Opportunities for Robots in the E-Commerce Supply Chain

    Tune in for a discussion with Ken Goldberg, Prof. of Engineering at UC Berkeley and Chief Scientist of Ambi Robotics, and Jim Liefer, CEO of Ambi Robotics on the tremendous potential for robotics across the e-commerce

    Jeremie Capron:

    Hello, everybody, and thank you for joining us today. My name is Jeremy Capron. I'm the director of research here at ROBO Global. I'm talking to you from New York city. Today we're going to focus on one of the most exciting areas of robotics and AI, something that our research team is very passionate about, that is logistics and warehouse automation. This is a very important topic for investors in robotics and AI, not only because it's already started delivering impressive investment returns, but also because it touches on many of the crucial aspects of the robotics revolution. We're going to be talking about some of the enabling technologies that make logistics automation possible. The sensing, the computing, the AI, and also about what's happening in terms of businesses adopting these technologies across increasingly large spectrum of industries. And to do that, I'm really thrilled to be joined today by two very special guests.

    First, professor Ken Goldberg, who supervises research in robotics automation at UC Berkeley. Ken has been researching robotics grasping and computer vision for many years, I think more than two, if not three decades. Ken is also a co-founder of Ambi Robotics, that is a startup that just came out of stealth mode a few months ago with a very interesting solution for supply chain automation with robotics speaking. Ken has been an advisor to ROBO Global for several years. And for that, we're very grateful. Ken Goldberg, thanks you and welcome. Where does this webcast find you at today?


    Ken Goldberg:

    I am in Mill valley, California.


    Jeremy Capron:

    Great. Our second special guest is Jim Liefer, who's the CEO of Ambi Robotics. Before joining Ambi earlier this year, Jim led Kindred, a very successful robotics company that focused on e-commerce an order fulfillment solution that was acquired by one of our portfolio companies, Ocado, last fall for more than $250 million. And before that, Jim had a long career in logistics including with Walmart and UPS. Jim Liefer, thank you and welcome. Where are you talking to us from?


    Jim Liefer:

    Thanks for having me today. I am in Emeryville, California at the Ambi Robotics headquarters in California.


    Jeremy Capron:

    Okay. So we've got both coasts covered today. Well, look, let me start with a question for you, Ken. I know you and Jim have some materials you want to show and some really cool videos of the technology in action, but here at Global, we have this view that it's important to invest across the entire value chain, not just the robotics, the robot manufacturers, but also the enabling technology providers. I think what I want to ask you is what's happened in the last few years that's enabled those dramatic improvements in terms of the robots dexterity and their ability to pick random objects? And why is this such a difficult problem in the first place?


    Ken Goldberg:

    Okay, great. Thank you, Jeremy. And thank you for putting this together. It's really a pleasure to be here, especially with Jim, who's the consummate CEO, the perfect CEO for our company, and I'll explain why later. What I wanted to do is... The question that you're asking is actually very subtle. What I'd like to do is share some slides of perspective and then circle back to why that problem has been so difficult and why we're making progress today. So if you don't mind, I'll start with some slides if I may?


    Jeremy Capron:

    Go ahead.


    Ken Goldberg:

    What I think is so interesting right now is that so much has changed in the past decade and then again, in the past 18 months. What we're facing right now is, there's a massive change that's happened in robotics and automation that is been catalyzed by COVID-19. This came out of left field for almost all of us. We didn't expect that something like this could completely shut down almost all travel and so many businesses and disrupt so much of our lives. And it's had a huge amount of effects in robotics. It's changed perceptions of how robots are viewed. We're now seeing them in many different environments like airports and home, in retail because people see them now as a safe alternative when there may be dangerous viruses around.

    Robots also came to the rescue in terms of providing some level of cleanliness and being able to spray areas when there was a lot of fear about airborne aerosols that could be harmful. They're also being used to disinfect things like stadiums using drone helicopters, which is a brand new application that people had not been thinking about before. And of course, in healthcare, robots have played a very important role where many patients are highly contagious and families are not allowed to go in and see them, but robots are able to provide both companionship in certain ways and also a communications medium so that family members could contact them and have interactions. Obviously not as good as being there in person, but given the safety considerations, it's much better than just having a phone call.

