Podcast

The Future Of Last-mile Delivery Robots

Touraj Parang, President and COO of Serve Robotics, joins Jeff Dance to explore how sidewalk robots are reshaping last-mile delivery. He explains how Serve spun out of Postmates and Uber, scaled to 2,000 robots across major US cities, and handles chaotic sidewalks with AI autonomy backed by teleoperation. The conversation covers merchant handoffs, falling hardware costs, smart city infrastructure, and multimodal delivery with drones. Touraj also shares how pushing delivery costs toward zero can create demand, revive local commerce, and unlock new uses such as reverse logistics.

Jeff Dance (Host), Touraj Parang (President and COO)

Podcast Transcript:

Jeff Dance: In this episode of The Future Of, we’re joined by Touraj Parang, President and COO at Serve Robotics, to explore the future of last-mile delivery robots. Touraj, welcome to the show.

Touraj Parang: Thank you, Jeff, for having me.

Jeff Dance: Excited to go deeper. By way of introduction, Touraj has been the President and COO of Serve Robotics since July 2022. Prior to joining Serve, he held a number of leadership positions at tech companies. He worked at GoDaddy as the VP of Corporate Development. He was the founding Chief Operating Officer at UpCounsel, which was later acquired; the VP of Strategy and Corporate Development at Webs, which was later acquired; and the co-founder and Chief Operating Officer at Jaxtr, some tech I remember using back in the day, which was also acquired. That’s a big string of successful runs, which is pretty amazing. He also has a Bachelor of Arts degree in philosophy and economics from Stanford University and a Juris Doctor from Yale Law School. It’s impressive. Is there anything else about your journey to landing in robotics that you can share with us?

Touraj Parang: Thanks, Jeff. As you just mentioned, I’ve had a pretty nonlinear path to getting where I am today. I started as a lawyer, pivoted into venture capital, and then moved into operational and founding roles as an entrepreneur. It’s been a wild ride. But there has been one through line in everything I’ve done, which you can kind of see in hindsight. It started with my parents being small business owners. In the early 2000s, and this is not part of my official resume, I bought them a Subway franchise. I actually went to Subway University, so I have that degree as well. I’m a sandwich artist, and I experienced for myself the pain that small businesses go through and how hard it is to be a small business owner in the US.

A lot of the companies you mentioned have all had some angle of supporting and helping small businesses in the digital world. At Webs, we had millions of small business websites that we were hosting, and we were helping them get on social and mobile, so their presence would basically be the virtual front door for their business. Then at GoDaddy, where I was for seven years, we did that at a much bigger scale. In fact, the mission of GoDaddy was to tilt the global economy toward small business. I feel like what those platforms were doing was helping small businesses reach customers digitally much more effectively and reduce the cost of reaching those customers. What we are now doing with robotics at Serve is helping small businesses reach their customers more cost-effectively in the physical world.

Jeff Dance: Amazing. Thanks for that background. That’s meaningful. I didn’t mention the degree from Subway University. That’s significant.

Touraj Parang: It is. I’m proud of it.

Jeff Dance: I love it. You pair that with Stanford and Yale, and it’s cool. As we think about the future of last-mile delivery, for those who aren’t familiar with the definition of last mile, my understanding is that it technically means the last step of the journey, which is often the most expensive. It’s sometimes the one that makes the most impact on customers, because they’re receiving something in that last mile. Is that how you define it? It’s the last step, and that’s a big place where Serve plays.

Touraj Parang: Absolutely. In the industry, “last mile” is sometimes used generically to mean the last few miles, but we literally mean the last mile. We started doing deliveries for the last mile between a restaurant and a customer, from a restaurant’s doorstep to a customer’s doorstep, with these Serve robots. So typically a mile, a little bit less or a little bit more, but usually around that distance. A generic industry term is actually a literal term when it comes to what we do with our sidewalk robots. Some people even call it the last yard or the last feet, but last mile is a perfect way to describe it.

Jeff Dance: It’s the last step. For those who aren’t familiar with the term, it can be 10 blocks, but it could be several miles. It could be 10 miles. It’s just that last major step.

Touraj Parang: Exactly.

