gene Demaitre: So that essentially, I mean, it helps reduce interactions from the human side to recover something, as you said, might jam things up elsewhere in the system. Yeah, it's a huge value overall, right? Because it means that no one has to jump in from the human in the loop perspective. And one of the big value drivers is how many people do you need per robot to monitor these systems. Additionally, if it's something that can quickly be solved with VLA, then you don't have to have this person waiting for this notification and jumping in. So you can solve the problem quicker as well. Josh, you had mentioned the challenges of each picking.

30:11once no magic's approached training its ai brain on billions of product configurations and then sharing that across different deployment sites and customers yeah so we start with um you know we've built up a more general model of of what uh different types of products are and classifications of products based off of what we've seen in the real world with our production data we start there so we can immediately begin picking when we come on site. That is something that is fairly new over the last few years, companies being able to just show up and start to pick and not have to actually train specifically on your data, on your items.

30:52And if the customers are open to improvements based off of learning from their data, then we can additionally learn as we go in this kind of ramp-up period to get better uptime, better throughput, better understanding of the product, right? So a simple use case, for instance, learning that, hey, this type of box that you have opens up like a pizza box and we can recognize this. Okay, well, we can learn on your specific data that these are the types of pizza boxes that you have in your SKU set so that we can pick them in an appropriate way and not open them up and have to have this operator interaction to solve the problem.

31:38So primarily, we have this general model and then we train on your specific data to help improve how we work on say. And in addition to talking about the software and AI and robotics and robotics as a service, you mentioned going directly into the sites. Are you also working with integrators? Yes, we work with some big integrators. We have a named partnership with the Wanderlunda, for instance, and we work with others as well. And really, it's a natural thing, right? They sell the big systems that our robots connect to and they want to offer an overall full solution. And for us, we want to work with the trusted integrator partners in each market, right?

32:24So it's a natural partnership for us to work with bigger material handling integrators. And also, I'll say, you know, the mobile robot companies that are out there, you know, that's another natural place for us to create partnerships and something that is, you know, an evolving way that people are buying, right? customers are starting to buy their ASRS or sortation technology in a more flexible way, especially for brownfield, especially for smaller volume sites. And, you know, sometimes those are the best use cases for robotics.

33:03Gene Demaitre: The way I understand NoMagic, and this is where I want a little bit more clarification, is that you're adding an AI layer to existing hardware and storage systems, like an auto store port, for example, rather than focusing on designing and building new robotics systems. So why do you think this approach is more scalable than some of the tactics, strategies of other companies? Yeah, yeah. So I'll say we have kind of the core elements of a platform, right? Of a software platform for robotics that can do each picking specifically. And additionally, we offer some robotic solutions that we know are market ready, you know, something to quickly bolt on to an ASRS, something to work with these types of mobile robot systems, something that we've done for Zalando in Europe specifically is a pocket sorter induction station, right?

33:58Something that's ready to go up against a pocket sorter. So I would say while we have the software layer, the AI layer, the thing that makes, that's the special sauce that makes this work, and we have the service element, really we're providing solutions to the market with these elements. Well, we know to enable that you raised Series B funding early this year, but since then there have been some really interesting breakthroughs like handling shoeboxes, for example, which might open depending on how you handle them. How challenging was it to develop systems that can manage these different types of containers?

34:39and how does that also then play into dealing with returns? Yeah, so I would say shoeboxes have been called kind of the holy grail of fashion intro logistics. It's a very tough challenge. It is, as you said, there are many different types of shoeboxes. They can open up in a variety of ways. A lot of times we even see that the shoeboxes are open a little bit. as they're moving around in the totes, coming back and forth from the storage and retrieval systems. So a lot of things to account for when you're trying to do this type of induction or order fulfillment with shoeboxes. We have a cross-functional team that built this.

35:24So we have an internal team that's focused on gripping technologies that's cross-functional from software to hardware. And this has been something that we've done over the last few years and has very hardened at this point. But it took a long time to get there. I think we're version 8, something like this, for the Gripper hardware. And it's something that took a lot of iteration and cycling to get there. And the software piece of it is not trivial as well. It's the vision technology is there to see these things, but really comprehending where on this box can I pick or can't I pick based off of what it is and the angle that it's in is a very challenging problem to solve and something that we've solved and it's been in production for over a year and something that's going really well and is being rolled out across multiple sites and multiple customers.

36:29The returns piece of it is a natural thing. when you are dealing with fashion, especially fast fashion, where there's a lot of returns coming in, there's the quality checks and that sort of thing. But once they are ready to go back into the system, usually everything is just kind of thrown into the same toad or the same trolley or something and needs to be inducted. So when we're picking up shoeboxes in these returns applications, There's also mixed apparel all around. So we have to have a tool changer there where we can pick many different types of fashion items and also pick shoeboxes that are oriented in different ways.

