Inside the $47M YC Startup Fixing the $380B US Commercial Insurance Industry

0:00 kind of scares me. I don't want to become the bottleneck of this company that I'm constantly figuring out what everything is going. That's not a good way to run a company. We can't ever make a billion dollars like that. That's never going to happen. >> For years, the obvious way to build an AI was to sell tools to an industry. But the best founders have evolved past that. They're using AI to become the business itself and taking the revenue that used to belong to the incumbents. The startup I'm visiting today is doing this in one of the oldest industries there is. We have to be fully in control of the funnel.

0:29 What's happening is that the funnel is in control of us. And there's a lot of smart people here. So, we can take control of the funnel. Until that thing gets out to market or until somebody bought it, it's in our purview. I don't love the outcome. I'll be honest with you. I have to make this company exist beyond me. >> Today, I'm going behind the scenes with Harper, an AI native commercial insurance startup in Silicon Valley. A traditional broker takes weeks to source quotes and find a business the best price on insurance cover. Harper uses AI to do it in as little as a day or two across more than 160 carriers.

1:03 >> Week is too long. I would say a day is a goal. >> They came through Y Combinator's winter 2025 batch and raised 47 million in a combined seed and series A led by Emergence Capital. >> Almost think of it as like when you boil water at some point the water starts spilling over. That's us every day. Insurance is a demand capture business. Literally every business in the country needs it. So demand was never the problem. The real question lies in understanding whether they can build the back end fast enough to capture it at venture scale.

1:36 So you may notice that it is super super early in the morning. It's not night time. This is morning light right here. And the reason we're up so early is because we're going behind the scenes with a company with a work ethic that we have not seen on this channel yet. We're going behind the scenes with Harper, an AI native insurance startup based here in San Francisco. They've raised a 47 million combined seed and series A. And these guys are on a complete different level. When I was on a call with them, you know, I was going to come in at about 8:00 a.m. That's usually when I like to come see startups. And they said, "Hey, we get in at like 5:15." And I was like, "Okay, I'm going to get in there with you."

2:19 >> The hats just keep showing up in different places. there's a hat here. There's a hat off from Char's desk. >> For a bit of background context, Dakota wears hats all the time apparently and just like leaves little trinkets around the office. >> We have another email going out today that'll just be like, "Hey, just a reminder like like we saw this and then if you'd like to get in touch with us and then the plan is to give those guys a call tomorrow." Did everybody see that? Does anybody have any questions about it?

2:43 >> I have a comment. >> Yeah. >> for contractors, I feel like for all of them, we can just reach out for Inland Marine and just let them know, hey, if you need your tools and equipment covered, we can get that handled for you. And then for the higher rev accounts, we can start talking about some professional liability. Maybe also looking into accounts with actual employees, specific payroll, maybe specific industries where we can look into their workers comp.

3:04 >> This is the end of the month, so typically this this should be our busiest week, right? A lot of people need an August 1st start date. So really push for that. Send emails, text messages, calls. We need to, you know, get people aware that if they want a start date for August 1st, they really should be paying today, tomorrow. >> Yeah. Yeah. Get ahead of me. >> Yeah. >> All right. I just stepped out of that morning kind of sales briefing again early in the morning. It's like 6:30 a.m. The team is on it. Like they are powering ahead and I'm like, "Wow."

3:36 Okay, I need like two coffees. All right, so now I'm looking at this thing and let's see, is the funnel getting leakier? I feel like maybe we got 14 construction contractors, nine ISC quotes, one same day bound, and seven applications built. What does this mean? So now I'm nervous. >> Let's get it going. And I heard there's some parsing issue now. >> So that one was like with Vinn's new changes, he makes the parsing like more like synchronous. So what happens is that it's happening like at the same time. So what I'm going to do is I'm going to refresh it and figure out why.

