First Time Founders with Ed Elson – This Company Uses AI To Help 911 Save Lives

24:54With AI, you can get to 100 % overnight. So not a very hard technical problem. It is what is being said. We look at the transcript of the calls. What ought to be said, which is the protocol. And did it happen, right? And if we look at that, now instantly 100 % of calls are QA'd. And we know, are they aligning with the city's protocols? And we can train those folks that aren't doing so great against them to do better next time and provide better service. I think that should be totally human out of the loop, done in the background. and it'll be less resources focused on it, we'll actually get better outcomes.

25:31So it's another automation opportunity. This hits at something that I think is so important about AI and that I think is applicable not just in this sector, but to basically every sector. And this is what I believe the true value of AI is, is that AI should exist to do all the things that are the mundane, menial tasks that you don't really care about. and I would go a step further to say what it really is, there are certain types of work and sometimes the things we do in everyday life where your heart is in it. We say that phrase, your heart's in it, but I think it's important because that's what makes it human.

26:10When you have someone who is suffering from a potentially life-threatening experience and you're their point of contact, your heart is in that interaction. You care deeply as a human what happens. And I find it so interesting that what you're finding is that actually we need to put the humans in those situations. And then we need to bring in the AI for all the other situations, all of the dumb, menial, rote bullshit where our hearts are not in it. And I think that's something that Mark Zuckerberg, as an example, doesn't understand when he says that we need AI to replace friends. Friendship is an example where your heart is in it.

26:51And what you need to be a human. I think what you're showing is, yeah, AI, it's not going to be good at trying to coach another human through a life-threatened experience. A human should do that. But what the AI is very good at is talking through a parking ticket. I totally agree with that. So, like, my potentially hot AI take is that the kind of general heuristics that we have used to apply technology in our lives today won't be that different in an AI future. Like, so that doesn't mean that there is an infinite amount of opportunity now and that this isn't the best time to be a builder because you can think about things first principally.

27:30The technology is fundamentally different than anything else. But I think the categories of things we will apply the technology to are very, very similar. And so I use silly examples on purpose just to, like, get your brain to think about it. It's like, I think it would be two buckets. I think it is things that humans ought not to do or should not do. Uh, and that is like my silly example there is like, I probably could clean my clothes, but I'm going to use like, yeah, I'm going to use a washer and dryer. It makes a lot more sense for technology roughly to go and do that thing. And there's this class of things that are just mundane that don't make sense for me to do.

28:03And I can free up my time to focus on the things I'm probably really good at. And then there is the second category of, um, use cases of technology that do fill in gaps that humans just could never do. And so, like, my silly example there is I'll never be able to run 60 miles an hour. I wish I could. You know, I, like, would love to be in the Olympics. I'd win all the gold medals. But I won't. And so I use a car. And in this scenario, me and a machine are just better together than they would have been apart. And for us, that's language translation, right? Like, I go to cities across the country, and they have this discussion of, like, where does AI make sense?

28:39Worry about, okay, if we apply AI, is that threatening to me? and I'll say, okay, quickly, show of hands, how many of us speak 30 languages? And I have not gone to a room where someone has put their hand up because we just can't do it. And technology makes a lot of sense there. And if we fill in those gaps, again, we're gonna be better at performing the core job. And so I think the AI opportunities of today are those two buckets. It is where can we address mundane things that I should not be doing, right? And some of those things, those mundane things were not technically solvable. You know, there was not low-latent voice and high-performing voice that could do those jobs.

29:17That's really exciting. But there's also a ton of things we just couldn't do, even if we wanted to. And I think AI makes a lot of sense there as well. It's also important that the buckets are sort of being reflected in your business, where you found that the really human important jobs, AI is there to be a co-pilot. And then the bullshit menial jobs, AI is there to replace the human. Those are the two categories that we're working with here. um i'd like to hear more from you about the business model what is the business model how does this make money so we um we have a couple different products so uh you can go and if you know like your city has a specific challenge like you know we have a high density of non-emergency calls you can just get that agent from us and um use it and it's a function of volume so call volume, right?