Jeff Dance: I believe I’ve seen Serve robots before. We had Microsoft as a client, and when we were down at Berkeley, I thought I saw a bunch of robots. I thought they were probably yours. You’ve been around for a while. Tell us more about Serve. I thought I read that you have the largest fleet of sidewalk robots, but give us a little more background on Serve.

Touraj Parang: Yeah, absolutely. I’m not sure the one you saw in Berkeley was ours, because to my recollection we’ve never really operated in that area, but it could have been an engineer testing something. Serve was actually started about nine years ago inside a food delivery company called Postmates. Postmates was subsequently acquired by Uber, and after that acquisition, about six years ago, the management team of Serve went to Uber’s management team and said, “Look, it makes a lot more sense for this technology to be independent and to be spun out as a venture-backed company. We can hire the right talent, and we can also use these robots in many other ways than just food delivery. When you do that, you have the opportunity to bring the cost of delivery down for everybody, including food delivery companies.” The industrial logic made sense to Uber, so Uber agreed, and we spun Serve out about five and a half years ago to be independent. I joined Serve right at that time and helped with that spin-out process.

We ran for a couple of years as a private company, venture-backed as well as backed by key strategic investors, including Uber and NVIDIA. 7-Eleven was also an investor, and Delivery Hero was an investor. Then we took the company public on NASDAQ two years ago so that we could have access to a larger pool of capital and build those 2,000 robots that you mentioned. With robotics companies, at least with our business model, robotics as a service, your revenue doesn’t come until you build a robot. So we built the robots, and now we are in the revenue generation and utilization phase of the journey.

Jeff Dance: Great. What about some stats? Did I hear 2,000 robots? I thought I read that online. Tell us more about the number of robots and the number of missions or deliveries that you’re doing.

Touraj Parang: For the longest time, we actually had just about 100 robots, up until we went public. Then it took a while for us to raise the money in the public market and build the 2,000 robots. By the end of last year, we had deployed 2,000 robots, and every day those robots travel, on average, the distance from New York to LA and back. That’s how much they travel. We’re in over 20 geographic areas, or municipalities. The major metropolitan areas we’re in are LA, which was the first place we started, then Miami, Chicago, DC, and we just started in San Jose. We have some operations in Dallas and Atlanta. Those are the large urban areas where we’re best suited to operate. We’ve also augmented our fleet with an indoor fleet through the acquisition of a hospital robot company called Diligent. The robot’s name is Moxi. Together with Moxi, and I think it’s in our quarterly reports, we’ve done millions of deliveries since we’ve had these robots, and we’re just getting started.

So, lots of deliveries already accomplished and lots more to be done. In the US, food delivery alone is millions of deliveries that people do today, and they’re all done more or less by humans riding in a car. That’s an absurd situation that we wanted to address when we started Serve. Our tagline, which kind of stuck, is: Why move a two-pound burrito in a two-ton car? There is a mismatch between the machinery and the labor and all the effort it takes just to move something small like a burrito from one part of town to the other.

Jeff Dance: That’s awesome. Lots of opportunity to scale, and it sounds like that’s what you’re doing. You’re on the scaling journey.

Touraj Parang: We are on the scaling journey, and it has multiple dimensions. There’s raw hardware: you have to build the thing. Then there’s the software, which is the autonomy. As your scale grows, you encounter more corner cases. Something that happens 1% of the time when you have 100 robots, one robot may encounter in a given period of time. If it’s thousands of robots, you encounter it more frequently. We are very cautious and conscious about safety. Safety is our number one priority as a company, so as you scale, you have to be very mindful about those corner and edge cases. Software and autonomy play a big role in that. You need to use all the machine learning and technology you can to anticipate what those issues will be, or, when you encounter them, to have a graceful way of resolving them so they don’t lead to a safety incident. We’ve never had a major safety incident in the history of the company.

Jeff Dance: Amazing. You’re out there in the real world. You’re on sidewalks and crossing streets, and there are people and kids, so it’s great that safety is number one. Tell us more about some of the challenges you’ve had to overcome navigating the real world. We all wanted autonomous cars. I’m an 80% autonomous driving person, and I get frustrated when my car’s not driving autonomously because there’s a little fog or rain and it needs to heat up and clear out. It has been challenging for cars. I imagine you have a lot of challenges too. Tell us more about some of those.

