William Santana Li, Founder and CEO, Knightscope security robots

Ali Tabibian:     Welcome, welcome, welcome everyone to Tech. Cars. Machines., and this time it’s going to be Tech. Cars. Robots. As you know at Tech. Cars. Machines., we spend a lot of time talking about how sensors, connectivity, and software advances are changing the worlds of cars and machines and in the case of cars obviously those cars already exist. In the case of machines, we’re typically talking about machines that already exist that are being influenced by advances in technology, but what classes of machines are uniquely enabled by some of the technologies we’ve talked about? Cameras, LIDAR, image recognition, autonomy, mobility. In what circumstances our customers actually paying for something made of this stuff and uniquely enabled because of this stuff? We’ve got a great example for you today. We are excited to bring you Knightscope, one word, Knight as in Teutonic Knights and then scope, Knightscope.

Knightscope make security robots that roamed the halls, parking lots, and facilities of a rapidly increasing number of large customer sites and by that I mean the customer sites of large entities and these can be corporate headquarters. You can see it outside of Samsung headquarters in the Bay Area, malls, hospitals, casinos and large institutions like that. Now, in the episode notes, we are going to link to pictures of the Knightscope robot, which already kind of fascinating to take a look at. I encourage you to take a look either via our links or via the company’s website at what these entities actually look like. What I’d like you to imagine a shape similar to Salesforce Tower in San Francisco or The Lipstick Building in London also known as the Gherkin building. Make that shape between three feet and five feet tall depending on which model you’re talking about, make it out of white fiberglass and you have the basic shape of a Knightscope robot specifically the K3 and the K5 model numbers.

These units sit on wheels and the wheels in these models, especially in the original K5 are somewhat hidden behind the white shell as the shell drapes toward the ground. There are a number of openings on these white fiberglass exterior. Generally, the openings are small and circular. There are a couple of mouth-like openings. Imagine the slit that’s in the chess piece called the bishop if you ever play chess and you have a couple of those openings, one along where the mouth would be or the face of this obelisk-like unit would be and the other one along the belt line. While the units are generally smooth in shape, they’re a little bit pointy-headed so to speak and that’s because the very important LIDAR unit, which is we’ve discussed before is a type of light-based radar system, which itself is shaped like a thick puck sits right on top of the K fiber, the K3 units where it has the best vantage point, the highest elevation of the device to be able to monitor things around it.

Knightscope refers to these things as autonomous data machines and that’s really quite descriptive. They go indoor, they go outdoor. Generally, they’re able to map the area into which they’re deployed automatically and therefore maneuver their way around the expected layout of the environment as well as unexpected entities of people or cars that might be in the environment, and they generate a whole host of data, obviously live video, live audio, but also thermal detection. They’re able to be remotely managed and many other features as well. Generally, by the way, the devices are autonomously recharging. They go find their own charging unit and basically charge themselves up and get back to work.

The company was founded about five or six years ago and about two years ago is when it started deploying its prototypes in the field and it’s been shipping production models for a little over a year or so. Let’s go to the very determined and the very passionate founder of the company, Mr. William Santana Li and this interview was done at the company’s headquarters in Mountain View, California.

Voiceover:       Tech. Cars. Machines. Subscribe here or at gtkpartners.com.

Ali Tabibian:     Bill, it’s so nice of you take the time here.

William Santana Li:       No, absolutely. Thanks for having us.

Ali Tabibian:     Bill, I’ve followed you for a few years now and I always felt you were emotionally connected with a problem you’re trying to solve that has run through a lot of things you’ve done. When I talked to you, you talked about the issue because you care about it and then you came up with Knightscope to treat the issue. Let’s go back to the company’s genesis and answer the why and why at that time question. Why did you think about starting it and what was right especially from a technical perspective for you to start Knightscope when you did?

William Santana Li:       Sure. Well, part of it is personal, part of it is professional. One of the things that drives me from a professional perspective is I think self-driving technology, autonomous technology turned the world completely upside down. As an ex-Ford Motor Company executive, I know the space, I know what can happen and possibly what can’t. I think a lot of the big players that have these humongous R&D projects, it’s going to take a very long time to properly commercialize it with not just the technical hurdles, but the regulatory insurance and legal frameworks required to put a 4,000 pound unmanned vehicle and bet your $100 billion brand that everything is going to be all right at five o’clock on Friday in Time Square.

Because Knightscope were building these autonomous security robots, we are now scaling nationwide. We’ve hold about contracts in 15 different states. We’re the only company in the world that is scaling autonomous technology in the real world with real clients, doing real work, and that gives investors an interesting way to get exposure to self-driving technology. That’s kind of the professional side of things. The personal side of things, when we start with our long-term mission is a little bit grandiose, which is to try to make the United States of America the safest country in the world and change everything for everyone. I was born in New York City. Someone hit my town in 9/11 and I’m still profoundly ticked off about it and I’m dedicating the rest of my life to better securing this country. When we started the company back in 2013, so in a couple of weeks is going to be our five-year anniversary.

Ali Tabibian:     Congratulations!

William Santana Li:       Thank you. As you know, 95 plus percent of startups fail, 80% never make it to their third anniversary, so we are going to have a little bit of cake and celebrate. Right now, if you look at self-driving and autonomous, it’s kind of hot. Robotics is hot as well as AI fad stuff is hot. It’s exactly what you’ve said, our clients have a problem, we want to fix it. I really don’t care what technology is required to do that. When we started the company, it took me 364 days to raise the seed round like no one wanted to do this. Something along the following, “Bill, you’re out of your mind. This will never work. Two, you’ll need $15 million to build the first prototype and it probably won’t work. It’s hardware and software, it’s too difficult, you should pick one. Lastly, physical security is not an investment thesis, so go away.”