4:14 I think it's getting like there's like a lot of volume. So it's probably getting like clogged somewhere. So I'm I'm going to figure out what's going on there. >> Tell me what that means. >> So I'm manually having to type in the premium >> for each of the quotes >> when normally it would automatically have its own line item. Okay. >> So, I'm just adding a line, which isn't I mean, it doesn't really matter, but from a Let me just show you from an email perspective. If I'm sending a normal quote >> Mhm.

4:39 >> this wouldn't look like this. It would have the specific line of coverage. >> Oh, now it says fees. >> Now, it just says fees and taxes cuz it's not parsing properly. >> Okay. Basically, what's happening right now, just FYI, is when you send out the quote, they have to manually type everything in cuz whatever Finn did with the parsing is up. So, basically, it's like it just shows fees and taxes. So, you send somebody a $10,000 quote, it looks like you have $10,000 fees and taxes. So, I think Arian's on it, but just make please like, please be on top of that because that's really problematic for the conversion cuz it just looks like all you're doing is paying fees and taxes. Send the 830 product safe. Okay, cool. And then let's make sure we have our good a good sense of what our agenda for that is going to be so we can model what we want to see in the world, which is good agendas.

5:19 Okay, let's think. Are we getting more things that are like good and going straight through and quotable? When I was doing some of the board deck prep though, I was kind of like the first pass rate is one measure, but then it's kind of like the quote rate cuz sometimes people still will have questions regardless like just a simple question that we might be able to answer ourselves. I believe the right metric for this is like how many quotes are we getting back and how fast versus the first pass rate? Like the first pass rate is like a starting indication of like the bigger thing which is just like are we getting more quotes back and are they coming fast? If you're defining first pass rate as like our the first pass rate we're tracking, that's like how good the intake was done, right?

5:55 Yeah. So then I almost imagine like follow-ups and the number of follow-ups and then how fast we get the code back. Like a combination of those two sort of determine whether the app checker skill is actually >> Yeah, agree. We should take this part of the conversation cursor cursor cursor cursor cursor or we should have a little bit more interaction with AI and figure out what's a good answer for this there. I really want to explore this a little bit more because I want to make sure we're tracking to the right KPI. First pass rate alone I think is a bad one.

6:18 >> Okay, cool. So, we got that all done. >> You have breakfast? >> No, I don't have breakfast usually. >> Oh, wow. >> Just keep the energy levels up. I feel like breakfast takes it and sucks your energy away. >> Really? >> Yeah. It's weird. I I really enjoy just you know, a load of caffeine in the morning. >> Interesting. So, are you saying like espresso coffee or brew coffee? What's the What's the vibe? >> You know what? My wife's not going to like this, but energy drinks.

6:43 >> C4. >> I know. It's it's you know dangerous scandalous. >> I was just with Dakota in the meeting where he's just briefing before this meeting we're going into and he says says cursor cursor cursor. >> Yeah. >> What is that? >> Oh okay. So one of the things that we're doing in in the company is that we are we monitor everything. All our emails our phones text messages they all kind of flow into a central place because we want to have the knowledge building of the entire company not have tribal knowledge in different places. Now when we have granas, Dakota is saying cursor cursor cursor because his cursor literally will be continuing to process everything that he's saying and he's going to send him reports. So in many ways he's just telling his cursor that hey this by the way please listen tune into this thing >> so that you can give me an answer to this. Oh, cool. So, >> it's a first party again cursor is almost like a first party thing here where so couple things ever since we met over the last many weeks everyone has been using cursor right so we have both engineering operations we have started like sort of like getting ops to be able to really build out their own applications as well so kind of moving for engineers are moving towards more like let's build a software factory and so then the operators are going and building their own applications and by doing So they can move faster, right?

8:01 Because they already know their pain points very well. They know what's the problem. So they can go ahead and like build these applications out and they can unblock themselves, >> dude. All right, real quick. Let's try this. Let's Let's give this a go. >> All right. Cheers. >> Cheers. >> Holy That's really good. >> Really dangerous. It's like cordial. team. I tried my first C4 this morning, by the way.

