Axon launched new cloud and artificial intelligence services as part of its hardware, software, and cloud roadmap and we caught up with CEO Rick Smith in a wide-ranging interview. Here's the transcript:
Larry Dignan: Let's just start at a high level. What the role of your company and the technology when it comes to police reform and some of the unrest that's been going on?
Rick Smith: Well, basically the way we think about it, police reform means we want to change the way policing operates and there are certainly political aspects to that, but there's also going to be very important technology aspects. For example, body cameras are one of the bigger police reforms of the past decade that now that officers are wearing cameras, the public can see very clearly what they do during these incidents.
And for example, I don't think we'd be seeing quite the energy around police reform from the George Floyd incident if there weren't body cameras there but because that incident was captured in such graphic clarity, that even the leaders of law enforcement around the country have come out. They saw enough to know that what happened was wrong. And they're saying, "Look, we've got to change." I think that's actually a great example of the role technology can play in enabling changes in policing.
Larry Dignan: And you've launched a lot of cloud services with evidence.com and taking that data from the body cams, putting in the cloud for analysis and all that. What do you think the role is of data in all this?
Rick Smith: Well, if you want to run an effective organization of any type in today's world, you need good data to do it, right? That's how decisions are made. That's how organizations manage towards goals. That's how they actually drive and measure change is by having good data systems to do it. So ultimately, I think that's one of the things that may not be obvious to the general public. When you think about police reform, there are the general concepts that, "Hey, we want police to be more accountable and find less dangerous ways to subdue people who might be resisting policing, for example.
But what that really boils down to is, "Okay, even if we all agree on, on those goals, how do we do it?" And that comes down to how are we going to train police officers? How are we going to measure what they're doing in a way that actually shows we're making progress? And how do we measure the poor performers in such a way that we can get them either remedial training or get them into a new career? And all of those business processes are going to require technology to play a big role.
Just give one quick example. When you think about the way we train police officers today, a lot of that's happening effectively in classrooms with an instructor doing PowerPoint. And it's pretty well understood that that style of instruction is nowhere near as effective as some of the newer technologies using virtual reality, where you can immerse somebody into a very intensive training environment. And I've seen studies that show that retention rates for people learning new skills are way higher in VR than in the old classroom. So that's just one area where we think approaching the way police are trained will have a huge technology element.
Larry Dignan: What's your footprint in the training area now? I know you're launching a virtual reality training system, to walk you through de-escalation and things like that.
Rick Smith: Well, actually, because of our history, we probably have the biggest footprint of any private company in training and law enforcement because we make the taser weapon that is used by well over 90% of US police departments. And when they use the taser weapon, they have to go through taser training. And so we provide all of that training material, content, we train about 10,000 instructors a year in the use of the taser. So we're very much already in that workflow. And every time we roll out a new taser weapon, we're rolling out new training, working with the national policy organizations, like the international association of chiefs of police. So we don't set policy for government agencies, private companies don't do that, but we work with them. And as those government agencies or these thought leadership groups, these NGOs help set those policies, we're the ones who translate that into actually building the tech that powers how officers are trained to those policies.
Larry Dignan: So you mentioned the data point and I know one of the 911 dispatch products is aimed to break down data silos as calls come in. Can you walk me through how that would work in practice?
Rick Smith: Yeah. So, as we sit here today across the country, if you get in trouble, you dial 911. 911 then goes to a call center somewhere in your County, most likely. One person will answer the call, they will then, while they're interacting with you, they'll start typing up information into what's called a computer-aided dispatch system. It's basically a computer program that allows that call taker to start entering notes into the system. Then another person called a dispatcher will be assigned to that call. It's called a call for service.
And now that person is in the same room, but they're actually not talking to each other because the call taker is very busy, frequently dealing with some hysterical person who might be having a very difficult crisis. So they need to focus on the person who's dialing 911. The dispatcher now starts reading what's going into that written record and they now start making decisions about, "Okay, do I send a police officer to this call or not?" And if I do, then they look in that same dispatch software to see where all the police cars are located because all the police cars have a vehicle locator on it.
They will then pick officers that are close by assign them to the call. And they'll then start doing research in the agency's records to see what do we know about this location? Is there a gun in the house? Does the person who is the subject of this call have any prior criminal history? They'll start looking all that, and they'll start typing that up into this dispatch software. Now on the other end of it, you have a police officer who might be out in a patrol car and they will get an alert, ding, comes up on their laptop, and now they will start reading these messages while they're driving to this location.