    Now, speaking of which also there's been a real change in the way healthcare is delivered. Doctors were forced essentially to have visits, many of them through teleconferencing. What many people discovered was that those were actually quite effective, that you could diagnose many different conditions over the phone, over a Zoom call, say, and that has changed perceptions and behavior in ways that we believe are going to be persistent. That tele-health is going to have a huge impact in the future. 

    The other is delivery. This was surprising too. There were a small number of startups that were looking at wheeled vehicles for doing delivery of food and products say on a campus or in a small neighborhood's environment in, say New York. This took off as well, because there was a shortage of workers and a huge demand and surge in demand for home delivery. Now, speaking in which this is the one that we want to talk to you about today, which is the huge, massive increase in e-commerce from organizations like Amazon and Walmart, and a variety of many other providers. Now, this is not atypical. This is actually a shot on my front porch, and it was a typical scene during the pandemic.

    What's interesting is that e-commerce has been growing for the last decade and on a steady increase every year. But this year, because of the COVID-19 conditions, e-commerce have a huge increase, huge search. And this is for obvious reasons. But what everyone believes is that this is not going to subside. That once people have understood the power and the convenience of e-commerce, that they're going to... They've fallen into new habits. They've developed the idea that if you want to buy toothpaste, you just order it because it's going to be there faster than you're going to remember to go out to the pharmacy and go get it.

    Now, the challenge is how do we actually meet that demand? This is a typical scene that is going on as we speak in warehouses and logistics environments around the country and around the world. It's very important to realize that this is currently a very human centric set of operations. Humans are essential for what you see here, which is manipulating the packages with all the variety of shapes and sizes and complexity that they have. This has been elusive for automation. And it's particularly been tricky now because of COVID-19, because now there's been a shortage of workers because of the virus, also need for distancing in warehouses that wasn't anticipated earlier. So there's a real anticipation, a real push for automation here.

    Now I want to turn it over to Jim, who's going to talk about the specific case of Ambi Robotics, and then I'll come back and talk a little bit more about the global trends, and then we'll go into the questions. Okay. So, Jim, can you take it over from here?


    Jim Liefer:

    Make sure you guys can hear me okay. Get off mute there.


    Ken Goldberg:



    Jim Liefer:

    Thanks very much, Ken, for the setup there. Yes, I want to talk about Ambi. But I'm going to start here with this the promise of e-commerce. This move to this direct to consumer model, it's been underway for many years, as everybody here knows. We've talked about this a bit here in that the pandemic that it greatly accelerated that movement, and Ken has talked about, it being something like a five-year acceleration. There's examples that I have, we can use grocery as one example. In that sector alone, there's data out there that show that the total ring for grocery, it saw about only 3% of it went to e-commerce for that total ring. And then after about March 2020, it jumped to nearly 35%.

    Even as we come out of the pandemic, thank goodness, I don't see that number going back down to anything like less than 20% or whatever it is because people now understand that that's okay to shop in that way, just in that sector alone. And while all of the e-commerce promise here provides all these great benefits that you see on the side to the consumer, there is an impact that's happening behind the scenes that's pretty tremendous that people don't realize. So I want to talk about that, what that impact is a little bit. What most people don't understand is that the corresponding shift in fulfillment centers from these things of what I call case pack quantities, which are boxes of products that are going to stores to the shift being into eaches. And that shift from those case packs to eaches requires manipulation of those individual items in fulfillment centers so that they can be picked and packed anFd shipped to those end consumers and in many cases, even to the store.

    It's the move of that each picking and placement that's really driven up labor and fulfillment centers and shipping centers to about 400% increase. That explosion of handling those eaches, it's really been a primary driver for retailers and shipping companies that they have to look for innovation solutions like Ambi Robotics. Because of the technology, and Ken can talk about that more, we were to build on this foundational work that was done at Berkeley with Ken and the founders here at Ambi Robotics. It allows us to move much faster to deploy our robot configurations and solve our customer's needs, which is manipulating products and packages. We fully recognize that the pain that our customers endure. They enter this pain to find good people, and not only to find those people, but to keep those people.

