Travis Deyle of Cobalt Robotics on how remote work and robots are converging

Travis Deyle of Cobalt Robotics - Podcast Transcript

Singer:	(singing) it’s Kruze Consulting, Founders and Friends with your host Scotty Orn.
Scott:	Welcome to Founders and Friends podcast this is Scott Orn with Kruze Consulting, and today before we get to an awesome podcast with Travis Deyle of Cobalt Robotics, a quick shout out to Rippling. Rippling is awesome at payroll, awesome at benefits, they also have a great IT service that integrates into all your favorite apps, helps you get new employees ramped up super-fast. Also, unfortunately if you let someone go, you can de-ramp them, you can take away all their access, it’s pretty cool. Travis is nodding his head. Maybe a future Rippling client here, but Rippling is great. Check them out rippling.com, and now welcome Travis. We’ve been friends for a long time. It’s awesome to have you on the podcast.
Travis:	Yeah, thanks for having me, Scott.
Scott:	Of course. So, maybe [inaudible] of you’re Kruze Consulting client, but let’s have a little, just take a couple minutes to kind of tell us how you had the idea to start Cobalt Robotics.
Travis:	Yeah. So, I’ll try to keep it brief, but it kind of goes back all the way to my grad school days. So, I did my PhD building healthcare robots. Robots the size of people with arms, helping people with activities of daily living. So, this was things like fetching and delivering medication for older adults at home. It was things like helping quadriplegics shave their face. Basically, getting robots into human environments, working closely with and around people. Right? And So, this was a big shift in the robotics world. It was fascinating, but the robots were incredibly expensive, think like half a million dollars a piece. So, I actually left and went and did medical devices for about a decade. Everything from cyborg dragonflies as a postdoc, to working at Google X Life Sciences building new medical implants. But about four years ago I decided, you know what, the group I was in at Google X was spinning out one of the first Alphabet companies called Verily Life Sciences, and I said, “Hey, it’s time to go look for new interesting problems.” And So, a buddy of mine, I convinced him to turn down SpaceX to come with me, and just without any clue what we were going to do, we set out in search of interesting problems. What we do is sit down and say, “If you could wave a magic wand, have us solve your biggest, most pressing problem, what would you have us do?” And we went through probably 50 or 60 of these, we had our own criteria, right? Like product to market in a year, paying customers before we built anything. Software only, clearly failed at that one. And then nothing we could see doing and loving for a decade. And as we were doing this, one of the people we spoke to was an old friend in physical security who told us during the day they have incredible support staff, right? They have security officers, receptionists, facilities managers, just an incredible amount of support staff. But when it came to after hours, right? So, like nights, weekends and holidays, they couldn’t find people or justify the cost to have someone wandering around an empty office building at night. And yet, many companies spend millions or even billions of dollars on hiring expensive security officers to go do this. And So, he asked us point blank, “Why can’t you guys take those robots you were building 15 years ago and use them for this task?” And we were naturally very skeptical and knew nothing about physical security, but we very quickly realized that, “Hey actually if all you do is take a Rumba, which is old 20-year-old technology now, add a computer to it, have a camera to do two-way video chat and then add a lot of this autonomy and AI features to detect things, you could basically achieve this exact task. You could do security patrols, you could check for facilities issues, like leaks and spills. And then at any time, the robot doesn’t know exactly what’s going on most of the time, and So, if it has uncertainty, you can just remote-in a remote human operator to sort of provide that guidance and intelligence. And So, you get basically observation and reporting on steroids. But beyond all that, it’s not even one officer, you actually get the entire collective intelligence of an entire remote team. And So, yeah, we took off and started executing from there.
Scott:	That’s incredible. And I’ve actually seen the robots, people might be visualizing a Rumba, but that’s not what your robots look like. It’s pretty big. It’s pretty amazing to look at, you want to talk about it a little bit?