Now, I have a teenage daughter who just turned 16 and of course as a parent, I'm busy training her, "Hey, you don't play with your phone. You don't text while you're driving." And yet officers basically have to be reading these very intense, fast-moving, text threads while they're driving to an incident. Now, if this sounds like a really clunky system, it is, and it can have catastrophic consequences. The shooting of Tamir Rice was basically a person called in and said, "Hey, there's a black man with a gun in a park. Oh, wait a minute. I think it's a boy. And I think the gun is a toy."
As that information went through this system. What got to the officer arriving on the scene was a black man with a gun. And that officer, at that moment, got out and within seconds ended up shooting and killing the boy. Now that's a pretty dramatic example, but it shows very clearly how when you have this process of human beings trying to translate information, you get this telephone game and if you get it wrong, you can have pretty catastrophic consequences.
So what we've done is we've now launched a new dispatch system. We actually won't even call it computer-aided dispatch because to even call an IT system, computer-aided anything in 2020, is crazy. Of course, we're using computers to do this. We're not doing pens and paper anymore. But what we're really doing is rethinking this whole system from the ground up.
Wait a minute, rather than tracking the vehicle, that officer is wearing a camera, that camera has location. And rather than relying on, you know, just text messages we're typing in, let's add some voice communications. Now they do have radios, but if you're on a police radio, that's not just the dispatcher and the officer. On the radio, you may have 30 different patrol cars sharing that channel. So we can now allow the dispatcher to tune in and watch a video, right from the scene of what's happening, rather than trying to decipher what's happening from a few cryptic calls over the radio.
So what we're effectively doing is we're re-imagining from the moment that call comes in, how information goes out to the officer, how that information from the officer gets live-streamed back to the decision-makers and we think there's just a huge opportunity to make the whole process more efficient and ultimately more effective.
Larry Dignan: And the idea was all this data sharing would be in real-time and then someone may be back at the office or dispatch center could make a call about what to do?
Rick Smith: Well, yeah, ultimately the officers on the scene are going to have to make the call, but the faster you cycle information back to them, the better. Certainly, in that case, I talked about earlier, if the next cycle of information had said, "Hey, it's a boy with a toy gun." If that had happened before the officer got out of the car, there very likely would have been a very different outcome. So again, that is a very intense example, but there are millions of maybe less intense examples, but where the rapid flow of accurate information is just supercritical to getting to the right outcomes and minimizing the risk of something escalating unnecessarily.
Larry Dignan: Do you have that dispatch system in beta somewhere or has it been rolled out?
Rick Smith: So we've launched with our first customer and I believe we're preparing to announce but I don't think we have yet, our PR team to fill you in there, but we believe shortly, we'll be announcing our second customer. And we're now rolling it out to general availability at our conference next week.
Larry Dignan: In your shareholder letter for your second-quarter results, you had mentioned that you're going to include community involvement with your product design and things like that. How will that work in practice?
Rick Smith: Yeah. So, luckily, we've been laying the groundwork for some of this over the past several years. A few years ago, we created this industry's first AI ethics board that is intended to bring in the voices of concern from the community, particularly around privacy rights and around the ethical use of AI and other technologies. So historically, right, he's a police technology vendor. I interact all the time with cops. And the reason we formed this AI ethics board was we were being pretty successful with body cameras, to the point where there were some people saying, "Well, Hey, wait a minute. Are they going to use these body cameras to do face recognition like they do in China?" And lots of other concerns.
And so we in thinking about that said, well, how do we respond to these concerns? Well, the first thing we need to do is we need to understand them better. So let's open up a channel of communication with civil liberties-oriented groups or groups like the policing project out of NYU that focuses on police accountability and oversight. So we formed that board and it's actually really turned out to be a really effective process for us in that, you think about it, under the way we did business before, we would work with police, develop new products, we'd then launch them and then we potentially get a lot of blowback from civil liberties groups or privacy groups saying, "Hey, did you think about this? Here's how this could be misused."
What's happening now is we're getting that input right at the beginning of our product development process when we can still change it. So we're getting ready now to launch, for example, a license plate reader system. Now license plate readers are of a significant area of concern for many privacy-oriented groups. Because we understood, we took the time on the front end to hear those concerns, we've been able to design many safeguards into our system that address most of those concerns. So when we launch our license plate reading system, many of those groups, I think, will actually be promoting, "Hey, we don't love license plate readers, but if you're going to do it, you should go with Axon because they built in good privacy controls that ameliorate many of their concerns."
And now what we're doing in the wake of George Floyd, was we're now extending that even further, where I'm now starting to hold round tables with groups that are not just privacy and technologist oriented groups, but groups representing black and brown communities that have concerns about over-policing or police brutality. We just held the first one of those round tables a couple of weeks ago. And we hired a new Vice President of Community Impact, Regina Holloway, who comes from the city of Chicago, where she was working on police reform.