    Our systems perform much more mundane, repetitive tasks within these environments that still have levels of uncertainty. Like various packages coming down a chute or a mixed input bin that need to be placed. You saw some of those images from the sites that Ken have. Not only is it remarkable that so much has gotten done, there's this compression of time that has happened at Ambi Robotics. So the team here, the founding team at Ambi was with Ken in Berkeley, working on these technologies, and that was in 2018 and before. 2019, the team came off campus, founded Ambidextrous Labs. We've now rebranded as Ambi Robotics partly because the difficulty of just spelling Ambidextrous by the way. But aside from that, they began conceptualizing the solutions in 2019. And then in 2020, when the pandemic hit and there was a lot of people going to the bunker, if you will, this team was figuring out and reaching out to various partners out there in the business world and developing systems, piloting those systems in 2020.

    As Ken said, and Jeremy, I started in 2021, early 2021 here at Ambi. We came out of stealth and secured these customer agreements, and we've been deploying these systems out there in the world. Again, it is this underlying technology that allows us to have these various configurations. And you can see here in this animation, the way that we think about the solutions for our customers is that they have to have a high degree of configurability. That is all of these fulfillment centers, sortation facilities and all the rest, they're doing actually the same thing, but they're doing it in slightly different ways and sometimes major different ways. Our hardware and robotic technology is able to adapt to those configurations so we can put them in between various mechanical automation that's already in place.

    We have three core components to our technology, again, that we built on over these last compress number of years. There is the Sim2Real component, which allows us to do all of the training of our AI 10,000 times faster. We do all of this in the simulation world. We have this generalized approach so that when we implement the ability to operate in the physical world, works on day one. The other part of this is that we have an approach where the hardware is quite modular. We can adapt to those different inputs and outputs that I talked about. We can also, if you think about the total solution, it's not really just the robot, it's all of the peripherals around it. There's scanners and cameras and lighting. So we're able to use those commodity components and bring them in as part of the solution.

    Then the other part of that middle tier is that we make sure that our approach with the robotic systems still has the people in mind, has the human in mind. Because we believe that you need that human there working with the robot to make sure that you have the throughput and all the other things are working. And then lastly, with robot dexterity where not only are we able to reach deeply into a bin like you see in the image here, but we can also deal with shoots or conveyance and those kinds of things. We're able to have different end defectors. And then lastly, we work with imaging technology as well, so that once we have something and we're moving it, we can understand what it is because it doesn't help that much if you don't know what you're holding.

    When I talk about the team a little bit, you can see the ones with the yellow rings around these people. This is the team that we've put together at Ambi Robotics. The founders here, Jeff Mahler, who is our CTO, Stephen McKinley, David Gealy and Matt Matl, all of them working with Ken closely to develop this technology. And then what we've done here at Ambi is rounded out the other verticals of the business, because it's not just about technology. We're solving the business problems that are out there. So we put together the manufacturing team, the implementation and support team, finance team, and those sorts of things so that we can get the right solutions out to our customers.

    This is just land on this page. There's a lot going on here at Ambi. We are scaling this business and we're getting our solutions out there in the world. There's a lot going on in the media. There's much more that's happening soon, but not quite yet. So my marketing person would be angry at me if I said more than showed you this page. So I turn it back over to you, Ken.


    Ken Goldberg:

    Okay, great. Thank you, Jim. Great job. To your point, we're out of stealth, but there's a number of major developments that we're just about to announce. We have to be a little careful about the timing of that. Joe Ruck has been our VP of marketing. He is tremendously active in guiding us in that aspect.

    What we wanted to show you is a video of one of our systems that we call Ambi Sort, and what this is doing, what this is going to show you is this system that combines three steps. One is to pick objects out of a bin, a deep bin in a real commercial environment. And then it has to scan those objects. Then it has to place the object according to the scan value into another bin. That's what we call sorting. This is a very, very essential part of logistics and e-commerce. Again, it's almost entirely manually done today, but it can be very tedious and actually injury prone for workers. So we were brought into to address this, and here's a video showing what our system or our commercial system looks like.

    Okay. So just before we get back to your question, Jeremy, what I wanted to note is that this article came out a couple of weeks ago, basically saying that post COVID, we're coming out of this, and there's a huge surge in robotics across the board and many different automation industries. One thing I want to also note is that the United States has just committed $250 billion to enhancing research and technology in this area, not just robotics, but also semiconductors. But they list AI and robotics as one of the key areas that the Congress is going to be supporting both in research and in small businesses. So we're really excited about that and glad to see that being recognized. Again, this is huge major shift in consumer behavior and how this is a huge opportunity for robots to address this.

