Latest / Code Story | Startup Podcast for CTOs, CEOs and Technical Founders / S12 E15: The Copilot Fallacy: Why Pure Automation Fails the Real World and the Rise of the Hybrid AI-Human Lifestyle OS with Meghan Joyce, Co-Founder & CEO of Duckbill

Transcript

S12 E15: The Copilot Fallacy: Why Pure Automation Fails the Real World and the Rise of the Hybrid AI-Human Lifestyle OS with Meghan Joyce, Co-Founder & CEO of Duckbill · Code Story | Startup Podcast for CTOs, CEOs and Technical Founders

• 0:00
  [SPEAKER_02]: It was such a jaggy mbp.
• 0:03
  [SPEAKER_02]: Thermophiles, I was actually putting it together nights and weekends, trying to figure out whether there was a there there, because it's not like this vision of the personal assistant for everyone, a way to scale human help is a new idea, a lot of people will try this, and to be totally honest.
• 0:21
  [SPEAKER_02]: They have no figured out the way to leverage technology to give enough vertical integration to get the human servicing to a place of high efficiency and high efficacy, and consistency that consumers expect of this.
• 0:36
  [SPEAKER_02]: I knew that there was a bit of a trail of tears in this category.
• 0:40
  [SPEAKER_02]: My name is Megan Joyce, and I'm the co-founder and CEO of Duck Bill.
• 0:47
  [SPEAKER_01]: This is Co-Story.
• 0:49
  [SPEAKER_01]: a podcast bringing you interviews with tech visionaries.
• 0:52
  [SPEAKER_01]: Six, six months moonlighting goes.
• 0:54
  [SPEAKER_01]: It's the last and all of the backgrounds.
• 0:55
  [SPEAKER_01]: Who share what it takes to change an industry?
• 0:58
  [SPEAKER_00]: I don't exactly know what to do.
• 1:00
  [SPEAKER_01]: It's the only goes to get right.
• 1:02
  [SPEAKER_01]: Who built the teams that have their bad company, is it's that teams help each other achieve this proud of our team.
• 1:09
  [SPEAKER_01]: Keeping scalability top of mind, all that infrastructure was not even fighting it as we grew up.
• 1:14
  [SPEAKER_01]: Total waste of time.
• 1:16
  [SPEAKER_01]: The stories you don't read in the headlines.
• 1:18
  [SPEAKER_01]: It's not an easy thing to achieve.
• 1:20
  [SPEAKER_01]: Because the shelf can decide it off and try to begin to ride the ups and downs of the start-up line.
• 1:26
  [SPEAKER_01]: Need to really want it.
• 1:27
  [SPEAKER_00]: Not just about technology.
• 1:29
  [SPEAKER_00]: All this and more on code story.
• 1:32
  [SPEAKER_00]: on your host Noah Labpart, and today how Megan Joyce is helping you be the most productive you can be.
• 1:40
  [SPEAKER_00]: Using AI combined with actually getting stuff done.
• 1:49
  [SPEAKER_00]: Megan Joyce comes from a long line of people living life to the fullest.
• 1:54
  [SPEAKER_00]: She takes a lot of influence from her grandmother, who was an entrepreneur making and selling dresses in the early 1900s, influencing her to take a hold of every moment in life and capitalize on the time you have.
• 2:05
  [SPEAKER_00]: She's led groups at Uber and Oscar prior to starting a current venture, but outside of tech, she is the mother of three children.
• 2:12
  [SPEAKER_00]: Her favorite hobby is to spend time with the people she loves, meeting them where they are.
• 2:17
  [SPEAKER_00]: But when she has spare time to herself, she enjoys being in nature, hiking, or walking on a beach, in staying active.
• 2:26
  [SPEAKER_00]: Megan was sitting on a bed in Amsterdam and experienced a problem with parental technology,
• 2:36
  [SPEAKER_00]: While sitting on hold with the company, trying to get another one available, she started to wish she had a solution that would help her with this, while she attended her meetings at Uber.
• 2:47
  [SPEAKER_00]: This is the creation story of Duckville.
• 2:54
  [SPEAKER_02]: Duck Bill is the execution infrastructure for the physical world, and that's a mouthful.
• 3:00
  [SPEAKER_02]: The way that most people think of us, frankly, is their place to dump the mess alive.