37:12So it's a tough challenge and something that we have in production. And it runs really well and really proud of the team that built it. Well, Matt, lucky Matt and Modix this year, NoMagic showed a bunch of different systems. Can you explain for a moment how you're using Spot or real-time vision to audit packed orders before shipping? Yeah, so Spot's a great product that we have. It's a vision-only product. So, you know, a lot of folks think of us as a robotics company, but really we are an AI company that has a heavy focus in robotics. And Spot is just a vision sensor that you can put in between your pick and your pack lines to be able to do item counting, order verification.

37:56It's something that allows you to have a lot more data about what's going on in your building without having to scan every single item that leaves the door. So you get a good sense of what is your error rate. You get to flag potential orders that have errors in them and something that can help you stop errors from getting to your customers. And so you're talking a little bit about that differentiation again. How would you say that you're different from sort of the classic warehouse automation suppliers that are, you know, everyone's trying to optimize their performance or success with their clients.

38:35And when we're talking about AI adoption, why is operational accountability so critical? Yeah, so I would say we're different because we're solely focused on this AI technology and bringing solutions to the market in a way that is service forward. Right. So making sure that what we put in to the market, into operation adds value for our customers and is something that they not only have have signed off on, but are continually signing off on. We put penalties in our contracts when we're not meeting KPIs on a regular, you know, on a monthly basis. We evaluate this. We really have a mindset to make sure that our customers, you know, want our products and continue to want to use our products.

39:28We, you know, something that we can claim that I think very few can claim is we have zero churn within our deployed systems. We've never uninstalled a robot. We have had a great success in our execution and service programs.

39:43Gene Demaitre: I think that's a great metric and certainly one worth shouting about because that's certainly not true for all systems across the board. But as we look to the future, there's always this. You talked about one holy grail problem in terms of handling shoeboxes. boxes. But the other sort of future that a lot of people talk about is the idea of a lights out warehouse. And I don't know, do you still think that that's realistic in the future? Or do you think there's still going to be, you know, humans at various points? I think you've hinted towards the, you know, the interaction and collaboration between humans and robots in some of your prior answers.

40:23Yeah. And I think it depends on your definition of lights out will dictate how far it is into the future. I think to some extent, there's already lights out warehouses, right? There are warehouses that the primary motion of the warehouse is done in a fully automated way. Now, there are still people that do the maintenance. There are still people that jump in to support the robots, wrangle the robots when necessary, right? So that's not fully they need lights for those things. There's not robots managing other robots and doing maintenance on other robots. And I think, you know, personally, I think that that's far out into the future.

41:03You know, robots that are capable of these, you know, highly generalized tasks that have a broad depth of requirements and broad range of skills and capabilities that you need to fulfill, like maintenance across multiple other types of systems. I think that's a very human-oriented job and something that would be difficult for a humanoid to master in the short term. I think, again, we come back to this mastery first idea. And I think if you think about every piece of the warehouse and focusing on deep diving into how to make it the most optimal, then it is ready today. and there are warehouses today that already do more or less the entire system in an automated way.

41:53And I think what's changing is how do we do it in a more flexible way to help, how do we go lights out in a more flexible way to help with that cost burden? Because most of those fully lights out systems are big warehouses with really big expensive systems. But the new way of mobile robot companies, I think is helping with this. Technology companies like us are helping with this to do the edge of the warehouse and it's becoming more and more of a reality. So instead of lights out, I feel like we should have some type of metric that tracks the level of automation across all warehouses and how do we raise that bar as opposed to focusing on the individual cases where we've spent a lot of money to make the whole, you know, big warehouse, you know, lights out, so to speak.

42:49But so, yeah, I've seen a lot of manual warehouses out there. And I think just taking the known steps that you can today is something that everyone should be focusing on.

42:59Gene Demaitre: Yeah. I mean, on that same thread, I mean, in your role, you get a chance to talk to many of your potential clients, existing clients as well, I'm sure. But I'm really curious about your observations from your seat in the industry about, you know, what are some of the unique pressures that you think North American supply chain executives are facing today? And maybe in that same light, you know, some of the smaller to midsize operators and their ability, as you just said, from going from a manual operation to some autonomy to stay competitive. Yeah. You know, to some extent, the job hasn't changed.

43:38You have to get the stuff through your building when people are asking for it. And that demand has changed for sure. That side has changed. And the available technology at your disposal has changed, right? And, you know, a lot of we see a lot more humanoid technology. We see a lot more overall robotics technology out there in the marketplace. And to some extent, you know, these operators, they don't have time to do the heavy research to figure out what's the right thing to do now versus what should they be potentially waiting on for the future. and I think it's our job as robotics technology companies to educate them on what is possible now and what does the future look like at least in the next you know three to five to ten years we should have a an idea of this and and what I'll say is you know I see a lot more companies you know focusing on you know while we have a lot of ASRS technology that's been deployed already There's a lot of companies that are just now getting to this in that midsize range in terms of volume.