    We're happy to take some questions. Again, I will try to answer your initial question, Jeremy, which is why is this hard and how are we able to do it now? Partly it's because grasping is something humans do very easily, even from a very early age. A few weeks after being born, humans are able to grasp. This is actually so natural for humans, but it has been incredibly elusive for robots to be able to pick up arbitrary objects. And it's because of one central factor, which is uncertainty. A robot cannot perfectly perceive the environment. It doesn't know the three-dimensional geometry in front of it with precision. Number two, it has errors in an uncertainty in its own motion. It's very hard to know exactly where the robot grippers or section cuts are in space. There's some uncertainty there.

    Then third one is, there's uncertainty in the physics of grasping. So when you pick an object up, there's things you don't know, like the center of mass of a box or the formability of a bag, and those are all of these combined together to make the job very tricky. Humans are able to account for this in ways that we still do not understand. So a human can reach in and grab a bag and somehow adjust their grip based on what they're perceiving, but robots are not there yet. But what's been, and I've been studying this problem for 35 years, I like to say we made very little progress. But until recently, and the breakthrough to the second part of your question is, as I see it is in deep learning. That the set of technologies that basically were harnessed for a computer vision initially, involving massive data sets, massive computation, and new set of algorithms known as deep learning, that we're able to train these deep neural networks to be able to learn functions that were very difficult from examples.

    In a nutshell, the deep learning revolution, the AI revolution that everyone's talking about is based on learning from example. And the challenge is, how do you get enough examples? Because in this case, you have many different products. So you have to be able to train the system and can train it for this one specific configuration you're working with. That is the sensors that Jim alluded to, and the robot specific of what kind of suction cups it has, and the class of objects, the range of products that you're going to be grasping and manipulating. One of the ideas that we pioneered was the idea of Sim2Real. So we do that training in simulation. We can talk more about that during the Q&A, but that was a breakthrough that allowed us to train millions of examples very rapidly.

    So overnight, we can change something like the gripper or the camera location, or the mix of packages and retrain our deep neural network system overnight to accommodate that environment. That's what we mean by day one, where as Jim said, our system comes in and because of that rapid configurability, is able to start working right away. Now, I want to emphasize that it's not perfect. This is a very important thing to keep in mind, which is that warehouse work is fault tolerant. And this is a big difference from, let's say self-driving cars. Self-driving car is not fault tolerant. If you have one error, that could mean driving off a cliff and doing incredible harm. In west, people drop things all the time, it's just a fact of life. Essentially, it's factored in.

    Robots will drop things occasionally, and that's acceptable and inevitable. It's one thing reason why we see this as a sweet spot for robotics, because this is one where there is some fault tolerance, you want to increase the success rate and the throughput of these systems. And we can talk more about that too, but the key is that we now have the ability to increase the reliability to a stage where it's on par or in some cases, even above that of human workers. I'll stop here and unshare my screen so that we can open up to other questions.


    Jeremy Capron:

    Yeah. Thank you very much, you both. I think it's very exciting what you've been able to achieve with Ambi in such a short period of time. I want to emphasize that at ROBO, we're really strong believers that we've hit an inflection point in terms of machine intelligence and what that means in terms of the expansion in the range of capabilities for robotics and autonomous systems in general, and their potential to be applied across essentially all sectors of the economy. Certainly, logistics being one of the more exciting ones in recent years. In fact, if you look at the ROBO index portfolio and its performance since inception about eight years ago, logistics and warehouse automation has been the best performing sub-sector in terms of total returns. Now, we've seen also great returns around the machine intelligence side, what we call the compute and AI. I think what you said about how deep learning has really enabled this progress in terms of robotic picking. That's something I want to go back to, and I'd like you to talk to us about the genetics of Ambi.

    I recall a few years ago, you were telling us about how you had presented the Dex-Net and the virtualization and assimilation approach to one of the biggest e-commerce companies out there. Is this what really was the impetus to start this new company? Tell us about how it all started.