• 3:07
  [SPEAKER_02]: The personal assistant that actually gets stopped done in the real world for them.
• 3:12
  [SPEAKER_02]: What makes stuck build different and what allows us to be that real execution and infrastructure for the physical world, we sit between AI agents and human operators to complete real-world physical tasks that AI alone can't do.
• 3:30
  [SPEAKER_02]: Whether it's making complex phone calls or running errands or basic maintenance,
• 3:36
  [SPEAKER_02]: Our job is understanding the ever-evolving execution frontier where AI hits the wall.
• 3:42
  [SPEAKER_02]: And to leverage AI as much as possible to get those things done, because that's what's going to give you the fastest, cheapest, easiest path they need to that task.
• 3:52
  [SPEAKER_02]: But ultimately, when something does need to be handed to a human, we do it in a way that is clear and discrete and set up for success, just like an override or a door-dashed or instacart delivery.
• 4:05
  [SPEAKER_02]: So that you can rest assured it gets done well on your bat.
• 4:11
  [SPEAKER_02]: This all arose out of a personal need.
• 4:13
  [SPEAKER_02]: I was an executive in Uber and then at Oscar and building this business I love, growing a family and realizing that I was spending all this spare time and energy, really trying to manage that life admin.
• 4:27
  [SPEAKER_02]: I remember the moment that I became convinced I would start this company one day.
• 4:32
  [SPEAKER_02]: I was actually sitting on the edge of a bed in a hotel in Amsterdam.
• 4:36
  [SPEAKER_02]: I was there for an Uber global leadership meeting.
• 4:40
  [SPEAKER_02]: I had just had my first kiddo, I think about six weeks prior, and I felt it was really important to be at that meeting.
• 4:48
  [SPEAKER_02]: So I flew over the Atlantic, and I landed on the ground, and plugged my pump, my breast pump into a European socket, and promptly blew the fuse.
• 5:01
  [SPEAKER_02]: And so I did this day of meetings, and was trying to make do as creatively as I could,
• 5:09
  [SPEAKER_02]: and was on the phone with the breast pump company at one in the morning Amsterdam time.
• 5:15
  [SPEAKER_02]: Crying to figure out was their way to troubleshoot could they express shitty and new one in Amsterdam?
• 5:21
  [SPEAKER_02]: And I'm thinking to myself, like, I already a multi-billion dollar piano.
• 5:25
  [SPEAKER_02]: I have a little kid at home.
• 5:26
  [SPEAKER_02]: Like, this is the last thing I want to be thinking about.
• 5:29
  [SPEAKER_02]: And yet here I am.
• 5:30
  [SPEAKER_02]: And I just wished there was something some service subperson that I could have offloaded this too in the moment, where I was like,
• 5:37
  [SPEAKER_02]: Here I am at this emergency.
• 5:39
  [SPEAKER_02]: I immediately figure out how to get back up risk-comp immediately, because I need to be in these meetings.
• 5:45
  [SPEAKER_02]: I need to be going this business.
• 5:47
  [SPEAKER_02]: I need to be flying back to see my kiddo in a few days.
• 5:49
  [SPEAKER_02]: Anything that we're waiting on will with customer support and trying to troubleshoot this eight hours later.
• 5:54
  [SPEAKER_02]: So that was the very visceral need that we set out to achieve.
• 6:00
  [SPEAKER_02]: And it's really in that combination of AI plus humans that we've been able to
• 6:08
  [SPEAKER_00]: I want to dive into the MVP, so that first version of Duck Bill that you created.
• 6:13
  [SPEAKER_00]: How want to take to build and what sort of tools were you using to bring it to life?
• 6:18
  [SPEAKER_02]: It was such a jerky MVP.
• 6:22
  [SPEAKER_02]: To me, I was actually putting it together nights and weekends, trying to figure out whether there was a there there, because it's not like this vision of the personal assistant for everyone, a way to scale human help is a new idea.
• 6:36
  [SPEAKER_02]: A lot of people will try this, and to be totally honest, they have now figured out the way to leverage technology to give enough vertical integration.
• 6:46
  [SPEAKER_02]: who get the human servicing to a place of high efficiency and high efficacy and consistency that consumers expected this.