Travis:	Sure. So, it’s about five feet tall, right? So, it can see out over a cubicle farm or anything like that. They’re exclusively an indoor robot, but they’re covered in fabric. Think of it like a giant Sonos speaker, like scaled up. Up at the top, it’s got a microphone array So, it can hear sounds coming from different directions. It’s got a touchscreen interface on the front. That’s how you have that two-way video chat. It’s got cameras all around. That’s kind of the security piece. And then it’s got various depth cameras and LIDARs So, that it can safely move around an environment. And So, we really designed the entire robot to be friendly and approachable, right? So, it’s this soft fabric, you can go up and push on it, it’s engaging and then at any time, you can tap the screen and get a remote person on. You’re not talking to like an Amazon Alexa or Google Home. It’s literally a human being, a real live person on the other side that you can speak to. And yeah, the thing navigates around. It’s pretty slick actually. We were always concerned about that cultural adoption piece. That’s the big change that has to happen. And So, designing a robot that is friendly and approachable first and foremost, right? It’s the friendly, helpful workplace robot, that’s the hard part. Making a scary robot is much easier. But yeah, you can literally jump-
Scott:	[inaudible] time for Halloween.
Travis:	Yeah. But just the insane ability to jump into a remote robot body anywhere in the world, see what’s going on and interact with people, it’s a pretty profound change in the way that businesses operate. And robots are all about exploiting constraints, and So, we target indoor spaces only, right? That’s offices, warehouses, museums, data centers, manufacturing facilities, basically any indoor space, hospitals, schools, our robot follows ADA compliance rules. So, we know what the mechanical specifications are, have good wireless connectivity. We can build a system that’s really robust and reliable and it doesn’t get sick. It doesn’t fall asleep. It doesn’t cause HR problems. It’s consistent in the pricing. It’s all the wonderful things you expect from technology, but still backed by people when and where you need it.
Scott:	Yeah, I was a security guard in college at night sometimes, it was one of my college jobs, and let me tell you, the not falling asleep part is very critical here, because I did take a few naps. So, you’re automating it, it’s a challenging job for anybody. I actually did that job in college 20 years ago. It’s not like the greatest job. It actually pays decently well, because it’s So, hard, but it’s not something that humans really enjoy doing all night. So, it’s actually, it’s a real nice benefit to society [crosstalk 00:07:16].
Travis:	And you know, we did a bunch of those positions too ourselves, just to understand the space, right? Both in-person graveyard shifts and then being that remote robot operator in the early days. And I can tell you it’s hard to pay attention, be alert all the time, and to deal with all of this frankly with boredom, right? Robots are good at the dull, dirty and dangerous. Nighttime security is like the dullest job you could ever imagine. And So, yeah, we pull that person remote, and then now they stay active and engaged all night, right? They’re talking to people, they’re basically playing a video game and sending the robots around and doing new patrols. It’s much more engaging and fulfilling, and people are generally, they want to be good at their job, they want to provide value, and we help them do that.
Scott:	Do you employ the operators? Or is that someone that the company that buys the Cobalt system hires? Or how does it work?
Travis:	Yeah, we do employ them. It’s part of our service, right? So, we offer security as a service, just by way of the robot. We have two command centers that are operated 24/7, or they’re staffed 24/7. And So, yeah, we employ those people directly. We train them up, they do a bunch of different things, it’s not just a security, right? They’re customer service agents engaging with employees and partners, they debug robots if the robots ever have issues, they adapt patrols. They’re doing analytics and sort of security intelligence on the data that is coming in. They look for facilities incidents, they manage these things. So, we call them specialists, and yeah, they do many different tasks. Yeah. Full time Cobalt employees and we really enjoy and love having them on board.
Scott:	I love the analogy a video game, because I can totally visualize what it must be like to do that job. And it probably is, especially the debugging or when you encounter something that’s off-script that you’re not used to doing, or someone you’re not used to seeing in the office at night, it must be kind of crazy.
Travis:	You’ll love this then. So, we worked a lot in this space in the past and during the early days of the robotics sort of revolution, we literally modeled it off systems used by NASA to control the Mars Rover and then crossed with video games like StarCraft. And So, you can literally go in and actually you as an end user, you get the same visibility. You can jump from your site here in the Bay Area to your site in New York, to out in Singapore or whatever. And you can literally just go and click on a map and the robot will automatically dispatch itself and go over and look at whatever you want.
Scott:	Wow.
Travis:	And So, we take that first line of control So, that you don’t have to worry about it. But if you want visibility, if you want to go in and see what’s going on, you as an end user can do the same thing. You have that power.
Scott:	That’s incredible. You also talked about the cultural aspect of this and how important it was to you and your co-founder and the whole team to have the robot be friendly. What was the process of that? Did you do like, have people come in and interact with robots and just watch how they talked to them? Or you had all this experience, but how’d you make it So, friendly?