    Ken Goldberg:

    Okay, great. I really love that question because it's very fun for me to answer. It really comes down to one very, very talented individual. His name is Jeff Mahler. Jeff came to my lab as a PhD student with extremely refined engineering skills. He had also been involved in a startup before he arrived, and we started looking at the grasping problem. At the same time, look at the grasping problem using deep learning, I should say. There are so many different directions you can go. Many different groups around the world are exploring this in different ways, but Jeff and I were co-teaching a class at Berkeley on robot manipulation or using all the, what we call the classical methods of analytic models. So we're looking at the physics of contact and forces and torques. So you could determine if an object would be securely held by particular gripper.

    Coming out of that class, we were excited to say, how can we actually merge that idea, those ideas with deep learning? That was the impetus behind, we started this project called Dex-Net, as you mentioned. Dex-Net is the analog to image-net for computer vision. And the key to image-net, as many people know, is that it was trained over on millions of labeled images. Then it was able to start generalizing to new images and working remarkably well. So our idea was to do something very analogous, but we needed millions of, in this case, three-dimensional objects with reliable grasps on each of those objects. That's where we put to practice the analytic methods that we had been just teaching.

    We put that in to give many examples, and then we were able to train the network to basically learn what psychologists call affordance, and learned how to just look at a set of points in space and see, where could you put your gripper to pick something up? That was, in my view, a breakthrough. It was really exciting, it worked much better than I thought it was going to. We started expanding that research, bringing in others. Matt Matl joined the lab, Steve McKinley and Dave Gealy are two of the best mechanical engineers I've ever met. The four of them started working together. We were then invited to an event held by, I think it's fair to say that the top e-commerce company in the world, Amazon, and we got a chance to present the system to directly to Jeff Bezos.

    It was a really inspiring moment because we had traveled down, we set up our system and we brought hundreds of objects for him or for anyone to try it out. He came in and tried it for a little while and he was very curious. He picked out a banana and he put it down front of the system and it just picked up the banana just like that. Then his assistant took off his shoe, Tye Brady from Amazon, he took off his shoe and put it in front of the robot. We all held our breath because it had never been trained on a shoe. We had no idea how it was going to react. Amazingly, it just reached over without hesitation and picked up the shoe and put it into the bin. It was an incredible aha moment right there with Jeff Bezos watching and his top robotics colleague.

    He was thrilled. He said, "This is a central problem we're trying to solve." That evening, as we packed up and we're driving home, Jeff basically made the decision that this was the time to start a company. That's how it started. I will say huge amount has happened since then, because you can take the system out from a lab, but now you have to make it practical and commercially viable. And Jim can talk more about this. One of the things that was really exciting is the team shifted modes into the amazing entrepreneurial group that really started to understand all of this complexities of working in the real world, which are very, it's an additional set, let's say, to the research aspects and making system that's viable, highly reliable and cost-effective. That's what we were working on. I really have to say this team is very lean and always under-promises and over-delivers.

    The technology is rock solid. They're the best engineers I've had them in my lab in 25 years. They're really the most hardworking, straightforward guys you can imagine. When we wanted a new CEO, we had a dream spec of the person who had so much experience in this industry, who really knew it from the warehouse floor side. So we talked to a lot of people, but we were very, very specific. We wanted a real world-class expert. When Jim came to meet us, we were absolutely thrilled because he hit every metric that we were thinking about. He had also had experience with Kindred, which was a similar company, but he had brought this... When he saw our group, he said, "This is a technology that can take this to another level." That's where that combination is very exciting. Jim, maybe you can say something about your perspective in terms of the market and how you see Ambi fitting into this logistics boom that we're just on the cusp of?


    Jim Liefer:

    I appreciate all of that, Ken. Thank you. I want to give one more moment of accolades to the team. You're absolutely right about these guys, Jeff, Steve, and David and Matt. That combination of all of them together, I've just really never seen anything like it. Day after day, it allows us to get a lot of things done. It's the team, it's the technology, it's the market that we operate in those markets. It's also just from a business standpoint, it's the unit economics. What I could tell you from living in the world of UPS and Walmart, where I came from, and dealing with all of the issues where you're creating new processes to handle more volume, because it's all about throughput. Every day, it's about throughput that you can get through a building no matter what building it is.