8:32 >> How do you like it? That's a good moment. >> It's crack. It's crack. >> Crazy. >> This thing is like I'm already like shaking my hands like >> if you start feeling ants on your skin. >> Can you help me understand where we are with like launching more verticals on the voice AI piece of it? let's see if we can start there cuz I also saw something over the weekend where it's like you're able to do workers comp alongside these things and others. So maybe walk me through what the product has developed to over the weekend and then where we're what we're thinking for today.

8:58 >> Correct. Over the weekend we actually added three more verticals. Also each vertical is now we added retail, smoke shops and consultants. Okay. >> Each vertical is now able to take the general application as well as cyber workers compent property. So if a customer wants all four lines, we're able to do that on the spot for them. right now the plan is to keep testing them for the day with the real customers. I've tested them all night yesterday. The idea for this week is to finish up to 10 verticals.

9:24 >> There's no reason why we shouldn't take every call out there right now in any vertical we can do with voice AI taking it end to end because in theory it's like there's no capacity constraint and we need to just quickly figure out if it's working or not. We need a massive signal. So my view is that an should be sitting there all day with like 50 call like I mean in an ideal world 50 calls going at one time. I mean we should take every single thing we can get on and literally just make sure you asterct it.

9:46 It is an R&D budget. That is a true literal non-marketing R&D budget. It is literally get this guy as many voice the eye calls as possible so we can see the thing works. I would think about it on par with like the spin we have on LLM or something like that in terms of how how much it can grow. So that's a top priority. I know for sure we can bid a lot for retail. I know we can bid a lot for consultants and make sure that we get it done. Even if it's $400 or $500 per one that we get in the categories that he can now actually do on the voice AI, the number one priority that we have is make sure that we completely get it fully tested and and buy every single thing that you can.

10:19 >> Another feature was implemented over the weekend was if the customer gives us their phone number, we're able to capture all the information. So we don't have to actually because we used to waste six to seven minutes on just getting the company's name, the address, all that. So now it's all eliminated with the phone number. We're able to prefill the application with what we have in Big Brother. The other thing I would say is that like I need more you know I get nervous because you're also have doing sales consumes out today.

10:41 There's you're doing the sales name there's a lot going on but if you hear the actual thing that's the most important for us to scale the top of funnel I mean if the voice AI can do a full intake itself it's already able to do workers comp cyber and all this other stuff. I cannot really think of a place that should be more well resourced. It's literally able to do all the calls. So, in my view, I think that like the the the speed of getting out the next one is it's seemingly a little slow. It sounds like there's so many things being done, but this is a super critical focus area.

11:12 So, I would just say in general, I think resource reallocation a little bit. I would leave it to Car and Tashard to think a little bit more about what that looks like, but I would really reallocate some of the focus because this it sounds like is working quite well and we need to get as many of these things stood up as possible as fast as possible because that's going to be the way that we really decouple >> literally like it actually is the decoupling factor of people to like the number of things we can intake.

11:34 >> To add to that, our target is by next mid next week, we need to automate the entire intake through voice agents. That's the target we have. And regarding scaling, I've been working on maintenance voice agents until now to automate internal operations. Skating is the most important factor right now. >> Stepped out of that meeting. Wild. We are going through everything. Anyway, next up is a growth meeting, but I wanted to quickly touch base on what I'm seeing. They deployed voice agents in under a week. I think it was like a couple of days. They deployed them live and they're running them across different industries that they're actually doing calls on and intakes on.

12:06 2 days, 3 days to implement that across multiple industries. and you know label it as R&D which makes a lot of sense because they're still figuring out how this thing works and if it actually works properly which clearly it does straight away and so they figure that out in a week and then the next week they're doubling down and they're going to roll this out over all industries and to move that quickly like that is a complete evolution of the company because suddenly you're freeing up the sales team like a ridiculous amount.