We got the no after no after no and so we did what most good entrepreneurs do is pretty much ignore everyone and just do what we said we’re going to do. Now, we are scaling our operations, our clients, our corporate campuses, malls manufacturing facilities. We put our first machine at an airport. We just signed our first casino and we are operating 24/7 in four time zones. It’s been quite a ride and what we’re doing is extremely difficult to have an autonomous security robot patrol outdoors or indoors 24/7 generating 90 terabytes of data a year that we analyze for the obvious monitoring surveillance, but also deter a lot of negative behavior. We have and still do have a very focused mission and nothing is going to stop us from doing what we intend to do.

Ali Tabibian:     You touched on obviously the robot itself, but also monitoring solution. Maybe just tell our listeners what are you offering the customer. Is it a piece of equipment? Is it an outcome or something in between?

William Santana Li:       I think we’ve got different clients with different needs. It’s a very fragmented marketplace. We have some clients that are looking to give their security teams really smart eyes and ears for them to do their jobs much more effectively. How do I cover more ground? How do I give my team some tools so they can cover more ground together with the machines? We have some clients that have a genuine crime problem, literally one to two criminal incidents a week. Theft, assaults, stolen vehicle, you name it. Then the third one is certainly a minority despite all the media talking points is purely on cost and how do I get rid of that guy. It’s a mix bag, but these machines, the K5 that runs around outdoors, it’s five foot tall, three foot wide, 400 pounds.

If you don’t spend a lot of time on security, you realize that very simply if I put a marked law enforcement vehicle in front of your home or your office, criminal behavior changes. Most of these guys and they’re mostly guys [inaudible 00:09:36] and they’re looking for path of least resistance to just get away with something. You put a 400-pound machine there that’s roaming around autonomously and people don’t actually know what it does. That’s going to deter a lot of negative behavior. So much so then a lot of cases with our clients, they’ve experienced one incident, two incidents a week or what have you. Put the machine there, it’s literally gone down to zero. We know that portion is effective.

Then the other is can you give 360-degree eye level streaming of HD video and record any kind of detections that might be of interest to me. Let’s say I’m a property manager. Between 11 p.m. at night and 6 in the morning there should be no one waltzing around our property. If you see someone, let me know or, “Can you run a thermal scan in these certain environments because there’s a high risk that that pipe might actually blow?” It actually has occurred a few times where we’ve been able to help. We stopped a fraudulent insurance claim where someone said, “I slipped and feel and we have evidence that suggest otherwise.”

Ali Tabibian:     Nice for quick settlement.

William Santana Li:       Yes. We’ve stopped the corporate vandal. We’ve helped the law enforcement agency issue an arrest warrant for sexual predator. We helped the security guards catch a thief. The list goes on and on where you can use technology and give the guards almost super human capabilities.

Ali Tabibian:     It sounds like at this point you have enough deployments where you can actually evidence with almost statistically relevant sample, the crime reduction benefits of your deployments.

William Santana Li:       I guess to my shock and surprise I would be happy to report that we’ve had over a dozen crime fighting wins thus far with just initial deployments we’ve done to date, which to us is fascinating and now that we’re having … We offer this on a machine as a service business model and now we’re starting to get the renewals coming in and it’s always fantastic to have clients not only experienced a return on investment in terms of either cost reduction, but they want to keep the machine because it’s actually been very effective.

Ali Tabibian:     There seems to be a fairly good amount of before versus after data now that you can evidence in terms of this is what has happened before we were there and this is what’s happening.

William Santana Li:       Especially for clients that are in rough neighborhood, let’s put it that way where it’s pretty obvious that we can be helpful.

Ali Tabibian:     All right. That’s pretty impressive and I think to go back a little bit to what you said about the role of autonomy and the applications of autonomy in the world. It’s not clear that the really big changes will come really soon, but if you find the right application and the right solution, it’s amazing how quickly restarting five years ago you’ve been able to make a difference. We’re still waiting for the cars to drive themselves as an example.

William Santana Li:       Today, there are a lot of regulatory requirements where folks working on self-driving cars needs to actually have a human inside the vehicle to take over because depending on what you believe 30% to 70% of the time the algorithms fail because they’re not yet robust enough to be operating. Despite what people might think, there are no people inside of our machines, so we actually have to be right 24/7, 100% of the time. To be able to do that and say, “We’ve operated more than 300,000 hours in the field with real customers and traveled collectively more than 150,000 miles. One of our machines have gone the distance of from San Francisco to New York and back twice over.” I mean, we’ve got a lot of field experience to be able to do that. We are certainly one of the pioneers in commercializing autonomous technology. It’s going to take some time to have that technology be able to mature that much more, but engineers are really good at solving problems with constraint boundary conditions.

Ali Tabibian:     Exactly.

William Santana Li:       Very simply we operate less than 25 miles an hour in private roads. We don’t need any regulatory framework or some other conversations that will allow us to operate legal in all 50 states for us to do what we’re doing, but the challenge that everyone is going out aggressively. Listen, I’m sharing them on. I want them to succeed, but you give a bunch of engineers, “Here’s a random location, random environment, random conditions, random everything, please go solve the problem and do it quickly.” I think it’s going to take a little longer than people would really expect. You’re going to see all those incremental improvements and the press release that someone built another prototype, a press release that someone added a feature incrementally over a decade or two something actually real might happen, but it’s going to take a long time. We know how difficult this can be because we’re actually living it for real.

Ali Tabibian:     You at Knightscope know how difficult it can be.

William Santana Li:       It’s different to build a prototype or run something on a track or do something limited testing. It’s a lot different to have an actual binding contract and have a paying client and need to know that you need to deliver 24/7. Here’s the funny part. Hollywood has done us a service and a disservice because most of our clients, the autonomous portion, which is fascinating to investors and fascinating to technologist and car [inaudible 00:15:03] and the like. Security folks don’t care. Like, “It’s a robot. I’ve seen it on the movie screen. Obviously, it should be autonomous.” That’s like just even have to start the conversation so they don’t understand the pain and suffering to actually get to that point, which is the hardest part like, “Okay, what else can it do for me?” That’s where we really need to shine with actual capabilities and value that we can provide the client because in this particular case, the client doesn’t care that it’s autonomous. It’s assumed that a robot can just go anywhere and do anything. We’ve got some expectation versus reality mismatch in the short-term let’s say.