    Getting processes in place, getting different forms of automation in place, whether that's mechanical or autonomous systems or whatever it is. I would tell you with so much confidence that what we're doing here at Ambi solves those problems in those facilities. I want to take a moment to say, from my years of experience on the other side, it would be impractical to think that you can set aside everything that you've built over the years, because that's what all of our customer base has been doing. They've been putting these different solutions in, and they've been connecting them. Then they have a singular solution that solves part of the throughput, and then it creates another opportunity in the facility in terms of the throughput. The way that the Ambi technology works is that we can fit in between all of those different pieces of automation that are already in place. We learn from that, and then we can move upstream and downstream and we can help other parts of the process.

    I've seen some of the questions coming in. I want to make sure that I talk about our approach. When I said earlier about a human centric approach, what I mean by that is there are tremendous number of mundane repetitive tasks out there in the world. Many of them live inside of fulfillment centers and inside of parcel sortation facilities and those kinds of things. It is those kinds of tasks that are, those are not well suited for people, for all of us humans. The way that we're helping our customer is to make sure that when they go out and take the time to find those good people and bring them in and train them and take care of them, they give them jobs to do in those buildings that are more appropriate for humans. There's more ambiguity to solve those problems. There's more creativity.

    When they are working in that area where you're having to do a mundane task, you're having a robot do a lot of it. So when I talk about human-centric, I mean we design these systems with the people in mind that are working alongside of them, because we believe that this isn't about we don't need humans anymore. Of course we do. We're all that, we're all human. What we need to do is have the ability to uplift those people and do jobs that are meaningful because we are all the same. We all want the same thing. I'll stop there.


    Jeremy Capron:

    So Jim, following up on that, you are providing automation solutions to solve those businesses problems. When you come up with solutions like what Ambi offers or what Kindred offers, tell us a little more about the return on investment that your customers expect. What can be expected? Are we talking about a very short payback periods and if you could create examples how the math works there?


    Jim Liefer:

    Yeah, absolutely. So the way that our business model is set up at Ambi Robotics, and it was very similar when I was at Kindred, we provide robots as a service. This isn't about selling the hardware and just licensing the software and walking away. By the way, in my former life, there was a lot of that. There was a lot of buying some new sort or a piece of condense or something and you'd pay the money for it, and you'd bring it in, it wouldn't perform like you expected it to. But we have this robot to the service model and we feel strongly about this that we stay there, side-by-side with the customer. The other thing that the robots as a service model does is it takes down the barriers to entry for these customers. Because a lot of them are challenged with deploying CapEx to pay for these big systems and where they have the ability on the operating expense side, on the OPEC side, they can pay a monthly or a piece picking fee and it makes it much easier to bring these autonomous solutions into play.

    Then as I said, when they take all the time to bring those people into their environments, they can give people that come in, they can give them those better jobs that are in the facility. We're very much about robots as a service. I think it's the absolute right thing to do especially with the customer, knowing that we will be there with them. We take on the holiday peak mentality that our customers do because that's where everything matters. I'll end with what I see in the places where we are most within in terms of the retail space and the parcel sortation space. The unit economics work incredibly well for both the customer and for Ambi. So we see it as a really great business model.


    Ken Goldberg:

    If I could just add to that, the human centric idea is so important to us, Jeremy. I want to emphasize that because I run a program at Berkeley called People and Robots. It really emphasizes this idea that robots are not going to replace the workers. That is a myth. There is a need. Humans are incredibly capable. So there's many, many jobs for humans. The challenge is that it's actually a shortage of humans. It cannot find enough workers to take on some of these jobs. They're dangerous, repetitive, there's injuries. So turnover is very high. What we want to do is improve these jobs for the human workers. The ones that are there, in fact, in our experience with real customers, the workers actually really love the system because it's taking over a part of their job that they never wanted to do.

    The analogy I want to make for that is, think about spreadsheets. When spreadsheets came out, there was a sense, well, this is the end of accountants and bookkeepers. Well, we just won't have them anymore. But no, it didn't work that way. What actually it did, was it freed up accountants and bookkeepers to not have to do all this sitting around with pencil and paper filling in cells, but they could automate this and then they could start enhancing. They could do their job better because they could do what if visualizations very quickly. So this is the way we're thinking about this. We're very conscious and very respectful of every individual worker in a warehouse and what their job is like, and of the customers. We are very customer-centric, which is how does warehouse actually use this? They don't care whether it's AI or something fancy. They want something that's going to work and work reliably and cost effectively.