Touraj Parang: In some ways, I would say cars have it easy. When you’re trying to navigate on a sidewalk, I would say it’s a negotiation, whereas on the streets you kind of just have to stick to your lane. You don’t normally have a car suddenly reversing direction, or stopping on a highway in front of you and slamming on its brakes. You don’t have a car chasing a squirrel across three lanes. On a sidewalk, you have a very chaotic, complex environment. It’s really hard to predict the movements of all the actors on the sidewalk. You have kids and toddlers. You have animals, like dogs pulling on a leash. Sometimes you have people who are curious about robots and suddenly want to approach them or jump in front of them to see if they stop.

Jeff Dance: To see what happens. I bet you get that a lot.

Touraj Parang: You also have scooters and mobility devices. There’s a lot that happens, and that’s what makes it both exhilarating and very challenging to solve. We are dealing with that on a daily basis. The good news is that every day we get better. The more robots you have out there, even our indoor robots, every encounter informs the entire network and the entire system, so all the robots become better as a result.

Jeff Dance: Last night, at the Northwest Robotics Alliance, which I started, we were on a panel about the future of fleet management. One of the topics of conversation was that it’s not really about the robot. It’s about the ecosystem and the operations you participate in. Have you figured out the ecosystem, the market, and how this thing is going to integrate for scaling? Lots of pilots have been done over the last 15 years, as new robot tech came into play beyond the historical arms that are 40 or 50 years old. That seems to be one of the big challenges, so congrats on making so much headway. If you can figure out the crazy sidewalks and streets and not have accidents after 100,000 deliveries, that’s really amazing.

Touraj Parang: Thank you very much. We take great pride in that. As you mentioned, when you have a couple of robots in a lab environment, your focus really is on the technology and making sure it works and performs the intended task. But once you have hundreds or thousands of robots, it really becomes a question of orchestration, fleet management, and operational excellence. This is where it really helped to have been born inside a food delivery and logistics company, because we knew the metrics that mattered, and we always had a keen eye toward optimizing for them. We made sure the merchants’ flow and experience were not compromised, and even provided a better experience than they have today with human couriers. Similarly, for end customers, we want to make sure they get their order intact, with great quality, on time, and hopefully with the bigger peace of mind that it was safely locked in a robot and no one tampered with it.

Jeff Dance: Tell us more about the interaction with customers. If it’s a good restaurant, it’s a fairly chaotic environment, and you have experience with that going back to your younger years. How does that work, and how have you solved the back-and-forth?

Touraj Parang: When we first started, we were worried. This is the only point in the interaction where somebody has to do something different from what they do today. The merchant actually has to come to the sidewalk or the front door of the restaurant and load the robot. We worried that might be a bridge too far for them. What we found is that the overwhelming majority of merchants actually welcome that experience, because it puts them back in charge of when the order is ready to go, and they hand it off. The paradigm with human couriers is very different. Couriers come in, and sometimes they take the wrong order or don’t take the full order. They may also be impatient if the order isn’t ready, because they’re on the clock and want to make sure they get the next order. So it’s been pretty delightful to see the reception we’ve had from the merchant community. Of course, in any neighborhood or geographic area there may be some merchants this wouldn’t work for, but we have more than 4,000 merchants and restaurants that we work with right now, and I’m glad to say it has been a delightful experience for them.

Jeff Dance: That’s interesting. I could see how it’s not different from serving a customer. If they have tables, they have to bring the food out. So it’s, “Okay, this is ready. Let me go deliver it to the robot.”