Ali Tabibian:     Bill, maybe actually that’s an interesting point. Tell us what’s under the hood. When I first visited you here, we talked about the 360 cameras, LIDARs, licensed plate scanners and I don’t know if those [inaudible 00:15:53] independent features. Give us a little bit of a sense what’s inside the box and then maybe walk us through what a deployment would look like. You go there maybe you light our map, the location. What do you do? How do you train the individuals who are going to work with this thing? If you think that’s an interesting thing to go through, please take us there.

William Santana Li:       Sure. It’s analogous to self-driving car. Obviously, we’re operating at different speeds and distances and that makes them actually material difference on how you approach it, but it’s a combination of LIDAR, sonar, wheel encoders or similar to higher odometer works, accelerometers and then bunch of crazy software to real-time dynamically map a location. We don’t ingest blueprints or architectural renderings or anything like that. The machines dynamically create their own map and then they need to figure out how to dynamically find themselves in the map that they created. That’s a little bit of a trick.

A lot of it has to do with probabilities and we do hundreds of simultaneous equations of probability that based on all the sensor input from all the variety of stuff that we get out of 504 calculations that we are confident that we know where we are and we are in a certain orientation. If there’s nothing in the way, then I can move. That’s kind of a very quick layman’s explanation of how this work, but they’re effectively self-driving cars but in my case a little bit easier to build. We don’t have to worry about seats and airbags and glass and all kinds of other stuff I used to deal in Detroit.

In terms of deployments, usually from signing of a contract, we tell most clients no more than 60 days to schedule and get stuff built and that sort of stuff. The fastest deployment we’ve done was 48 hours, which is not the norm. Usually probably the more sane is 5 to 10 days, maybe five days on site and it’s also highly dependent on the location. Some of these deployments are very, very large. We’ve got to think through the whole thing. Something is simple as, “Okay, where do we put the charge pad? I know there’s an outlet out there. Does there actually power?” We also run on Wi-Fi and/or LTE and then you’ve got to go through the dance of, “Well, is AT&T, Sprint, or Verizon, or T-Mobile great?” Then people that don’t have field experience wouldn’t know this but sometimes telecoms aren’t consistent.

You at 5:05 in the morning and in SoCal one of the telecoms is down. We can’t operate that way. We actually have failover backups. We’ve got Sprint as a primary and Verizon as a secondary or depending on what the signals look like in that area. That kind of experience is really important to be able to do that. It takes some time. In a lot of cases, some of the time is the interaction with the clients because our primary client is typically the chief security officer. Because of the nature of the product and it’s so new and everything else, it’s actually interesting organizational behavior because the CFO wants to come down and understand the economics. “Again, how much is this saving us?” The CEO is fascinating with forward-looking technology and then the chief marketing officer shows up and is like, “Hey, can we brand this thing? What can we do in terms of PR messaging?” Then the HR people come down and is like, “How do we explain or how do we share?”

We have one of our clients like make a custom cake and have a big thing about it. I think almost all of the machines are branded and almost all of them have a name like they have name tags issued, employee badges issue. One of our clients has a Twitter handle for the machine. The other one has an Instagram one. Another one built an entire website dedicated to their machine. It takes a whole life of its own. That eats into some of the “deployment time” if you’ve got to go setup for the photo shoot and the media won’t have you. Then we come back and we try to remotely take a few days to clean up some of the data that we gathered to get things to go forth.

Now, we’ve start to build our own tools because no one has ever done this before. We have our huge pre-deployment checklist that we go through before we sign a contract to make sure that we can actually do what we said we’re going to go do and then we have a huge deployment checklist of all the stuff that needs to get done and give status reports to where we are with the clients or they have some visibility and transparency. Yes, actually this week, we are adding the Mountain Time Zones so we’ve got folks in Colorado. We’ve got East Coast. We’ve got Central Mountain and Pacific.

Ali Tabibian:     You’ve got basically a repeatable template both from in terms of the device and its feature sets as well as the deployment to rinse and repeat.

William Santana Li:       Rinse and repeat with the small exception that as we get into new verticals or locations, you always learn something new at a different location like this is our first winter we went through. That brings some learnings with you. Different clients have different ways they communicate so we have to build some tools on how the team interacts. One of the things, we’ve been rinse and repeat as you say, malls kind of pretty custom. Hospitals, we’ve done a lot, manufacturing facilities. We got at the point that we’re starting to scale up and we’ve got some pretty large contracts in the works, which is going to be a lot of rinsing and repeating.

Ali Tabibian:     Excellent. That’s great. One interesting thing that occur to me as I was listening to you, you’re describing essentially a pretty welcoming approach by where these deployments are going and then there’s the part of which you can read in the news, which is about the robots are going to kill us.

William Santana Li:       They’re all coming here to kill us.

Ali Tabibian:     To kill us all and then there are the stories, I don’t know how much [inaudible 00:22:13] at the SPCA and the [inaudible 00:22:17]. Tell us about what you’ve learned about human machine interaction both maybe some of the features you’ve build in to make sure these devices are both present the security posture, but at the same time aren’t threatening to good people out there or how they avoid collisions with them, etc., but also just some of the things that are purely a social dynamic of how people get used to these machines and whether they do or not and what style do they wind up getting used to the presence of a 400-pound outlet.

William Santana Li:       When we first started, probably the highest risk in the business was would “society” allow us to do this? We are scared in our minds. The first time we put a machine on the field was May of 15 and we just sat back like I don’t know what’s going to happen. What was fascinating was it became like robot selfie time. It’s definitely a kid magnet, which has its own challenges and opportunities which we’ve fixed. We ended up with machines with lipstick on them. Girls kissing them, hugging them. Families driving four hours to go see a machine and take a family photo. I mean, just never ending and it’s been for the most part extremely positive and welcoming. The others at lovely media talking point were X, Y, Z technology is going to kill us and every 15 years we have the same ridiculous episode where when electricity first came out, obviously that’s the work of the devil and that’s going to get us killed. Then the first automobile went over some dirt road in Detroit and that thing is going to get us killed. Then the computer, the ATM, the internet, every 15 years-

Ali Tabibian:     [inaudible 00:24:09].