30:07How many calls are going through the system? Is it managing? And that right size, the large city is going to have larger volume and pay more versus a small city who has lower volume and pay less. We have other products that sit next to the call taker and help them do the job better. So things like language translation. So now if we're talking about the roughly 30, 40 % of calls that are emergent, again, it fluctuates city by city. We sit next to that person. Every single call that comes in, we would transcribe it. If it's non-English, we translate it. And then we create this instant summary of the interaction.

30:45And that summary takes the load off people. They don't have to take notes of every single call and try to catch every single detail. And that's what gets sent to the responder. And so that's based on how many emergency calls that we would be processing. And so we try to understand how many calls the system would process. And then we base the pricing off that it's almost like a enterprise contract like where it's usually per seat but this would be sort of per call exactly right you've seen this company grow a lot um from you and your co-founders i know you were playing football at yale and then you got the teal fellowship you dropped out um you've grown this thing from you've you were knocking on doors to now a company of over 100 employees.

31:31What has that experience been like? And what have you learned about being an entrepreneur and also a manager of people? I always start by saying it's good to acknowledge, like, I certainly did not come out of a dorm room ready to build the company. So you just had to learn really, really fast along the way. I think there's three distinct phases that a founder has to go through. And like, I'm in the middle of them. So I'm still trying to learn them as we go. But like, the first is you are doing the thing, right? And that thing is building the product, talking to customers, selling the product. There's nobody else but you.

32:06All of those disciplines are crafts you gotta get good at. Like sales is hard. It's especially hard in government. There's a lot of oops and things to jump through. Building a product is hard. Understanding users, what do they really need, right? And what are the biggest problems and how to sequence those things? Like you are doing that task. very quickly as you grow, it becomes, can I manage people who are very good at that discipline to do the thing, right? And so you have small teams, you may have a couple reps that are going out and talking to customers and are they going to say it exactly like you?

32:39No, but can you like talk about this thing in a very, very, or in a consolidated way or in a way that's cohesive and makes sense? So management. And then I think like we're starting to go into this filter now where you have middle management. It's really about culture and what is the operating system, the set of, again, values or things that you all share that are going to be how you make decisions, the framework of how you make decisions. And so all three of those very different muscles. And so it sucks, especially when you're growing fast, because you're like, oh, I just got good at that thing.

33:13You know, I was so good at sales. And it's like, oh, right now I got to be a manager. I got to learn that thing. You start to get good at that. You're like, oh, that's useful. But now Now it's like multiple people got to be managing. Yeah. Am I a small company manager? Am I a large company manager now? Yeah. Yeah, exactly. And so those are the things that we're, you know, that I'm thinking a lot about. And so today as we were getting close to a hundred people, it is about what are the frameworks that we set up where we have a very, you know, diverse set of folks, diversity where they live. They're across the country and they all have different backgrounds.

33:45What do we share? Like one, like what are the goals of the company? What are we trying to do? And is that clear? And then secondly, what are the set of like cultural norms that we're all going to say, yes, we're going to hold the line. And these are going to be the things that we stand on and do. And how do we, you know, meetings that I'm not in make the types of decisions that are going to continue to move us forward in a similar way that we've seen over the last few years. We'll be right back.

34:18Rinse takes your laundry and hand delivers it to your door. Expertly cleaned and folded. So you can take the time, all that time once spent on folding and sorting, watching and waiting, to finally pursue a whole new version of you. Like birdhouse building you. Yeah, the birds are going to love this. Or dollhouse building you. Yeah, the kids are going to love this. Or maybe house DJ track building to the bass drop you. Oh yeah, the kids are going to love this.

35:01Do you love it? Oh, that's a good you.

35:11it's up to you. We'll take the laundry. Rinse. It's time to be great.