    Jeremy Capron:

    Right. Okay. I see we have an interesting question. It's something we've done a lot of work on here at ROBO with the research team, but I'd love to get your view on that, you both. What is the percentage of warehouses out there in the country that are automated? Of course, it's a difficult question. What does automated mean, a fully automated, a little bit of automated conveyor systems or whatnot? But before I let you go on to this one, I want to say that we've published some of our research. It's available on the roboglobal.com website, where we make some of our research on companies and automation technologies available to investors. We believe the logistics warehouse automation market is about $50 billion in annual revenue. And it's been growing in the mid teams annual rate. I'm curious to hear what you think, Jim, in terms of the addressable market for your tech, but also that question from the audience around how automated are warehouses currently in the country?


    Jim Liefer:

    Yeah, absolutely. As you said, it depends on the definition of automation. There's been a lot of it around for a long time, like back when we started having powered conveyance in facilities instead of having warehouse workers use hand trucks to move case packs around and that sort of thing. Well, from all of the buildings I've been in, both retail operations and parcel buildings, and there's been many of them, there's a high degree of automation that's mechanical. There is a small percentage, a very small percentage right now in terms of the piece picking and placement, which is the hard part, everything that we're doing, and Ken was talking about that earlier. However, there's other kinds of autonomous systems in there. So going back to the time when Amazon acquired Kiva Systems, which was maybe 2012.

    That problem, they were solving for travel. People having to travel around the facility. So they had these autonomous systems that would move the product around for them. That was an important problem to solve, but an easier problem to solve, frankly. And then there were other systems that were in place, like Six River Systems that would improve the efficiency of the worker by helping them through the pick path through the facility and those kinds of things. The general answer there is, there's a lot of autonomous systems in there. I'm sorry, a lot of automated systems in there, but then there's those points in between, which is where the Ambi Robotics piece picking and placement fits in to solve for that. There was another question that was embedded in there. Remind me of that.


    Jeremy Capron:

    How do you think about the addressable market for retail operation?


    Jim Liefer:

    Yeah. Oh, thank you. So from a retail operation perspective, I don't want to say endless, but it's enormous. There's an enormous opportunity there. And the reason that I say that is, not only is it the size of the big retailers and the middle tier and the smaller ones, but also when you think about the functions that are happening within those fulfillment and sort facilities, which are by the way, if you've not been in one, for anyone who's not been in one, it's like being inside of a city like Manhattan. There's just so much happening around you all the time. So there's different processes, different places where you can inject this autonomous capability.

    If you look at just, I can give you a number here, if you look at parcel alone, for parcel sortation, I see that market as about a two billion dollar market opportunity where they're just sorting packages. And I'm not saying that Ambi Robotics would replace all of that autonomous capability, but there's many points where it doesn't make sense to sort packages down to a specific level that's so detailed. Like a zip code, for example, where in those locations you might use a human today to do that. Again, very mundane repetitive tasks, it's better to use an autonomous system for that. Ken, anything to add there?


    Ken Goldberg:

    No. I'm not the expert on the market size aspects of these things, but I do feel that there's been this shift in the demand over the past 18 months. It's due to the changes of COVID. As we come out of it, there's a huge surge of interest. The other thing, as Jim alluded to, is that this is very seasonal. There's a strong season at seasonality. And it's that from October to through December is a peak period, and you get this huge surge in demand. Everybody is sending things around due to the holidays, due to ordering things and packing them up, and then shipping them off to friends and family. What happens is that surge, it makes it very difficult on the retailers. They can't find people to do this. So this is where there's a real sense of ramping up now starting really year ahead for the next season. We're very excited to address that.


    Jeremy Capron:

    I see we have a question about public companies that play around supply chain, warehouse automation. I want to take this one to reiterate that within the ROBO portfolio is about 10% of the companies [in there that are in our automation sub-sector. And certainly the public equities reflect what you just said, that there's been a step up. Although books are full, companies can't keep up with demand in terms of providing the automation equipment. So to give a few examples, there's quite a bit of variety. There's around 10 companies in the portfolio, and they are as diverse as a Zebra Technologies that provides the track and trace technology, marketing and RFID tagging and the likes. We have companies like Manhattan Associates that provides the software that powers the warehouse automation. Or Daifuku over in Japan that's a giant in material holding or material handling equipment. There's good diversity. We will have a company out of Europe, it's called Cargo Tag, that provides full automation equipment to move container boxes and things like that.