Travis:	Well, So, it turns out, from our research with older adults, people are actually very receptive to robots. If you look at a Rumba vacuum cleaner, people name them, they-
Scott:	We have a name, I forgot her name, but we have a name for our one too.
Travis:	Yeah. So, people inherently, they will anthropomorphize a robot there, and it becomes part of the team, right? And then it’s really just about trying to make it not sci-fi and scary. So, like a big plastic shell of a robot, it’s not approachable, it’s not friendly, it’s not sort of the soft engaging thing. And So, we tried to just lean heavily on our background there. We have some wonderful advisors like Leila Takayama is an old friend of mine, human robot interaction researcher. We worked with Yves Behar, Fuseproject, brilliant designer, to try to make this thing sort of fit into a high-end office space, right? Because that’s the most demanding from a design perspective, right? It has to look professional. It has to be approachable and friendly. If you’re in an empty warehouse, it doesn’t matter as much, but if you design for the hardest spot, then you can repurpose it to these other locations. And So, yeah, we tried to lean heavily on all of that, and then really to emphasize the interaction through the screen with remote people. Because if you think about your Amazon Alexa, if you ask it a question, for example, a significant fraction of the time, it doesn’t have an answer. That’s really bad when you’re in a safety and security context. Let’s wait for the technology to mature to the point where it’s really good, and for now, we can leverage a human to come on and provide that interaction when and where it makes sense. And So, it’s just being very diligent and disciplined about how to deploy these things. What things to emphasize, what things to de-emphasize.
Scott:	Yeah. Do you ever in our COVID world now, do you ever think that you guys were like a forerunners of this? Because in essence, every time I talk to someone now it’s that same type of interaction. I’m talking to everyone through Zoom on my computer, on my phone, are there lessons for all of us that we can repurpose for our just general lives?
Travis:	I think So. If you think about remote work, technology is letting us be anywhere on demand, right? We can [inaudible] project our presence anywhere, and the robots are a physical manifestation of that, right? I don’t need you to grab your laptop and move around and show me something. I don’t need you to walk around with your phone and do it. I can literally use the fact that mobility is cheap, to create a robot that can move and do that for you. That’s the observation and reporting part. I think we’re still in the early days, but you have to remember, we were building robots 15 years ago that would grab a razor and shave someone’s face, or reach out and open the door. And So, robots really represent mobile computers that can reach out and touch the world. And that’s, that’s a profound change, right? So, if you look back over the last 30 or 50 years, we’ve had robots in manufacturing, we’ve had them in warehouses, we’ve seen a little bit in our daily lives, right? So, you have some of the parking assist and lane assist in your car. You have a Rumba vacuum cleaner. Heck, you could even count a dishwasher or washing machine as a robot. It’s just as soon as it works, it’s no longer called a robot. But these things are pretty commonplace. It’s just that people have a lot of misconceptions because of Hollywood, right? You think like, Terminator and Robocop and whatever. But what you really need to think about is like WALL-E or a Big Hero 6, right? Like the Baymax helpful, friendly healthcare robot. And So, what we have to do is change our mindset, because robots have the potential to be hugely impactful, right?
Scott:	Oh yeah.
Travis:	To help with healthcare, cleaning, delivery, safety and security. Literally, every single aspect of our lives will be impacted by robots. And it’s not just labor replacement, it’s literally taking a microwave and the utility of a microwave and now making that utility apply in a vast array of additional tasks. And So, that’s-
Scott:	Also, like friendship. Like I was talking, I mean Japan, I know I think it’s robot dogs or there’s robots that are kind of are thought of as family pets and things like that. Maybe it’s too cliche or stereotypical. But I was talking to my grandma last night who’s 80 something years old, she cannot be out seeing people right now. And So, I was just thinking how great it would be if she had a friend in her house. There’s just So, many applications for robots. It’s probably… You must wake up every day and be like, “I could start three companies right now if I wasn’t doing Cobalt. There’s So, many other opportunities to go after.”