35:21We're back with First Time Founders. What is the investor pitch? The big question that I would have if I were an investor is, how much money can you get in revenue from these underfunded local government call centers. So like, what's the pitch in terms of shareholder value? Like how valuable can a company like this actually be? This company that primarily works for local under-resourced governments. When big tech inflections happen, often startups or companies undershoot the potential. Don't overshoot. And so you think of a company like Figma. When the internet, now this happened over a long-term time horizon, But when the internet first came out, people would not have thought that the browser was going to be so good that you could build a$20 billion company, that people are going to collaborate in real time.

36:15It was provocative and controversial for Bill Gates to say there's going to be a computer on every desk when the internet came out. So that was the kind of provocative, controversial statement at the time. Nobody was thinking Figma, right? And so it's important to step back and say when these inflections happen, not only are they rare, not only are they big, but often people undershoot what ultimately they'll become. And so I think AI is similar in that it's an iceberg and like 10 % above the surface and like 90 % is still not. And there's some like general rules or things, common ground people are starting to find around opportunities like voice could be big and underlying models would be interesting.

36:52But like even that, not fully known. Is it going to be open source? Is it not? Like, that stuff is still to be determined. And so we are at one of those times, right, where everything in an industry not only can but ought to be rethought, first principally. Why do we use the technology that we use? Why? What jobs should a person do versus not do? And so, like, that's a macro trend. When you take that lens to emergency services, a ton of opportunity. I think they're along these two buckets. One is you have this fragmented set of tools, right? Today, you have eight or 10 screens. I think in an AI future, those systems converge and that they come together.

37:34And that, you know, not only will that be a convenience thing, yes, it's inconvenient that I have 10 screens as an operator, that I can't think of a better word, but I think if it does not change, it would be a dangerous thing. You know, you think about what is AI really good at? it's really good at taking a lot of valuable information, aggregating it together, pulling insights quickly, and then, okay, now we can go and use that. If I have 10 different systems in 9-1-1 that I'm using, and all of them are from different vendors and the data is siloed, you're never going to be able to capture all of those data sources and get the intelligence from it.

38:12And so to just give you some of the rough systems, it's like today, you have a call come in and that 911 call is in one system. The data we punch in out to the responder in another system. The radio is happening in another system. The protocols are in another systems. What you would really want is an intelligent system that knows all those things happening in real time and can suggest what responders should we send based on a bunch of different factors. Isn't that a hardware product though? I mean, the thing that comes to mind is what Sam Altman and Johnny Ive are secretly building, which is one product, a piece of hardware, replaces the screen.

38:45It's not going to be glasses that you put on, but it's something that just consolidates all of the data, whether it be audio or text or video. Isn't that what you're kind of describing? I don't think so. I think it'll be still software. I think that the hardware will be the computer, the desktop, and that it'll more be that all of these disparate systems come together and that the AI layer on top of it will now have all of the data intelligence. So I'll give you an example. Maybe we'll bring it in front. So like, if you have an emergency today, right? Call comes in and, you know, like we find out that there's weapons unseen and that a person unseen has autism.

39:26This is a real example of this app. So person unseen has autism. You know, we're trying to make a decision on who we should send. Here's a couple of failure modes. One is because that is a small detail in the broader thing, there's weapons unseen. We could just miss that there's somebody on scene with autism. That could not even find its way into the notes out to a responder. We catch that, right? We put that in. If you have a transcript of the entire call, the AI would not miss that, right? So that's one thing. But then the second thing is now we have to make a decision. Who is the best person to send to this situation, right?

39:57Let's say there's one responder two minutes away and there's one responder one minute away, right? Today, because that's such a fast-moving event and emergency, we are going to send the person one minute away. You can't take in account that many factors other than proximity. right? But if I knew the records of that, both of those responders, and the guy two minutes away, I saw in his report that he has an autistic kid at home. This is his life. He's done this every day. Yeah, he knows exactly how to handle this. And the other guy, one minute away, it's his first day on the job. He's never responded to emergency ever, let alone something as complicated as this.