Touraj Parang: To the robot. And the robot has the name of the customer on it, so they know exactly which robot to load. Each of our robots has a name and a personality too. It’s a very seamless experience. The challenging part with robots is for the merchants to know that there’s a robot out there waiting for the order. That’s something we’ve worked really hard on with Uber, DoorDash, and the other delivery partners we work with, to make sure merchant notification is done properly. Sometimes it’s easy to integrate with the platform they work with, and sometimes it’s not. To that end, we’re actually creating our own hardware device for merchants for whom it would be hard for their tablet or POS system to show that a robot is outside. We can put a little device on their counter called Beacon. We announced that in our quarterly earnings last quarter, so we’re very excited about that too. We’re constantly trying to think of ways to make life easier for merchants and small businesses and to make these robots an integrated experience for them.

Jeff Dance: Nice. Tell me more about how you navigate, from a robot perspective, getting up sidewalks and things like that. There’s been a lot of movement in the last two years. I’m assuming you’ve seen the Spot robots that are on wheels and how they can navigate difficult terrain. That’s been a huge advancement. How are you thinking about it? Are you planning to get into stairs? Even that sidewalk hurdle can be an issue for some of these robots. Are there any existing capabilities or plans you can share that aren’t confidential?

Touraj Parang: There are actually a number of sidewalk robotics companies out there today. One of them, which has a very large fleet, Starship, started before we did. This may also be similar to the one you saw on the Berkeley campus. Their robots are usually smaller, low to the ground, with smaller wheels, usually six wheels or so. They were built more for a campus environment. Companies did that because there were a lot of student food orders, and a campus is a very contained environment, so it’s easier than the chaotic sidewalks in urban areas. A lot of robotics companies went there.

Because we were born inside Postmates, we had a different problem to solve from the beginning, which was the urban environment. Our robots were purpose-built from the beginning to navigate sidewalks, and sidewalks can be very uneven. Our initial robot designs were inspired by the Mars rover. They had drivetrains that could keep all four points of contact even if one point went up, so the robot wouldn’t wobble or get stuck. They have larger wheels, so they can climb a six-inch curb. Before we go into any new area, we map it. We determine the right intersections for the robot to cross and all those things. Today we don’t do stairs. Of course, that would be an optimization we’d need to tackle down the road to get even closer to both merchants and customers. The good news is that there’s so much room for improvement. It’s going to keep us busy for a long time. The existing form factor is great for the vast majority of orders, but there’s always room to improve and even expand the market size.

The market is also not just food delivery. You can imagine these robots being used to deliver anything, especially once you bring the cost of delivery down. It opens up so many more opportunities: convenience items, pharmacy items, packages, even reverse logistics if you want to return something. None of these things are really happening today, because the cost of delivery is so prohibitively high. If you’re ordering something that costs five or ten dollars and you have to pay another ten dollars for a delivery fee, it stops you in your tracks, no matter how rich you are.

Jeff Dance: It doesn’t make sense economically. Plus, there’s still a human in those vehicles, which is even more expensive.

Touraj Parang: That’s what contributes to the ten-dollar price people pay. And merchants right now actually pay something like 20% to 30% of their income to these delivery platforms to cover the costs of delivery. So there’s an expense for the merchants too.

Jeff Dance: Right. For those who are skeptical about robots being in their neighborhood, there are a lot of community benefits we just talked about. You’re reducing cars and car travel. What would you say to these communities about some of the benefits?

Touraj Parang: If you look at it, we’ve spent over a century designing our cities around passenger vehicles. We have big roads and lots of parking spaces and parking lots. All that space could someday be reclaimed by pedestrians and the people who actually live in those neighborhoods if we offload a lot of these delivery activities to vehicles that are purpose-suited for those items. If we can do even just food deliveries alone, imagine how many fewer cars you would need in a city. The benefits of that are that cities become safer, because cars are responsible for a lot of pedestrian deaths. Cars also create a lot of pollution, noise, and congestion. We should reclaim our cities from cars, in a way. Cities should still use cars to move people, but not tacos.

Jeff Dance: That’s interesting. Thank you. Let’s talk a little more about the future. You’re an interesting person, given that you’ve scaled companies and you understand economics, strategy, corporate development, and operations. You have the background to think about how you scale last-mile delivery to become infrastructure. How do you make this infrastructure, so you can have some of those community benefits you mentioned? Let’s reduce a lane and make that for people. Let’s add more space for parks, et cetera. What do you think it’s going to take to make it more like infrastructure, something that is commonly accepted in our cities?