44:52And I see a lot of companies that have already done this focusing on what's next. Is it the autonomous fork trucks or truck unloading or mixed palletizing? So how do they make those decisions? I think it's good to get advice from your integrator that you're already working with. But also make sure you talk to the robotics technology companies, because the difference is that we spend all of our time focusing on what is happening in the market and what's the latest and greatest technology. And the integrator splits their time across many different technologies. So I think you should both deep dive in your area of interest and talk with your more generalist, with your integrator, to get the best of both worlds and figure out the right path forward.

45:38Well, and Josh, to follow up on that and to wrap up, what's one piece of advice you have for supply chain leaders who are looking to evaluate physical AI, sort of getting back to the AI piece, without them getting caught up in all of the vendor hype at the same time? Yeah. I think one of the biggest challenges is, you know, the economic buyers of robotics technology are there. They're treating this still like it's a very, very large capital purchase that they're, you know, going to go through some 10 to 20 month decision cycle on. And in 20 months, the technology has changed, right? So I would say try things faster and work with vendors that are willing to help you de-risk the first kind of stage of technology adoption and not make it some huge capital expense, but to do it more as a service in an ongoing way.

46:42That's the approach that we take. That's the approach that others take. and you should be moving forward with something that makes sense, right? So instead of spending a lot of time and treating it like a capital purchase, try to just figure out what you think makes sense and make it work for you and de-risk it as much as possible on the front end and understand that you're going to have to make a lot of decisions along the way over the next few years about what robotics technology works for you right now in this current moment.

47:15Gene Demaitre: Well, Josh, thanks for joining us on this show today. It's been interesting again to hear your observations from where you're sitting and to learn more about NoMagic and what NoMagic is bringing to warehouse operations going forward. So, thanks for being here. Thanks for having me, guys.

47:36Well, that was a really in-depth look at warehouse automation. I love how he defined only on to mean continuous flow and not just zero humans. Josh's version of the always-on warehouse is one where robots keep systems running continuously, but humans still step in for maintenance and exception handling. I thought that a strong moment from the interview is his perspective that fully autonomous warehouses still need people for maintenance and exceptions. Mike, what were some of your takeaways?

48:06Gene Demaitre: Yeah, no, I agree. I think that, you know, He makes a strong point there about our path towards the concept of lights out warehouse and what that means. And, you know, of course, the other part that NoMagic is doing is gathering a lot of data and they think of their production data as the moat for the deployment. And, you know, the other part of it, too, is their advantage is they're training their physical AI models using real-world deployments, not just simulations. So again, they're crossing that bridge as well. And that helps, you know, make their solutions more accurate. And so when we talk about, you know, the idea that their moat is their production data, I think that's a difficult one for, you know, a lot of competitors to cross as they continue to build experience and expertise in the rural world here.

49:00Gene Demaitre: And, you know, I think warehouse automation, as we've been tracking on Automated Warehouse over the last couple of years, moving from flashy demos now to really productive systems that can handle exceptions, can continue to meet KPIs and the ability for the systems to continue learning, continue evolving and to do this in measurable ways. So again, I want to thank Josh for coming on the show this week and sharing his perspective. And he knows it was really an opportunity for me to learn, for all of us to learn just now, you know, what NoMagic is doing here in this part of the space. So thanks to Josh.

49:40Well, before we go, we have just another reminder about the upcoming RoboBusiness event in October. It's going to take place October 20th and 21st. And this is our 20th anniversary show.

49:52Gene Demaitre: Yeah, right. Yeah. 20 years of RoboBusiness. We'll be looking back at how the industry has changed over the last 20 years. It's a lot. We're expecting over 100 exhibitors, over 2 ,000 attendees, and you can register now to save money with early bird discounts. So don't wait. Get that discount. And to learn more, you can go to robobusiness.com. Great. Thanks. Well, hey, Brianna, thanks for joining me this week. It's always great to have your voice on the show. I love all your reporting, so thank you for that. and that's going to do it for episode 256. Thanks again to everyone for tuning in. Thanks again also to Josh Clore, general manager at NoMagic for joining us today.

50:36Gene Demaitre: Brianna, thanks as always for joining me on the show. New episodes of The Robot Report drop each week. You'll find us anywhere you get your podcasts. Please subscribe, leave us a rating, but most importantly, tell your friends about us. That's how the show grows. So share this episode with someone who you think might benefit from listening. The show was produced by me, Mike Weitzman, and edited by Tyler Kubisti. Our executive producer is Amanda Buhner. Thanks everyone for listening.