12:37 It's just crazy, man. The speed is just incomprehensible until you're in it. Until you are here. Anyway, we're going back in for another meeting. My hands like shaking on C4. >> So, we paid $246 for this lead. There is no application, but it got instant quoted on what portal? Somebody open. So, you all going to have the parallel path. Somebody open a big brother to like hunt down what's going on. So, this got instant quoted >> just for 2K.

13:03 >> It was for 2K. So, Jessica presented a $2,000 quote to somebody and then what happened? >> She agreed to pay within an hour. >> Okay. >> Of the last phone call. >> And then what? What time is the last phone call? Be careful, Kieran, with your Z 20 27 T000000 cuz you can't really see what the time is. So, I can't really It's not really time. And now it's 9:47 and nothing happened. So, how are we tracking that in the system? Jessica was told that within 1 hour, somebody would pay her, right?

13:34 >> Mhm. >> We have this in our system as we have no application. Application built. We gave her an instant quote for $2,000. The person says she's going to pay an hour. So, I'm also a little bit nervous because you all don't really have control of what's happening. If you all think the prices are $2,000 and Harsh is finding that it's $600 and you are running the sales team all day long, every every hour and a half, we're just going to keep doing this. I'm going to go through every single one. I'm telling you, until you all get systems that answer all these questions I'm telling you, you will never have a stop to your leak of your funnel. You're always going to have a leaky funnel. We're always going to be sad at the end of the day talking about how we failed to keep up with what was going on. It'll never be good because until you get this right and you think like this, it's not going to happen. Next one.

14:09 >> Is this a normal morning for you? >> Yes. This is a business that's always going. It's fine. There's a lot going on. >> I like it. >> And so, how are you feeling now after these two minis you've just run in? It's all systems go. How you feel? >> I feel optimistic. I mean, I think when people know I feel good like it's like when you know a problem, you can fix it. I think like it's clear what the problems are. I made a clear system of delegation like you own this, you own this, you it took me a while to learn that. I think I used to I'm a person who like is obsessive about controlling everything myself. So another version of if this was 2 months ago, I probably would have taken all those outcomes and just gone and done all of it myself.

14:43 Then I'm super bottlenecked. I can't think about growth. I can't talk to Joanne about partnerships. I can't do anything. So I feel like I feel happy and excited and optimistic because I feel like we have dissected the problems. Now people have them, but I'm never I never let it go. Like I'm like, "Okay, I said all this stuff, but I always have this subconscious thought there. Something can still get up." So I'm always thinking like, "Okay, like now let me go monitor. Let me do some more other things, but like in an hour I need to be on top of it again to make sure it's going." Then when I see after two or three checks that it's still going, I'll be like, "Okay, this is going to end up in a good end of day." I can really shift my focus to board prep cuz right now I have to remember MY BOARD MEETING HOME AND STAY and I have my memo, but I've not sent it out and I have to get it out today.

15:20 >> Yeah. >> Yeah. Wild. Wild. >> >> I am very impressed with the level of hyperfocus from Dakota. He promised to us at the beginning of the day that the one goal was focusing on fixing the leaky funnel, but just the determination to stick to that one problem and try and fix it. Because you have to have that level of focus at a company of this size because things will always get pulled apart and put by the wayside and forgotten and like tickets get missed.

16:06 If I wanted to get commercial insurance from these guys, knowing that one of the founders is sitting there going through individual quotes and individual customers is so important and and it's cool to see that these guys have not lost sight of the reason they exist, which is the customer. Also, they move at a million miles a second. Like to the point of where like I'm trying to keep up with what's being said. >> What happened? >> Well, I I I need my charger. I was just speaking to the camera about how how fast this company moves and then I just get this running.