Touraj Parang: I think we actually have the beginnings of that. One of the things I notice in areas where our robots have been for a while is that they’re sort of part of the urban infrastructure. Locals don’t even really notice them anymore. They just accept them like they would accept a stroller or a shopping cart on the sidewalk.

But that’s the human interaction component of it. What’s even more interesting is the interaction between robots and city infrastructure. Right now, for instance, you have a lot of intersections with push-to-cross buttons. It’s challenging enough for some people to even push them. Our robots, and all other sidewalk robots, don’t have an arm or a finger. And you don’t really need to build an arm and a finger to do that. If the city can control it, or if there’s smart infrastructure with some communication between the robot and the traffic light, then these things work seamlessly with each other. The same goes for gates, some doors, and zone barriers. I think the city of the future, the smart city, will be one that recognizes these robots and is able to interact with, adapt to, and communicate with them.

I don’t think it’s that far off. You saw it with broadband. When the first iPhone came out, the 3G network was extremely burdened, and all the cell towers and infrastructure had to be upgraded so that 4G and 5G phones could work. I feel like there’s going to be an upgrade to our city infrastructure to accommodate all these robots that can do great things for us.

Jeff Dance: So a bit more infrastructure and a bit more integration will put you in a position to scale last-mile delivery.

Touraj Parang: It would definitely accelerate it and open up a lot of opportunities. Right now, for instance, if there’s a push-to-cross button and a robot needs to get to the other side of the road, sometimes we have to navigate around it, and it slows down the delivery. It’s an unnecessary obstacle.

Jeff Dance: Right. How do you deal with charging at scale? Do you have a central hub where they come back when they need to charge? I’m assuming they’re all electric.

Touraj Parang: Yes. Right now we have a hub model, and in certain urban areas like Los Angeles we have many hubs, each with fifty to hundreds of robots. What we did was design a robot that doesn’t need charging midday. Our robots, especially the latest generation, can operate for 10 to 12 hours on a single charge, which usually takes them through the lunch rush and the dinner rush. We usually deploy them in the morning, they do their thing, and then they come back at night to be charged, cleaned, and disinfected, with any maintenance issues taken care of, ready to go the next morning.

Jeff Dance: Nice. I recently had a podcast on the future of batteries, and it’s really exciting where some of the battery tech is going, both in how fast things charge and in longevity. Obviously, we’re seeing it in our phones. They’re putting more and more of this material in our phones, so they’re lasting a little longer and charging a little faster. But for vehicles and things like this, it’s pretty exciting too.

Touraj Parang: It’s interesting. Where you see the cost of labor going up, we have the opposite trend with the cost of hardware, and the capability of the hardware is really increasing. We started building the current version of the robot in 2019. We actually deployed them, and they were doing commercial deliveries for Postmates at that time. It was a very expensive robot to build. It has the full autonomy stack. It has a lidar on top. It has cameras with a 360-degree view. It has all sorts of bells and whistles you would see in a Tesla, but in a shopping-cart-sized form factor. It also has a very strong NVIDIA GPU processor inside it. It was very expensive, so a lot of people were wondering whether the economics could ever work. But all of those components came down dramatically. That was a bet we made on the future. We looked at where the puck was going and said, “Okay, in a couple of years, this is going to follow Moore’s law. It’s going to become cheaper and cheaper and more powerful.” And lo and behold, that did happen. Just as an example, the cost of lidar technology has come down by 90% in less than ten years. What was totally inaccessible and super expensive for many startups and robotics companies is now commoditized in some ways.

Jeff Dance: We’ve been working on industrial autonomous vehicles for 10 years, and we remember when those pucks cost about 30 grand each. We were doing a lot with industrial vehicles originally, and I thought the brain we’re going to build on top of this is like 250 grand. That doesn’t scale for a company like yours. But when you can build something for much less, you can translate that into cost savings for everyone else.