    Anyways, I wanted to ask you both because we're getting quickly on to the hour here. You've made a very strong case, which is this automation is very promising. When you think about what's happened in the world of robots and automation beyond logistics, what do you think are important areas to look at as an investor? As you know, our audience is mostly investors today, so I'm curious to hear your perspective on that?


    Ken Goldberg:

    Well, I can tell you my perspective on that one, Jeremy. I think that we've talked a lot about logistics today. That to me is, as I said, the sweet spot. But there are two others that I'm very excited about. One is, and this came up in the questions, is agriculture. There is an analogous problem there, which is getting workers. Again, there's these often back-breaking aspects of the job that the jobs could be improved. We'll always need human workers, but they won't have to do these very low level things like picking strawberries in the future. Now, the challenge is very, very steep there because strawberries are in some way, even harder than picking packages. They're hanging and they're covered up by vegetation. You have to find the strawberry, not pick it too early, not pick it too late.

    There's a variety of huge challenges to agriculture automation, but there's a lot of new work in that area. Again, it's relying on this critical part of being able to find and grasp objects securely. That has been the missing link for automation in my view. That's what we're starting to see closing that gap. And that's very, very exciting. So I think there's going to be a massive increase in automation for ag. The other is healthcare, and specifically for home automation. I think that we're seeing tele-health as mentioned before, but we would see a huge increase in the shift in the demographics, huge increase in the aging population. I can relate to that because I'm a boomer myself, as my daughters remind me.

    One of the things is that there's no speculation here. It's just the demographics are, we know exactly how many people are born, when they were born, and you can actually watch the curve and this huge swath of population, not only in the US, but all around the world, is now moving into their senior years. People don't want to go into retirement homes, they want to stay in their own homes. And this then it becomes very difficult to do things around the house. So we're not going to have a robot like Rosie from the Jetsons that's going to do everything. That's not coming anytime soon. But one challenge is just keeping the floors clean and clear. Now, we have Roombas, obviously, that sweep the floor. That's one thing, but tripping hazards are very big issue for senior citizens.

    The next step is have a robot that can move around the house and pick things up and put them away. Just keep the floors clear so that there's no tripping. This tidying up is another sweet spot, I believe, that we will start to see those robots in the next five to 10 years, because the demand is there, the technology is very close, we're making enough progress. The key, as we've been talking about for the whole hour, is in getting the economics to work, getting the price of the robots down and making it valuable and cost effective for a person who's retiring, or in their senior years to want a robot that will actually help them do something very helpful around their homes and keep them safe. I think that's going to be a big market in the future.


    Jeremy Capron:

    Great. Jim, any views here?


    Jim Liefer:

    Yeah. Instead of talking about the various markets, just from an investor standpoint, I'll put it as if I was an investor, I would look at three things. I would look at focus, that is that company that I wanted to invest in. Does that team have the focus of caring about solving real world problems versus creating interesting technology? It doesn't matter if you create an interesting technology if you can't apply it to real world problems, that's number one. And then the second one is the team itself. What are the dynamics of that team, what is the culture, all of those things around the team, which are super important? And then finally, the underlying technology. I think that when you think about the technology being adaptable, because there are so many different issues to solve in the world, you have to have that ability to adapt quickly. That's how I would think of it.


    Jeremy Capron:

    Excellent. All right. Well, I think it's time to wrap up here. And before we do that, I want to remind everyone that if you want to learn more about investing robotics, automation, AI, please visit our website, roboglobal.com, where we share some of our research and companies in the ROBO to THNQ, and the HTEC portfolios. I want to thank you both very much, Jim and Ken, for sharing with us today. It's been very insightful, and we wish you good luck to you both and Ambi Robotics. Thank you all for joining our call today, and we'll look forward to speaking to you again soon.


    Ken Goldberg:

    Thank you, Jeremy. Really a pleasure.

    Back to All Entries