40:29Who would you send, Ed? Okay, you know all this information. You might send the guy two minutes away, right? AI can look at all of that data, can search through the last 12 months of calls and what this person has sent and provide, hey, maybe we should send this guy. Now the human will still make the decision, okay, I have all the information in front of me, who am I going to send? But AI can go and hunt all that information and pop it up. I think you will have that single intelligence system that takes into account all of that data and helps us make more informed decisions. I totally agree. However, the AI must receive access to that data.

41:03And that's probably the other question that I need to address here, which is what you're describing is a world where we know a lot about the personal lives of citizens. And, you know, I'm sure China's crime rates are a lot lower, and I would bet their call centers are better than the US. But I think that's partly because they are down to put cameras everywhere and to have a profile on everyone and know all of the personal details on everyone. We are less willing to do that in America. and I would imagine that governments are especially less willing to do that when a private tech company backed by VCs comes in and says, hey, we need to get access to this data if you want to improve your product.

41:47To be clear, I'm not obsessed with the data issue, but I know a lot of people are. So what would be your response to them, to people who are concerned that if we go your route, we're sort of going into a big brother type situation? I think you should be concerned about the data route. It's a thing that we think a lot about. Here's how we approach it, right? So I think, and I can give you practical examples, is on something like video, right? Like, let's say we're going to capture a video. You are, a number of things are happening. One, you have a call taker who's deciding to send that out. They're saying, hey, I want to send this out.

42:25There's a person who gets a link, the caller, and says, yes, I want to turn on my video. and they have multiple permissions where they can say yes to doing that. And so only if they say yes, are they doing that? So any data collection from the caller is happening in a permission-based way. Now, on our side, right, when we are transcribing or introducing translation, etc., an important thing to remember is this person called 911, right? So they're calling 911 on... So it's not like we are just putting up cameras And as they're walking down the street, we are capturing that interaction. This is somebody who's saying, hey, I need help.

43:03Help me. So that data already, the transcript of that call, in every city in the U.S. is being stored. It's actually a mandate nationally that you have to store the calls. And I think it's actually a good thing. Citizens can FOIA request them and say, I want to review the call. And they come up in court all the time. And so that data is already being stored. I would say we are just taking information that's already being stored and now using it for the full potential of serving. citizens, right? So there's no new data, to put it, shortly, unless a caller is saying, yes, absolutely, I want to share this information.

43:36We're taking existing data sources that are being stored already when somebody calls 911 and making them useful in real time. This has been very informative. I'm just going to begin to wrap us up here. You are using AI to transform 911, and I'm in. I don't know if the audience is in, but I certainly am. What are some other government, public sectors, public services that need an AI makeover? I'm most up to speed on the ones adjacent to us. I think the first is just quality of life issues. You're seeing New York talk a lot about this in that how can law enforcement not just be reactive to crime and, you know, large-scale emergencies?

44:26How can we get more proactive about the things that most people face every single day? These are services like 3-1-1. What is that? I don't know what that is. Yeah, 3-1-1 is, so you call 9-1-1 for an emergency. It doesn't always work that way, but you should call 9-1-1 for an emergency. You call 3-1-1 for things like a pothole, right? Or, you know, like I have a question about a parking ticket I got, So that 3-1-1 is normally the line that you call for that. In most cities, because they're already having trouble staffing 9-1-1, those lines just are pretty poor. They face really long hold times.

45:02It's very hard to get an operator. So a place like LA right now, if you call the non-rudency line, you'll have a 47-minute wait. You're just going to be on hold. So most times it doesn't get answered. And I think that's the storefront of our cities. You know, I think of like an Amazon public company that does really, really well, or, you know, one of the most valuable companies ever. Bezos was calling the support line because he wanted to know what the service was like. And again, that's the service line of our cities. And I think we can do a lot better. How do we promote more innovation? Because I agree, like, I would love if I were clued into local government and if I were interacting with it at a product level on a semi-consistent basis and was happy with the product.