Touraj Parang: Absolutely. Layered on top of that, you have more and more broadband connectivity in urban areas: better, more reliable broadband, which is critical. We haven’t even talked about teleoperations yet, but we should at some point, because that’s where broadband connectivity really matters. Then you also have all the latest developments in AI. If you think about it, robots are embodied AI. They are the physical manifestation of what AI can do in the real world. So you have this convergence, a perfect storm of all these trends, that makes these robots quite timely.

Jeff Dance: Timely. Tell me more about how AI is shaping what you can do. Obviously, exceptions are a big thing. You mentioned teleoperation. There are people who need to handle those, but now that we have agents, tell us more about how you’re using AI or planning to use AI, and how that can help last-mile delivery.

Touraj Parang: Definitely. Our autonomy stack is all AI. AI learns from data, from real-world encounters, and from simulation. But there’s only so much you can do in simulation. You have to be out there collecting real-world data on the sidewalk. We actually have a company value called “Truth Is on the Sidewalk.” It goes to the experimental mindset we have: we put things out there, we observe, we make sure it’s safe, and we collect the data and improve on it.

Teleoperations and autonomy go hand in hand. You can’t really train your autonomy stack without having a teleoperation stack in the robot. Teleoperation basically means someone can remotely come in, take control of the robot, and drive it around. Some robotics companies did only that. They didn’t have any autonomy; they were just teleoperated. That has downsides. If your connection drops, or if you need a very quick, split-second reaction, it’s hard to do that via teleoperation. That’s where onboard autonomy is table stakes. You’ve got to have that. But then there are the corner cases I mentioned previously that you may never have encountered in any model, simulated or otherwise. In those cases, robots basically ask for help, and a human being steps in, takes control of the robot, and resolves the issue. Maybe the robot encountered police tape it had never seen before and didn’t know what to do. Or there may be conflicting signals, like someone telling the robot to go this way while another person tells it to stop. There’s a myriad of situations a robot may never have encountered before and needs to react to. With our safety consciousness, we really care about the robot behaving safely. When there’s uncertainty, we ask a human being to step in and help.

Another area where we have humans taking control is intersection crossings. We do those fully teleoperated today, because that’s where a sidewalk robot has the highest potential risk of a safety incident. Of course, we think that as autonomy improves, more and more of that can be done autonomously, but again, we’re taking a conservative approach. The same goes for the robot’s speed profile. Our robots can go seven miles an hour on a sidewalk, but we usually don’t push them that hard, because we want to make sure that whatever speed they’re going is a speed our autonomy stack is comfortable with.

Jeff Dance: That’s interesting. One of the thoughts I had yesterday when we were talking about fleet management was about the future, where we have a lot of robots out there, including last-mile delivery robots. In some sense, you’re replacing what people do, but you’re enabling a better experience, and it’s better for the community. The thought that came to mind was, look at cars. The car industry is a big industry, but the service industry for cars is slightly bigger over time. So I was thinking about a future where we have fleets of robots and a growing industry where people have to maintain them and, as you said, come in and teleoperate. Tell us a little more about that side of the business, because it’s the unseen side. How do you think about scaling it? As you scale up, is it a linear thing? We’re bringing in more jobs and different types of jobs, and that’s an important part of the market.

Touraj Parang: Definitely. There’s an immediate effect when we go into a city: we set up a depot, and we hire local technicians, field operations people, engineers, and so on. Then there’s also the teleoperations dimension. But I would say even bigger than that immediate job creation is the fact that autonomy isn’t just a labor-saving technology; it’s a demand-creating technology. DoorDash did a study that I saw, and you can see this in different ways, showing that a one-dollar drop in the cost of delivery translates into 6% or more growth in order volume, or demand. Delivery is pretty price-sensitive. So if you can radically reduce the cost of delivery, say down to a couple of dollars versus what it is today, you’ll see a lot more orders flowing to local restaurants and local merchants, possibly even for non-food items. You have a lot more local commerce, which creates a lot of local jobs. Right now, small businesses are suffering. They have very thin margins, and they’re losing a lot of business to online ordering from Amazon and others. We can bring that thriving local economy back. That’s what I’m really excited about from a jobs angle.