William Santana Li:       Everything is going to kill us. I would suggest that your listeners plot the employment levels from 1900 to 2018 and note the two dips in ’29 and 2008 and I can probably assure you that neither one had anything to do with technology. This will continue and to me I’m the counterpoint person like, “Guys and gals, let’s get real here. This technology is going to be profoundly helpful.” Let’s fast forward and say today crime has a trillion dollar negative economic impact on the United States of America. It’s a hidden tax and for some reason society says, “This is okay.” Crime levels go down by X at some city and everyone is like, “Yes, so we’re all good.” “Trillion dollars a year, you’re all good? I don’t believe the founders of our country ever expected us to build a society where going to school, going to the mall, going to work, going to the movies came with a risk of literally being shot or killed. This is not acceptable.”

Let’s fast forward and suspend reality for a moment and let’s say we’re right. Knightscope is going to be able to help our country be one of the safest or the safest country in the world and cut crime in half, let’s just say. Talk to me about the impact on housing prices. Talk to me about the impact on insurance rates, the volatility of financial markets, the viability of someone’s local business, the safety of your family like literally everything changes for everyone. When that technology is pumping out at volume, that technology will not only be priceless for society, it would be priceless for those investors because you’ve got a business model that actually can throw out a lot of cash flow, but is actually doing a lot of good for society while we’re at it. We can suspend all the ridiculous assertions of the machines are coming here to kill us or what have you. This technology allows people to be that much more effective.

Ali Tabibian:     Has there ever been any injury as a result of your machinery?

William Santana Li:       We are pretty transparent folks. We’ve had three incidents and I’ve assured our investors there will be a lot more. As you know, I’m ex-auto guy out of Detroit. You would agree with me that the auto industry is a mature industry, 100 plus years. How many accidents occur every day?

Ali Tabibian:     I don’t know the number, but the number of fatality is around 40,000 a year.

William Santana Li:       15,000 accidents occur a day in a very mature industry. Knightscope, we’re in uncharted territory with autonomous security robots operating out in the wild 24/7. We’ve had three minor incidents and people want to go focus on those. They’re going to be more incidents because no one has ever done this before. You’ve not had the opportunity to have an autonomous machine running on a Saturday at a mall is a completely different experience than four o’clock in the morning at a hospital in a parking structure. It’s very different than a manufacturing facility when the shift ends and more stuff is going to happen and all we can do is, “X happened. Did you fix it? Then let’s move on to the next bet.” In some cases, it’s just basic algorithm stuff. In one case, it’s semi-humorous put.

We had assumed two years ago that in our algorithms the earth would not move. Meaning, the floor would not move. You would think that would be like a sane maybe assumption. Then we go to this location where it’s all brick pavers, just bricks as a flooring and more than half of them are literally lose like you can go pick them up and take them out of their slot. A robot might misjudge something because the “earth moved.” Meaning, one of the bricks turn the machine 90 degrees or thinking it turned 90 degrees. I mean, that patch took us 48 hours to fix, but obviously assumption from two years ago that the earth wouldn’t move or is not valid. I’m sure this probably a high and full if not a lot more than that where things are going to continue, but we’re going to make the technology more and more robust and it’s really important that we continue to operate in different environments because, again, as I mentioned, each deployments very unique and very different.

Ali Tabibian:     Let’s talk a little bit about what’s changed because you talked about it over a couple of years. Obviously, there’s some improvements you’ve made as you’ve learned more.

William Santana Li:       We dropped new software every two weeks and a hardware every three to six months. There’s a lot of version releases and upgrades along the way.

Ali Tabibian:     For example, I didn’t hear you mention cameras or is there a time when cameras where in the units and aren’t no longer there. Are there changes in terms of the significant component technologies that have made sense to you over time?

William Santana Li:       No, not really. I mean, we’ve gotten more efficient on somethings we’ve had to bring in house to build that we’ve taken stuff out. The feature is still there, but we’ve taken stuff out. I’ll give you a good example. Detecting a person or detecting a … or reading a license plate or something like that. That’s a kind of known problem. It’s been solved primarily with fixed cameras. If you’re doing some analysis and you’re using a fixed camera, you’re doing pixel differentiation and just saying, “Okay, the foreground is not moving. The background is not moving. The camera is not moving. This object came in and out. Can you detect that?” Well, in our case, it’s a little bit more difficult. The camera is moving. The foreground is moving. The background is moving and the object moving. Now, tell me that’s that person or tell me that is a plate. We’ve had to bring stuff in-house to basically build stuff that is more attuned to what our clients really need as opposed to taking stuff out from what you’re asking.

Ali Tabibian:     Is that basically a LIDAR-based detection?

William Santana Li:       In some cases, we used multiple sensors to do something, to do ACCEL, let’s say. A good example, a human would never walk down a hallway only using their ears. Similarly, I’m not a big fan of you have X sensor and it’s going to tell you everything about the world and everyone should invest in that one thingy because that’s going to solve everything like this is beyond foolish to me. You really need to be thinking this through a little bit more holistically and looking at different sensor parameters what different things can be really effective in certain situations and perhaps combine that with other data to help confirm something. Very crudely, we don’t do this yet, but just to make the point. The cameras can detect that this is a person. It would be really nice to just run the thermal scan as well and it says 98 degrees. Now, I’m a little bit even more confident that that’s a person. That’s a good example.

Ali Tabibian:     Great. Great point. Thank you. Speaking of things that have evolved over time, I remember a few years ago when we were talking, it wasn’t clear what the rationale for adoption would be and there was a potential that the rationale was going to be primarily cost reduction, reduce the labor for security [inaudible 00:32:04]. That hasn’t really turned out to be the case at all in terms of being a prevalent reason for the adoption of your technology. Am I correct?