45:46How do we just promote more innovation? Is this an issue with VCs that they're too obsessed with private sector products? Like, why is there so little innovation here? I think there's a lot of variables there too. I think you hit on one. Part of it is funding, and we're doing the hard work to try to convince people to say, let's put it in new places. And we actually think there's a ton of opportunity because a lot of stones have gone unturned from a technology perspective in spaces like emergency services. So it is a funding issue. I think that's changing, which is really, really exciting. I think the second thing is it's a talent issue.

46:23It is not a lot of folks with our background are jumping into public safety. I remember when we first started the company, a lot of people told me I should not do it. And it was not obvious that it was gonna be successful. When I first did the Teal Fellowship, it was like, okay, I'm gonna live in a New Haven apartment. I'm gonna talk to public safety, who's really slow at moving and like, that's going to, going to work. And, uh, you know, a lot of people were like, why are you doing that? That's never going to make it. So like, there's a lot of scrutiny around it. And what my counter to that is like at, you know, startups going to be hard regardless, maybe this is harder.

46:55Um, when you have an inflection technology, you likely are going to do your life's best work. Uh, you can make the most impact ever now. You should be very deliberate in the decision you make and the things that you work on. You know, like Uber Eats is cool, but if you get somebody their food faster, is that really important? Whereas here we have the opportunity to save lives and you should do it. And so I think it's a talent issue and we're doing a lot of work to try to convince the best people to do it. And then I also think the space, like we talked about for, yeah, I think logical, thoughtful reasons is less receptive to new things.

47:28But if you get companies that really do it well, that take it very, very seriously, that are going to build things over a long-term time horizon, we're going to change hearts and minds. And they're going to see the new companies as credible. But that all reinforced itself. You get the funding, you get the talent, then you're going to become off, you're going to come off more credible. But I'm confident we can fix it, man. And I think it's very important that we do. But I think those three things will really move it. Finally, any advice that you would give for any young entrepreneurs or wannabe entrepreneurs who are listening to this podcast, maybe people who want to follow in your footsteps, what would your advice be to young people listening to this?

48:10We are at a very, very important inflection. And it doesn't happen often. And, you know, regret happens retrospectively. You don't often see it in the moment. I would nudge people to say, in 10 years, 20 years, name your time horizon, what do I wish I did today? And do that thing. Where that's led me is to make sure that, Okay, if now every industry can be thought about first principally, we talked about that earlier, what industry actually matters if we rethink it? And then make sure that every ounce of your work is going towards something you actually care about it changing. For me, that is lifesaving work.

48:47I get excited and I'm inspired by the fact that every line of code we write can be led up to somebody getting better help. And so I would just nudge any founder or early employee to be very, very thoughtful of that decision, especially in the time that they're in. Work on something that matters, man. And it might be harder. It might be more arduous. But it'll matter a lot more. And you will not regret it. Make sure you spend time on things that really matter. Michael Chime is the CEO and co-founder of Prepared, work on something that matters. I love that as a closing line. Michael, thank you very much for joining us on the show.

49:22Thanks, Ed. Appreciate it.

49:27Our producer is Claire Miller. Our associate producer is Alison Weiss. And our engineer is Benjamin Spencer. Thank you for listening to First Time Founders from the Vox Media Podcast Network. Tune in tomorrow for Prof G Markets.

49:48Thank you.

50:12Show me all the things PDFs can do. Do your work with ease and speed. PDF spaces is all you need. Do hours of research in an instant. With key insights from an AI assistant. Take a template with a click. Now your Prezo looks super slick. Close that deal, yeah you won. Do that, doing that, did that, done. Now you can do that, do that, with Acrobat. Now you can do that, do that with the all new Acrobat. It's time to do your best work with the all new Adobe Acrobat Studio.