Jeff Dance: The economics of enabling efficient delivery, and what that unlocks for our lifestyles. Who likes waiting in line for a prescription? It’s like, let me tell my watch, “Go pick up my prescription at Safeway, Rite Aid, Walgreens, whatever,” and let that talk to the MCP over there, which then talks to your MCP, and then it shows up a couple of hours later. I’m not waiting for an hour and trying to figure out whether they’re busy or not.

Touraj Parang: A lot of the building blocks of that kind of future are already in place. It’s just a matter of them integrating, coming together, and creating that seamless end-to-end flow. But you’re right. We’re not too far from the future you just described.

Jeff Dance: That’s really neat. As we think about the future, there’s been incredible investment in humanoids in the past couple of years, much to the chagrin of people in the robotics space like us. We’re like, that’s exciting, they’re cool, but they seem farther out. There’s a lot of opportunity for businesses now with intelligent machines, especially with physical AI. But there are also drones. I feel a lot of momentum in the drone space, with things happening at the city level and with companies. I interviewed the CTO of Zipline a couple of years ago, and there’s just so much activity going on with drones. Do you envision other types of vehicles or partners in your network doing the deliveries?

Touraj Parang: Delivery has always been multimodal, and we think it will continue to be that way. You have all sorts of different form factors doing deliveries today: vans, ships, and so on. For the last mile, I think we’re going to have the same thing. There will be certain deliveries that are best done with drones, and certain deliveries that are best done with autonomous vehicles, probably where there are multiple packages, more like a mail run, a large Costco run, or catering orders. And then there will be certain types of deliveries that are perfect for shorter-distance sidewalk robots.

There’s also an interesting interplay between these form factors. I think the video is still up on our website: we did a pilot with Wing, Alphabet’s drone company, where we picked up a food order and handed it off to a drone without any human interaction. One of the challenges with drones today is that they can’t really go to the front door of a merchant and pick up an order. You still need some kind of runner to go a block or two to a drone launch area to hand it over to the drone. A sidewalk robot could be a perfect fit for that use case. Similarly with autonomous vehicles, perhaps there’s a staging area for them, and sidewalk robots can go back and forth. It may seem a little science fiction-y today, but I think that future is on the horizon as well.

Jeff Dance: I loved your parallel to worldwide delivery, where people use trucks, ships, and planes. If you take that concept into the last mile, what are we using? Are we using the autonomous car, the delivery robot, or a humanoid that needs to run up all the steps? Are we using the drone plus that, plus that, and they’re going to meet each other at the park? That’s pretty cool to think about as we envision the future. The tech is there. We just need a little more infrastructure and a little more collaboration with the cities. And with people like you, there’s capital behind this. We’ve seen how delivery has changed with things like Uber Eats. That’s pretty exciting to envision.

Touraj Parang: It’s a multi-party coordination issue as well. You have to work with the food delivery platforms and with POS merchants, and between these different tech companies, to make sure there’s a seamless integration or handoff between them. It’s just a matter of time, but there are a lot of small steps that each party needs to take to collectively make that vision a reality.

Jeff Dance: How have MCPs changed your thinking on that? It seems like in the last year, MCPs have been all the rage for connecting one system to another and then handling, with intelligence, some of that negotiation or an exception. Are you thinking along the same lines as you connect with other systems?

Touraj Parang: Right now our integrations are much more at a B2B level. We don’t have a direct-to-consumer offering. It’s more about our robots plugging into a delivery network or platform like Uber’s, DoorDash’s, or Grubhub’s, and being another entity on their system. Those platforms themselves probably use MCPs for interconnection between their parts. We’re just an integration on a delivery-as-a-service platform. But you’re right, I think we’ll see more and more of that in the future.

Jeff Dance: Maybe eliminating some steps of negotiation. It seems like AI has made integration a lot faster and a lot more possible. I’ve never seen anything like it in my short 20 years of doing emerging technology. It’s been really fascinating. How do you unlock a robot that was built in 1980? Now you can do that overnight with AI and not worry about a whole bunch of integration parameters, and it can make the connection. Really interesting. You mentioned hospitals and your acquisition of a robotics company.

Touraj Parang: Yes, Diligent.

Jeff Dance: Tell us more about that and how you’re thinking about expanding into different environments.