William Santana Li:       You are correct now. We’ll see. We are working about 100 contract right now. A significant majority are, “How do I add security? I have exposure. I have limited budget. How can I do this efficiently?” We have two or three very large players that are really looking at genuine cost reductions, but now they’ve also realized it’s not just the cost reduction is. These machines can do 10x what a human could ever possibly do. Try to actually do the comparison is a little odd.

Ali Tabibian:     Give us some examples of the differences.

William Santana Li:       One of our about to be client is 40,000 parking spaces. We can read 1200 license plates a minute, run it against the database. If we were working with law enforcement, we can run it against the NCIC database and tell you that vehicle is stolen. That’s a stolen plate. That vehicle is tied to a felon. We can tell you that vehicle has been parked in that exact parking location for the last 17 hours and 57 minutes and 10 seconds. We can blacklist. Let’s say you have a domestic dispute, the spouse keeps causing all kinds of problems. We can obviously flag that plate. There’s no way a human could ever possibly keep track of all that stuff, process it and then pinpoint exactly where you need to be looking.

We’re really focused on actionable intelligence. We don’t want to be spitting out 90 terabytes of data to a user and go, “Hey, go figure this out.” Let the machines do the monotonous, computationally heavy work and do it consistently 24/7 will show up no ifs and their butts. No sick days, no any that stuff, and then let the humans do the strategic work. “Okay. Well, I got this complaint. It looks like the spouse might be here. Let me track that person down and let me call the other guard and go do whatever they needed to do.” It’s almost apples and oranges discussion. Most human could never do what these machines do. These machines in a lot of cases will never be able to do what a human can do.

Ali Tabibian:     It really is very complementary. Just to be respectful of your time, let me maybe ask you a little bit about money. In other words, what do you charge [crosstalk 00:34:48]?

William Santana Li:       A financial guy asking me about money? Interview is over.

Ali Tabibian:     I’ve held myself back for quite a while. I couldn’t do it anymore. The medicine wear off. What do you charge for your service, list prices whatever you are comfortable with sharing, and tell us about the fairly unique way you’ve chosen to finance the business?

William Santana Li:       Where a business is going to end up scaling and ramping very nicely is when there’s the intersection of value for the client, a price point that’s attractive for the firm. They can generate value for both the client and the company. This technology is brand new. Back in Detroit, we used to sell a lot of hardware at prices that we can barely cover our cost of capital and I don’t want to repeat that movie. We offer a nice offers that’s technology and a machine as a service business model. If you want an armed law enforcement officer off duty, depending on what part of the country you live, plus and minus your 85 bucks an hour. If you want a security guard, again, plus or minus the regions because cost of living can be very different across the country, but you’re looking on average plus or minus 25 bucks an hour. Mind you it’s not the guard salary. It’s what the client is actually having to pay. You need to pay the guarding company or what have you. Let’s loosely say it’s 25 bucks an hour.

We offer this on a machine as a service business model. We sign one, two or three-year contracts depending on which machine, so we have several. It can range somewhere just north of $6 an hour to tops out of maybe fully loaded with all the features at about $12 an hour. You’ve got a significant cost reduction where you could possibly pay that guard even more money and get someone more highly skilled and more engaged. The security industry is about 100% to 400% employee turnover rates. The jobs are almost untenable because no human certainly wants to sit there at three o’clock in the morning twiddling thumbs where 98% of time there’s literally nothing going on where a machine could be a little bit more productive. That six or 12 bucks an hour, it can be very cost effective for the client. Now, how’s that attractive for us as Knightscope and our investors? Well, if you kind of loosely or run 24/7, so the contract is somewhere in the 70 to $98,000 a year.

Ali Tabibian:     Per unit?

William Santana Li:       Per machine. Today with no economies of scale, no efficiencies, no design efficiencies, la, la, la, plus or minus the bill of material, one machine is around 60 grand. Can you recover the variable cost of the machine in the first calendar year on way or another? About 30% of our clients pre-pay the contracts, which is nice from a cash flow standpoint. As an ex-auto guy, I want to make sure this last for five years in service or a few 100,000 miles in service. You recover the variable cost in the first calendar year. Then the second, third, fourth, fifth year, you’re basically printing money. You are meeting obviously do the upkeep and make sure all maintenance and operating stuff is working well, but you’ve funded an asset that now is generating some nice revenue stream downstream. You want to be very efficient on how you finance that because all of our clients are credit worthy folks. Most lenders will look at that and go, “Okay. Well, I don’t think X, Y, Z Fortune 500 company is going to renew on the contract. I have his asset here.”

Sure, I’ll lend you the 60 grand for 18 months or what have you at some de minimis interest level. Now, you can take the debt and fund the assets and then take our precious equity and use that for R&D and SG&A and growing the rest of the company, so we can be very efficient with the capital. To dates, we just closed our forth run funding so we’ve raised $40 million in equity.

Ali Tabibian:     Incredible.

William Santana Li:       I would love to challenge anyone to go up and down Silicon Valley and see who’s been the most efficient with their capital for the type of business that we’re building. We’ve been very mindful and careful with our investor’s money and we’re not the guys like, “Oh, well, there’s always going to be another huge equity around and just keep burning money till it runs out probably.” We’re trying to be very careful to be much more conservative and methodical about growing the business. I don’t like the hockey stick or step function business plans like, “Voila! Tomorrow, there’s going to be a thousand machines and we’re just going to burn through 500 million bucks and we’ll figure out later.” I’m not interested. This is already really difficult and we need to be very thoughtful on how we execute. Our big focus right now is scaling the company and making sure our clients are happy and continue to better the technology. We’ve got a lot of more stuff coming in the pipeline so we can actually help fix our client’s problems.

Ali Tabibian:     If I recall correctly, it was mainly a strategic capital and then smaller investors that have helped you come this far.