Travis:	Well that’s an interesting example, because there is this companion robot called Paro, P-A-R-O, that is this little seal and it’s been in Scandinavia in nursing homes as a companion, and it’s a little stuffed animal, and it actually has big psychological benefits for those end users. The real trick for all of robotics, in the end, is to identify a clear value proposition where the technology of today can actually address it. Because everybody’s seen the crazy videos for Boston Dynamics of the wild robot doing absurd things. And they are some of the best robotics and controls and mechanics people in the world, but they have no application for them. And So, what you have to do is you have to find that application where you can deliver more value than the cost. I mean pretty simple business economics. And then from there you can layer on new capabilities. And So, when you think about Cobalt, we really started with just this observation reporting, very simple anomalies that we would detect, right? So, this robot rolls around, builds a model of normal, and then flags, things like people, motion, sounds, open doors, leaks and spills, unattended packages, whatever these safety and security things are. But we can keep layering more and more on. So, to give an example, we recently rolled out cyber security checks, right? So, if you are a big enterprise, you’re required to protect your information, your IP, and that means things like not leaving unattended devices on your desk.
Scott:	Oh, I didn’t think about that. Computers [inaudible] right?
Travis:	Yeah, not leaving your screens unlocked. You’re supposed to erase your whiteboards. We go through and we do those checks, and we do it with the machine, right? So, you get a hundred percent compliance checks to make sure that people are adhering to these standards.
Scott:	Wow. I hadn’t thought of that. The whole reason we’re doing the podcast was like three weeks ago you and I were emailing, because I was just kind of curious. I was actually like, is COVID actually spiking demand for Cobalt. I could totally see that. I’ll leave it to you, you can report some of the good news here.
Travis:	Yeah, So, the challenging news was initially, no, because every security director on the planet was like, “Holy crap, we have this giant existential crisis happening.” And So, they all started worrying about how to do that. With the exception of our existing customers, who all started saying, “We want to go full 24-hour coverage, we want to add new robots.” And So, we were like, “Wow, this is obviously a very good solution for this time.” And now what we’re seeing is, as people settle into this new normal, people are realizing that this is the ideal time to not put people in harm’s way, right? And the same things, the same capabilities that we were doing before are extremely relevant during the time of coronavirus, right? So, doing temperature checks. Our robot has a thermal camera on board. We can heck to see if someone has an elevated body temperature.
Scott:	Just let me interject, because before we turned the mics on, I was going batshit crazy that you guys can do temperature checks. Because my wife and I have talked about this, really her, she’s smarter than me, she’s like, “That’s the future. That’s what’s going to need to happen.” And So, you told me before we turned on the mics that you can do this now. My mind was totally blown, because that’s just So, amazing. That’s just exactly what we need right now to be able to get people back to work.
Travis:	It’s certainly part of it, right? But we already know that at least a quarter of people are asymptomatic, right? So, it’s not a panacea, it’s not a silver bullet. And you can’t just have a camera there doing the check, you actually need to react and take actions. So, to operationalize the camera, it’s great because we have a remote security officer. So, if they detect an elevated temperature, you can remote-in a medical professional. You can remote-in a customer service agent. You can remote-in HR to say, “Hey, please go home. You’re covered under your benefits.” You can say, “Hey, please go over to this extended screening area.” So, it actually helps you operationalize the body temperature checks. And So, that’s what we’ve been rolling out right now. But we already do PPE checks. So, for face masks on coronavirus, making sure that everyone is wearing their mask.
Scott:	IT could go around a hospital and just do that and just see. Oh my God.
Travis:	Hospital, office, wherever and say, “Hey, is everyone wearing their mask?” And if they’re not wearing it, we can go up to them and politely request, “Hey, can you please wear a mask? It’s really important.” And So, you’re driving the behavior change that is required to actually address coronavirus. We were already doing checks like that, right? So, in manufacturing facilities, we would check to see if people were wearing their hard hat wearing their safety glasses. For us, it’s just a slightly different machine learning model on the computer vision to check now for face masks. So, that’s great.
Scott:	That’s incredible.
Travis:	And then finally, social distancing, right? So, we do people detectors is one of the computer vision algorithms we run, right? To detect bad guys and good guys in your space. And So, we can tell you if there’s aggregations of people, right? So, if people are sitting in a lunch area or something, they’re all too close to one another, the robot can detect and flag that, and then it can come up and politely remind people like, “Hey, we need you to spread out a little bit more. The social distancing is really important.” And again, it drives that behavior change to actually get that response. And So, yeah, we literally are at the frontline of this stuff, and it can be really helpful as people figure out how to go back to work.