16:42 >> Yeah. Yeah. Yeah. >> It's actually so fun here. >> I love it, man. Are you new or are you you've been here a while? >> I've been here for like 3 months now. >> Hell yeah. >> Yeah. >> Super fun. >> You keep running. You keep running, >> dude. So, run me through It's all hands today. Like, it's happening. >> It's always all hands. >> Okay. I do think that we're going through a phase where, you know, we've been very founder mode, right? Yeah.

17:06 >> And so because we've been very founder mode, we're like, okay, like, you know, we're very hands- on deck into every single thing. And of course, the company has evolved and we've had a lot of people come in to, you know, lead different efforts. One of the things about me and Dakota is we're very detail oriented. >> Yeah. >> So, we really care about, you know, Dakota is really detail oriented when it comes to go to market, life cycle, the customer, service, basically everything.

17:28 Yeah. you know he's he's one of those unique people who holds a lot of context in his head and then I have to think about okay how do I like you know tango right it's all a tango and you know because in the sense what is going on right now and how do I best support the organization in the sense of like building out capabilities >> they can provide leverage so if somebody goes left I go right so whenever there's like an organizational shift that's happening it's very much like okay like how do I sort of bring my organization behind it and how do we have a lot of impact and force behind it >> future Right. You get your right. Kids are all right.

18:01 >> Yeah. Hello everyone. How's it going? >> Good. Good. Good. >> I see some things here where it's like Anas was testing the voice AI and the person just dropped off the voice AI. For me, the way I think about this is the highest level. There are 38 things that are still sitting there that we paid for in some way or another that we still need to figure out what's going on. And all I really it's like the same kind of probe to you all every day. am trying to understand is like give me the narrative of how we're going to recover all these 38 things.

18:28 Like I'm getting closer but kind of like not still getting a clear sense of what exactly is happening with these things and that kind of concerns me. So I'm just trying to figure out what's actually going on. >> Can I share you >> if I and also please if I'm being unfair I promise you please tell me I'm being unfair and please help me understand. I'm I'm genuinely just trying to understand cuz we get to the end of the day every day. We have some amount of pending applications. All I'm trying to do is make sure y'all know what's actually going on and I can have some hope that we have a path to recovery of some of this stuff. That's all I'm trying to do. Okay. Anything else? Kudos to Garrett. I think Garrett closed a pretty day.

19:07 We have a lot of work to do. So, I will start my night shift. but it was a great day and I look forward to seeing you guys tomorrow. What would you say to these founders out there that I want to do this, want to have this and and and properly go and change the world. I think the simple advice I tell myself every morning is get into the arena and fight like hell. There was so many moments me and Dakota's journey that we could have stepped away from the arena. You may not know if your product is making it. You may not know who your customers are. You may be trying to figure out who your co-founder is. Me and Dakota for religiously have done 18 to 19 hour days for the last many years.

19:50 >> Crazy. >> Just like it does, you know, there's no stopping. I do this every day because like I saw directly when the little guy gets hurt. Like what happens when this doesn't go right when it's not the happy path? And I genuinely do believe that this is a business that's AI for good. I don't think there's any substitute for hard work. I say like, okay, I'm going to my night shift. It's like my night shift is I'm observing all the things that happen throughout my day. I see a team that's in need of direction and some more guidance. I think they'll get there. I need to help them get there. Let me spend my evening across a number of things from product to what you experienced in the growth meeting. Let me make sure I come back in this evening that they see at 2 or 3:00 in the morning that final message for me. It's hey, here's how we're operating tomorrow so we can achieve what we need to achieve.

20:28 >> And then people see the success and the wins and they get excited and they keep moving and then eventually they learn these pieces and you just compound and compound and compound. >> Yeah. >> Yeah. >> And everyone wins. >> Yeah. That's to be honest, people want to be on a winning team. My job is not to be anybody's best friend here. My job is to make sure it's working, but that's it. >> I love it, man. Well, I'll get everything I need. This is perfect.

20:46 >> Perfect. >> Thank you for joining us. You're always welcome. >> Thank you, legend.