William Santana Li:       We have four corporate VCs and we’ve got now nearly 6,000 investors. We did the largest Reg A plus mini IPO in history. For those of you who don’t know, the federal government changed some rules that allow private companies to do a public offering but remained private so stocks are not trading anywhere. Every six months now we have the SCC filings and audits and all that good stuff. It’s like a stepping stone to a full public offering, so put some good disciplines and controls needed within the company, provides a good amount of transparency. One of the things I love is our investor base. Like we have folks that are chief security officers at major firms or vice presidents of leasing at major facilities or ex-Department of Homeland, ex-FBI, ex-NYPD folks. We’ve got all kinds of friends now that are literally invested in the company’s success nationwide. I think that’s a good stepping stone.

I’m probably one of the few and only folks that have raised $40 million of equity in Silicon Valley and I’m still the sole director, which is going to be an interesting and exciting opportunity for a founder to be able to architect the right board of directors for the right and appropriate level of governance. I’m starting to recruit the four independent directors that I want to make sure can stick with the company for 5, 10, 15 years; can have the right skills, be a customer facing in the security side of things, be a technologist probably on the data side of things, certainly someone that the SCC would view as a financial expert to be the charity audit committee, and then a successful founder who’s sat in this chair and has been through this hell. I’m sure I’m in the very single digit minority in this opinion, but in this particular case, I’ve got the authority then to pick a diverse board in thoughts, in gender, in race. There are actually no VCs on our cap table.

Ali Tabibian:     That’s interesting. Bill, you talked about how in the first year or so how difficult it was to convince anybody that any component of the business model would work and you’ve proved everybody wrong. What’s interesting is I don’t think anybody would have bet that you could pull off the financing approach you had.

William Santana Li:       We’ve always been pretty much oversubscribed. I think the last round we ended up, don’t quote me here, 123, 128 million post money. I got nearly 6,000 people put their money where the mouth is and bought in a three bucks of share, or the less folks came in at 3.50-ish share. I think the market set the price. One of the other reasons why I did this is let’s go back to where we started. What’s the long-term mission? If we want to make the United States of America the safest country in the world, change everything for everyone. This is going to take a long time. This is a 10, 20, 30-year commitment, a long-term capital that is focused on the mission and we’ve got a good running start to what we’re doing, so all the theory of having a physical deterrence or the theory that robotics and autonomous technology and sensors and data can actually help is no longer theory. It’s reality so now the thing is to find the next group of folks that want to make sure the company can scale and grow the company properly a bit differently than most startups have grown.

Ali Tabibian:     I would have never guess you would be able to pull off many of those and now you’ve pull off all of those. That’s pretty startling and that’s why you’re the entrepreneur and the rest [inaudible 00:44:17].

William Santana Li:       Listen, we’ve had a lot of help along the way, yourself included.

Ali Tabibian:     Thank you.

William Santana Li:       You’ve been very kind to invite us to some events, met some folks. We’ve got some awesome investors who take time a day to introduce us to new clients or recruits or other investors or “Hey, you should really talk to this person” or got a, I don’t want to say kitchen cabinet, but we’ve got a few investors that are so thoughtful who’ve been around the block who often reach out and like, “Hey, how are you doing? What about this? Did you think of this?” or “If it’s okay, I’m happy to take a look at that term sheet you just got.” We’re surrounded by some really passionate people that want to help.

As much as Silicon Valley likes to build up the CEO and she or he did the whole thing like I’ve got a kick ass team downstairs that are bunch of crazy engineers that are determined patriots who really want to get this done. We’ve got ex-marines downstairs. We’ve got ex-law enforcement, ex-navy, air force, army. We have some serious people here. We’re not playing around. I think that’s what most of the investors bet on like, “Do you believe in the technology? Do you believe there’s a market there and do you trust the management team to execute and God forbid something goes wrong they’re smart enough to go figure it out.” It’s a simple bet.

Ali Tabibian:     At this point of the conversation, we moved our discussion outdoors and spoke for a few minutes in the presence of a K5 unit that patrols Knightscope’s parking lot. We then had an insight for a tour at the network operating center, the manufacturing floor and a prototyping area. As we move from inside to outside, we walk by a stationary model, the K1, which startles me by saying, “Hello.” It’s not on the recording, but on my way out after finishing up, the driver who picked me up was startled when the parking lot K5 yelled, “Good afternoon” after it saw me leaving.

William Santana Li:       This need to be acceptable and open society, but second, it’s also a psychological deterrent. You’re trying to use all the senses, what you see, what you hear, what you see happening is going to deter a lot of negative behavior. We didn’t talk about this earlier, but you can also as a security guard you can use a machine as a mobile PA system. God forbid there’s a live shooter situation you can speak through a machine or all 50 machines at one time. You can test one [inaudible 00:47:03] machine and have a two-way intercom call. With the security operation center, the machine can also announce stuff either randomly or manually. Actually, in one case for the pre-recorded messages, we’ve had to put them in Spanish as well so the guards could announce stuff.

We can have, depending on the client, someone a male voice, someone a female voice depending on what they name the machine. We’re just getting started. The really hard part was getting the autonomous stuff to work consistently. I guess we talked earlier Saturday at a mall or a parking structure or a busy parking lot can be pretty challenging, and it might be easy to run a 30-minute demo, but you need to do that for 24/7 is a whole other game. As we add more and more detection capabilities, we’re also working on a concierge feature that allows human to have a two-way dialogue with the machine. That can also be helpful. It’s a little bit of a science project that we are working on, but the concept of doing a concealed weapon detection to be able to detect if someone is concealing a weapon in an area that they shouldn’t is also something that’s going to be very valuable for some of our clients.

We want these machines in the long-term to be able to see, feel, hear and smell and gather all that data and be able to figure out what is “normal” and abnormal in an environment and anything that’s abnormal, they better let the humans know so they can go do the enforcement aspect of it.

Ali Tabibian:     That’s pretty impressive. Let me ask you, has this unit mapped his area itself?

William Santana Li:       The machines mapped the area. A human would guide the machine around.