Touraj Parang: From a robot’s perspective, a hospital corridor where people are walking around is not that different from a sidewalk where people are walking around. We realized there was a lot of commonality between what we were building and what Diligent Robotics was building, and they had found a very useful place for robots to help. Nurses are in short supply. They’re highly skilled workers, and their time is much better spent tending to patients and providing lifesaving care than walking down a quarter-mile corridor from one side of the hospital to the other to deliver a tray or a lab sample. Robots are perfectly suited to take that task off their hands and free up nurses’ time to do what they’re really hired to do.

We’ve seen great success with the deployments. I think we have close to 100 robots deployed with Diligent. The robot’s name is Moxi, and I’d really encourage folks to check it out, because their robot is also very cute. Ours is cute on the sidewalk; theirs is cute in the hospital. It has a much different form factor. It actually has one arm, which it uses to push elevator buttons, open doors, and things like that. There’s so much commonality between what we do. We can help their robots be better at navigating, and help build their robots, using our learnings, and we benefit from their data as well. We have this data flywheel. All this data can support every other robot.

Jeff Dance: That’s valuable, and I think that’s where you get ahead. People underestimate the value of the data Tesla has been gathering for the last 10-plus years. I imagine that as an early player, there’s a lot of value in that data as you think about navigating unique environments. If you know how to operate within a hospital from a data perspective, from a VLA and computer vision perspective, and that translates into your end effectors, that means you can work with lots of different robots. And once you figure that out, it’s how do you scale into other industries?

Touraj Parang: Absolutely. That was the thesis for us, and I’m glad to say that almost a year into the acquisition, it has borne out to be true. We’ve also uncovered other use cases. We’re basically building physical AI infrastructure that we want other people to be able to use as well, and Diligent is sort of our first customer for a lot of the things we’re building. But it can be used by others as well. We want to be a physical AI platform for other form factors and other robotics companies, because we know we can’t build all the various forms ourselves. We can share the learnings we’ve had from these robots in unstructured, chaotic environments among people, so that people don’t have to adapt to how robots behave; robots are able to navigate around people.

We also acquired a company called Voysys, which was part of Phantom Auto. We acquired the assets, and Voysys is now a very popular teleoperations and connectivity solution for a lot of robotics platforms as well. So we’re happy about that.

Jeff Dance: That’s really cool that you’re thinking about turning your vertical stack into more of a horizontal stack for others. It seems like that’s where the cloud went, and robotics has been so vertical. I think that’s one of the reasons for what’s been a 99% failure rate: everyone has had to build all the layers of the stack. As you start layering in these horizontal layers, as the cloud did, it allows other people to build on top of that and get there faster.

Touraj Parang: Certainly. We look at it as if we’re almost at the books stage of Amazon’s e-commerce era, and AWS is still to come.

Jeff Dance: Amazing. I really appreciate the time. I only have one or two more questions. What have we not talked about that you want to share as we think about the future of last-mile delivery?

Touraj Parang: I think we’ve touched on a lot of things. I feel really excited about a world where the cost of delivery goes as close to zero as possible, and what that unleashes in terms of possibilities. Reverse logistics is something we’ve always talked about, and it doesn’t exist today. Let’s say you buy two pairs of shoes from your local shoe store, but you don’t know which one fits. The robot brings them to you, you try them on, you put the pair that doesn’t fit back in the robot, and it takes it back. Or peer-to-peer: you may want a robot to take a wrench from your house to your neighbor’s house down the road, or vice versa. There are all sorts of use cases for robots once this utility becomes accessible and affordable. That’s a very exciting future. We couldn’t imagine the existence of Uber when the smartphone first came about. All these use cases emerge once you have access and affordability, and we’re excited about that kind of robotic future.

Jeff Dance: Let’s end on that note, Touraj. Thanks so much for your insights into the future, and for your leadership and vision on how to get there. I’m envisioning Subway University’s curriculum in 10 years. It’s going to include how to partner with robot companies like yours.

Touraj Parang: How to load a robot. Yes, absolutely. Thank you so much, Jeff. I enjoyed the conversation.

Jeff Dance: All right, thank you.