And so it's just another way for us now to extend out and hear more voices early in our product development process, when we can still do something about it, besides argue our point of view after we've released it. And I think it's making us much more effective as an organization and it's positioning us well to be able to help with police reform.
Larry Dignan: How early in the process will the community be involved? Is this the case where it's still a PowerPoint or is it a prototype or, how do you get pulled in?
Rick Smith: Yeah, so what we did, for example, with license plate readers, we actually showed that very early, over a year before launch to our AI ethics board. And then they had the opportunity to ask questions of our product managers who were literally showing them the PRDs, the product requirement documents. And then that created these feedback loops, where we started changing the product requirements in order to meet some of the concerns about how the system would operate. So this is being done very early in the process.
Now, when we talked about community interaction, this is a newer initiative, but our intention is to have the same interaction where we're getting this input when we're in the design phase because as you know, it's much easier to make changes when you haven't yet really built the product or invested in tooling. It's easy to make changes upfront. So what we've learned is, it's better for us to understand concerns when we can still do something about it.
Larry Dignan: Right. So as far as product development goes, how has it been over the last few years? I covered your digital transformation. And now last quarter, I think, revenue was pretty much evenly split between software and sensors and hardware, like taser and body cams. How has that transition, how has it been? And what are some of the lessons learned for other CEOs?
Rick Smith: Well, number one, I would say it was hard. It was much harder than I thought when you're transitioning from a manufacturing company to one that does hardware and software. You do need to be prepared to rethink everything in your business. We had to set up a completely separate and parallel sales team. We're now starting to bring those teams back together. But as you could imagine, the way you sell nonlethal weapons is very different from selling IT solutions to CIOs, the way you staff things, the way you manage your business across the board, really significant changes.
I think we were able to do it partially because we're a founder-led business, meaning, look as a founder, I get a lot of deference from the board and from the organization to push through those challenges. But I would tell you about three or four years in, I was in a company meeting where the number one question at that company meeting was, "When are we going to shut down this video business?" Because we were losing money. The organization was struggling to figure out... We had never built software before and we were struggling early on. We were able to push through that. It was very painful. Many of the management team had to be turned over. It was a difficult process to go through.
Now, coming out the other end, there are a lot of benefits. We're very proud of what we've done and we're excited now that we do hardware and software, it's opening up a ton of opportunities for us. We're about to roll out what I would say is our first major AI service, we have a previous AI service that helps redact faces out of videos. But that I would say is used only on a small subset of videos that have to be released to the public, our new AI service, auto transcribed, will be creating hypermedia transcripts. Meaning transcripts that aren't just something you print out on paper, but where the transcript is linked deeply into the audio-video record. So that when you playback the audio and video, you have the text transcript right next to it.
This is something that could apply to every video in every incident and could become really the core of the police report going forward. And so we think this is the one that has probably the first significant revenue implications from our AI investments.
Larry Dignan: So in terms of revenue and the public sector, has the public sector made that transition to operating expenses and cloud subscription model relative to I have a budget that's approved for two years and go?
Rick Smith: Yeah, when we first started, man, it was hard because our customers were not used this subscription model. However, what we found, once we started selling body cameras and software on a subscription, then our customers really, once they got some experience, they opened up to it and then they started coming back to us and saying, "You know what? Can you just put tasers on my subscription too?" And the advantage, particularly for the public sector, is procurement processes are extremely painful and difficult and expensive. And they take a long time.
And that's one reason many agencies, their tech is just really old because if it takes you three years to get through a procurement process, by definition, you're at least three years out of date. With these new subscription models, we're able to do software updates. Every 30 days, we're giving new software capabilities and now we've bundled in the hardware as well. So every two and a half years, we replace a customer's camera. They don't have to go back to another procurement process. And once they've come to understand that, I would say now, probably 80, 90% of our public sector customers are now on subscription plans.
Larry Dignan: Is there anything I didn't ask that I should have? Any final points you want to make?
Rick Smith: I would just say the thing that gets really exciting about this business, when we're doing it hardware and software, they're both difficult to figure out, but being able to do things like design our cameras, because they're being warned by police on the side of the road. If you want to transcribe audio, it's really quite helpful. If you also can go all the way to the edge of the camera and design the way you're capturing that audio to account for the fact that there's going to be vehicles and a lot of noise driving by, and we need to think about how we design the camera to optimize so that downstream, we can use an AI engine to transcribe the text from the voices in that audio and having that broad reach across the whole platform is something that I think is giving us a real strong advantage going into the next decade that we're able to basically do things that would be impossible if you only were doing the hardware or only the software.
Larry Dignan: All right, thanks for joining us.
Rick Smith: This has been great. Have a great day.