Ali Tabibian:     I see.

William Santana Li:       Then we might demarcate certain areas like, “K, don’t go in the lake. Don’t go in the street. Don’t go over this humongous pothole.” Other than that, it needs to dynamically plan and re-plan. Some of the environments we operate in, 18-wheeler truck will just show up and plot itself there or a mass of 50 people will come out at one time or we’re at a major stadium. Imagine when the event comes out, so you got to be able to do that and do it dynamically.

Ali Tabibian:     What are all the various openings and circular features?

William Santana Li:       The new feature we’re announcing today, it’s an ATM [crosstalk 00:49:34]. You can just stick your card here and we’ll take them. No, just kidding. The idea with the lights is also visual indication. The future model that we’re working on will have a little bit more lighting that’s similar nature to a car or a truck so that people know that a machine stopping. In some cases, we’ve got it setup where there’s a certain orange or red alert. The lights will actually turn on if the guard possibly looking at their phone or the like. They can be used for a variety aspects or when the machine might be speaking to you, it can also be an indicator.

Ali Tabibian:     It’s pretty impressive. Where there has been one of those I think you mentioned three incidents? What’s been the renewal rate for those clients?

William Santana Li:       For two of the three incidents, we’re looking either at coming back or having a different location. We’re always very transparent with what happens. They have a full understanding. I think one of the locations, the terrain wasn’t a good place to put it, so let’s put it somewhere else. It’s going to be interesting as we scale up, but I think one of the concerns from the earlier clients were, “Will this work and can you scale?” When we were just before 2017, we’re operating in California only. We are here at Knightscope headquarters in Mountain View, in the heart of Silicon Valley and the first time we ever put out a machine here at our own headquarters, we all pulled all-nighters, sleeping inside the building and doing shifts because we’ve never done this before.

I think that first few deployments were bloody painful. I think for four months, someone on the team or half the company pulled consecutive all-nighters for four months. We try to keep these machines running. That was a bloody painful experience. That was in 2015. Those were version 2.0 machines. We got smart, quickly released the version 3.0. That’s what we’ve been building at. This one that’s patrolling here is version 3.7 that has a lot of the upgrades and capabilities and improvements and navigation and stability and the like. At some point, we needed to go home. Then we got a few working up here in Northern California and then one day we’re like, “Hey, let’s put one in San Diego” and the team turn pale. “How are we going to do X?” “We have to pull a Band-Aid off or you’re never going to be able to figure out how to get this to work.”

Then we got brave and went outside of California. Now, we’re in Massachusetts, New Jersey, New York, Florida, Kentucky, Mississippi, Texas, Washington State. We’re just putting one in Colorado this week and you start learning. “Hey, someone forgot to make the change for the time zone.” It’s like basic stuff.

Ali Tabibian:     We’re in a parking lot here and the K5 is going up and down the path the car would use for egress. If a car pulls out, will it move out of the way?

William Santana Li:       That’s a good example building what I was just talking about experience. We have an early version of a car backup detection algorithm. Who would decide in the conference room and go, “Hey, we should like figure out how to analyze if a car is about to back out? How do we turn around or re-plan?” Unless you’re actually on the field, you wouldn’t know how to do that or what happens when a girl decides to hold on to the machine and just hug it for a selfie like what do we do? We have this funny experience back in 2015 where the algorithm was if you got in the way, it would try to go to the left of you. If you went to the left, it try to go to the right. If you go to the right, it go to the left. Imagine a 10-year-old kids going to have a field day with that. How do you fix that? All engineers pulling their hair out like, “How are we going to get over this problem?” and someone said, “Just stop.” “What do you mean just stop?”

Ali Tabibian:     Take away the attraction.

William Santana Li:       Literally stop the machine, turn off the sound, turn off the lights and just do nothing and sit there for three minutes and see what happens and voila!

Ali Tabibian:     Somebody knew how to deal with kids.

William Santana Li:       When we’re talking about earlier about putting this in society, these machines have the same challenge. A police officer should stand up straight, shiny shoes, command respect and authority, but shouldn’t scare a child or grandma. We have a little bit of that same challenge. It needs to be large enough and command enough presence that someone is going to think twice before they do something silly, but it can’t be an ominous thing. If you would have pulled up today and we had 10 of these machines running around here, all painted black with red lights and a very ominous sound and moving 10 times faster, we wouldn’t be having this discussion right now. We have to be very thoughtful on every radius, every surface, every curve, every fonts, every color that we pick so that you’d have the right psychological view of what you pull in the market line you’ve never seen one before. I don’t know how many accidents we almost had here on our street where the UPS guy is the first time they drove by and they keep looking and they almost crash. He said, “What is that?”

Then upfront here we’ve got the K1 so this is a stationary unit that can do all the detections and stuff that we’re talking about earlier, but is intended for ingress and egress locations and then the future we hoped to put our concealed weapon detection capability on this machine. A soft target like the luggage area and airport or actually an emergency room in a hospital can be very, very problematic in difficult neighborhoods. This is our first-

K-1 Robot:        Hi.

William Santana Li:       First production unit and, yes, it talks. Sorry to startle you.

Ali Tabibian:     At least he was friendly.

William Santana Li:       This is if you recall the Knightscope’s security operation center. This is the user interface. Those were all the deployments across the country. That’s all live and then here you can route the machine in different patrol schedule depending on the time if it’s weekend or Tuesday or what have you. The guards can figure out what to do. Long-term, from an AI machine learning standpoint, we want the machines to create their own schedules based in all the data that they’ve learned over the last few years like, “Okay, actually, I think on Friday at five o’clock I should go hangout here instead of doing what you told me, which is go hangout over there where there’s nothing going on.” We can also run a scan of any device in the area so let’s say you’re at a data center where someone is trying to do a cyber attack from across the street. We can detect that new router and give you the device ID, how long it’s been sitting there, when it was last seen, how far away it is from the machine.