Scott:	It’s incredible. You mentioned something that I thought was great, I can relate to this in a different way, but you said everyone was freaking out except for your customers. Your customers, that must’ve been a really great moment. Because people who listen to this podcast know starting a company is hard and building a startup’s hard. In a way, that’s the ultimate validation for you and your team in that moment where they’re calling you and just saying, “Hey, thank God we have you deployed already. Can you go 24 hours? Or can you do X, Y and Z?” That must’ve felt really good.
Travis:	Yeah, it’s certainly validating for us in many ways. And we’d seen coronavirus coming pretty early, and So, we had taken steps to operationalize and address everything on our side. So, to the extent that we could, many of the remote operators were shifted to work from home. So, in the same way that the security director might be at home, now our remote specialists or remote operators were also able to work from home. We couldn’t do that across the board. Some clients had very specific regulations about not allowing that, So, we were able to accommodate. But it was really fascinating to see that this entire distributed model that we’ve been referring to is actually ideally suited for what we’re trying to do right now. And I think the entire world is going to have to change its operating procedures to address social distancing and coronavirus and risk and things like that. So, it’ll be interesting.
Scott:	One of the other just kind of takeaways, we’ll wrap up here in a second, but I also feel very happy for you personally, because just going back to your story, you worked on robots 15 years ago, So, you’ve been in this industry for a very long time. You’ve seen it evolve, and just some of the lessons I’ve heard you articulate here about using robots for an application that actually has an ROI and that does create more value. That’s obviously something you learned in your early twenties when you were working on the medical robots that just weren’t cost [inaudible] on a cost-wise. That also probably makes you feel pretty good looking back, it’s kind of rewarding, right?
Travis:	It is. Ironically, when I finished my PhD, I thought I would never work on robots again. There is this weird thing where roboticists love making cool robots and not solving real problems. And So, I kind of swore it off. I was like, “I’ll never go back to robotics.” And then when this application popped up I was like, “What do you mean no one is doing this? How can no one be working on this already and doing it in a sane way?” And yet it’s a very old idea, right? The very first mobile robot ever was in the mid 1960s, it’s actually at the Computer History Museum down in Mountain View. And literally this was one of its intended applications, right? In the 1980s, they had security robots patrolling around here in Palo Alto at like HPs campus. They would bury a wire under the ground and the robot would follow the wire and look for fires and stuff like that. And So, it’s this really, really old idea, and it’s just that all the component technologies are finally at a level of maturity and a level of cost-effectiveness that you can put them together. So, if you think about what those are, you have a wireless connectivity, WIFI and cellular are just everywhere, you have mobility. So, think like your Lime scooters and things like that, you can get good motors. Your sensors are amazing, camera systems are great. LIDAR is still a bit expensive, but you can get depth cameras like the Microsoft Connect, right? These sensors-
Scott:	Oh, no way [crosstalk]
Travis:	Yeah, they used to be like $19,000 a piece, now they’re like 200 bucks. Computation is cheap, and the machine learning is vastly better. Video chat is a phenomenally robust and capable technology that enables it. And So, it’s all of these really simple, straightforward technologies. And then you just have to do a really good job at integrating them together. But yeah, it’s eminently feasible. And the unit economics are solid.
Scott:	And you’re going to be able to ride the component cost reductions on all those things for many years. They’re all getting cheaper, constantly. So, that’s amazing for Cobalt’s gross margin and the future and it’s going to really help the company. That’s really cool. Well, this has been awesome. COVID has affected all of us, and So, it’s really cool to kind of talk to someone who is, in a way, that’s not benefiting from it, but helping other people who need a new solution. You guys are stepping into the void, which is super exciting and I’m just really happy for, you. I know how hard you’ve worked on this and it’s really cool.
Travis:	Thank you.
Scott:	Yeah. Well, do you want to tell everyone where they can find more information on Cobalt Robotics and how to reach out.
Travis:	Cobalt Robotics, again, we’re safety, security and facilities management robots. You can find us online at cobaltrobotics.com and you can personally reach me, Travis Deyle I’m just travis@cobaltrobotics.com. Happy to reach out and talk.
Scott:	Awesome. Thank you again for coming on the podcasts. I really appreciate it.

Travis Deyle of Cobalt Robotics on how remote work and robots are converging

Explore podcasts from these experts