Then you can blacklist that so that device ever shows up again, you can send an alert to the guard. If you want to run a thermal scan, so for most of our clients we set this at 400 degrees. If there’s a high risk of a fire, we can send you an alert. In a couple of cases already we’ve been able to avoid a mini disaster for a couple of our clients where we were able to detect something that obviously shouldn’t be that hot, i.e., a pipe was about to blow up or someone left a curly iron in a kiosk at a mall and would burn out a good portion of the place. No human would have been able to do that. As we are talking about, we can do parking enforcement, so we can give you the utilization of your parking lot. I’m sure 10 years from now, some hedge fund is going to be asking for this curve.

Ali Tabibian:     It’s like a [inaudible 00:57:58] Christmas parking lot information.

William Santana Li:       Exactly. What’s the parking utilization at this particular facility or we can run a parking meter. This will give you the top 10 plates that have been sitting in a certain location for a certain period of time, so this vehicle here has been sitting there for seven hours 44 minutes and 35 seconds. We can set that no one is supposed to leave a vehicle more than 24 hours or if it’s a loading area, it’s four hours or this is a handicapped spot or this is an electric vehicle. Why they’ve been charging for 27 hours? Somebody go do something with that or you can white list everything. Let’s say at a school, you could white list all the mobile devices. You could white list all the license plates for the faculty, parents and students. Anything out of the norm that shouldn’t be in this “sterile environment” should get flagged.

We recently announced that do April 20th for all the students out there, we’re going to donate half a million dollars’ worth of services. We’ll put three machines there, K1, K3, and K5 at school we end up selecting that sends us a compelling asset as to how an autonomous security robot could be utilized in an educational institution. Part of that is we want to do a good deed and help. Two is for us to learn. Three, hopefully, inspire some students. I mean, what a cool way to get STEM education promoted where they can actually learn a lot through the process. I think a lot of good will come from that. We’ve got a lot of initial interest thus far. I’m super anxious to read all these essays that are coming in. Then over here, we want to go this way, so that’s the KSOC, the Knightscope Security Operations Center for the clients to utilize a technology.

We started building tools internally here for our network operation center. You probably know how data center works. We’ve got servers in the massive warehouse basically and you’ve got human, software and hardware trying to monitor what’s going on with everything and making sure down to the millisecond that that server didn’t go down, something overheat, something didn’t die, whatever it is. One of our operations specialist here that run this here 24/7 can login to any of the machines nationwide, reset stuffs and software patches, do software upgrades, re-localize the machine, change a path, do something special for a client. If there’s an issue with the hardware, they can actually send out a technician. They’re on-call 24/7 to go make a repair out in the field if needed. Because we offer this machine as a service business model, the hardware, the software, the autonomous charging system, all the software upgrades, firmware upgrades, at times hardware upgrades, any maintenance support is all [inaudible 01:01:06], it’s all included.

Ali Tabibian:     Trouble free for the client.

William Santana Li:       Try to make this as easy and simple as possible. Here’s the slightly louder production area, but we design, we engineer, we build, we deploy and we support 85% US content. We physically built things here. As you can see, a lot of the machines are here to get upgraded or new machines getting built for the next batch that are going to be going out in the field. It’s flattering as an entrepreneur for massive major corporation to ask very politely, “Is it okay if we put our brand on your machine?” And somebody get super intricate. There’s one back there it’s going to a casino and they had a little bit of a field day with how they want their machine to look.

Ali Tabibian:     How long before charging on average for the K5?

William Santana Li:       Think of this the opposite of your electric car.

Ali Tabibian:     Between charges, I should say.

William Santana Li:       If we ran like your electric car, you drive all day and then you’re going to sit in the garage all night trying to charge, which is not how our clients want to utilize a security technology like they need 24/7. The machines ran for two and a half to three hours and then on their own autonomously go recharge and they’ll sit on the charge pad for 20 to 30 minutes. We ask our clients to put those pads in a prominent location like where security guard would normally be or in ingress/egress location because the machine is still operating, it’s just not moving. We take a coffee break let’s say every two and a half hours.

Ali Tabibian:     But still you can see things.

William Santana Li:       Where you can see things and were running 24/7.

Ali Tabibian:     Impressive. This was a more prototype fashion the last time I was here.

William Santana Li:       Our last stop here, this is the K7. Just a recap, we have a K1 that’s a stationary unit for ingress/egress locations. K3 that’s four foot tall, two foot wide, 340 pounds, primarily used for indoors. The K5, which is the most popular model, which is the five foot tall, three foot wide, 400 pounds, primarily for outdoors although some clients do use it indoors. Then the K7 here, which hasn’t shipped yet. We’re still in development. This is a four-wheel version of our technology for multi-terrain use. Think of dirt, gravel, sand, really large parameters, let’s say eight-mile parameter of an airport that today operates with no monitoring, no guard. There’s no anything and you can just jump over the fence and you’re on the tarmac or even something as simple as an avocado farmer where they pay the share 50 bucks an hour to patrol their farm because people are stealing their crops or solar farm or wind farm or power utility substation. One of our incoming client is 3,000 substations that they don’t want to pay a guard $28 an hour to stare the fence in the middle of the desert.

Ali Tabibian:     I’m impressed with you clearly take a lot of pride in the design. I mean, [inaudible 01:04:16]. It doesn’t look like you strap something together and just toss it on the road, aesthetic to it.

William Santana Li:       We try to be thoughtful with the design because these aren’t military products. They need to operate in society. They need to be consistent with a client’s brand. Thanks for coming over to Knightscope headquarters. You’re going to pretend you didn’t see any of that stuff.

Ali Tabibian:     That’s right. Bill, it was such a pleasure.

William Santana Li:       Absolutely.

Ali Tabibian:     It’s a pleasure every time I come here. Thank you.

William Santana Li:       All right. Thanks for having me.

K-5 Robot:        Good afternoon.

Voiceover:       Wasn’t this episode inspiring? Keep the inspiration coming, click subscribe. Subscribe with the little button in your podcast app or click the three dots and a little circle or visit us at gtkpartners.com.