• 6:54
  [SPEAKER_02]: I knew that there was a bit of a trail of tears in this category and I didn't want to take that leak without doing a bit of delgent until I took a page on Travis Calenix's plate of the founder of Uber.
• 7:07
  [SPEAKER_02]: So what I did was hire a personal assistant.
• 7:09
  [SPEAKER_02]: and made her their role to a handful of friends at cost.
• 7:13
  [SPEAKER_02]: I didn't want false plan of do people use this at the free.
• 7:16
  [SPEAKER_02]: I wanted to charge them her real rate, where real are really rate.
• 7:20
  [SPEAKER_02]: There was a very likely weather, but what we found from that.
• 7:26
  [SPEAKER_02]: God's tape and bulgown version of the product was that people walked this product.
• 7:33
  [SPEAKER_02]: In fact, they loved it so much that the group of users very quickly expanded beyond that tight-circle of friends because people were like, I can't really make it really use this.
• 7:41
  [SPEAKER_02]: Her brother would like it.
• 7:43
  [SPEAKER_02]: I have a colleague who's been talking about this and could really use someone.
• 7:46
  [SPEAKER_02]: And so the word in mouth just took off on wildfire.
• 7:50
  [SPEAKER_02]: And then, the second thing we found, that when we were dealing with these really idiosyncratic things in our own personal lives.
• 7:59
  [SPEAKER_02]: Truth be told, each one of us had 70 or 80 different types of things that we dealt with a recourse of the year.
• 8:06
  [SPEAKER_02]: And so no wonder anyone human feels overwhelmed, there are 70 or 80 different types of things that the average adult does on their to-do list at anyone given time.
• 8:18
  [SPEAKER_02]: and that's a lot for a human, but it's actually not a lot for an AI system.
• 8:23
  [SPEAKER_02]: To manage at the New York 80 different types of tests, some by the way, the types of firms that are distilled down from those.
• 8:31
  [SPEAKER_02]: And so what we did over time was build increasing data collection and technology around all those human activities, and at the point at which that we had done about
• 8:41
  [SPEAKER_02]: 10 million a fan, which was just about a year ago, we could train our models to be good enough that we got the prior to a point where it was consistently delightful, totally reliable and skimpled.
• 8:58
  [SPEAKER_00]: So you've got that MVP, you've got it to the point you want it.
• 9:01
  [SPEAKER_00]: How did you progress it and mature it from that point?
• 9:04
  [SPEAKER_00]: And again, you touched on this at a high level, but want me through that.
• 9:06
  [SPEAKER_00]: And I think to wrap it in a box a little bit when looking for it, how did you build your roadmap?
• 9:11
  [SPEAKER_00]: How did you go about deciding that?
• 9:13
  [SPEAKER_00]: Okay, this is the next most important thing to build or to address with DuckBuild.
• 9:19
  [SPEAKER_02]: Because we work, participated in this widely fast-moving, well-funded AI economy.
• 9:26
  [SPEAKER_02]: We felt that it would be unwise to get too big or too commercial without having a product that we know of work.
• 9:38
  [SPEAKER_02]: The best way to grow our business was to have a product set.
• 9:43
  [SPEAKER_02]: People were in love with.
• 9:45
  [SPEAKER_02]: Because when people are in love with your product, they do use it more and more every time.
• 9:50
  [SPEAKER_02]: And that's the best way to get a product off the ground, especially when you're building a category, especially when you're doing something new.
• 9:58
  [SPEAKER_02]: Because those people are not the takers or read actors who want to try untest his things for fun, they want something that is recommended by commonly trust.
• 10:09
  [SPEAKER_02]: We actually had a wait list up.
• 10:11
  [SPEAKER_02]: We were in cell for a long time and had a wait list up.
• 10:14
  [SPEAKER_02]: And we allowed people to refer others.
• 10:16
  [SPEAKER_02]: We tinkered with what it would take to serve the well.
• 10:20
  [SPEAKER_02]: The leader caught totally flatfooted by the time we got to our birth base.
• 10:23
  [SPEAKER_02]: But I
• 10:24
  [SPEAKER_02]: It was a very controlled for him.
• 10:27
  [SPEAKER_02]: But it still did two things.
• 10:28
  [SPEAKER_02]: Number one, until we had nailed reliability.
• 10:33
  [SPEAKER_02]: That was our wrth star at Uber.
• 10:34
  [SPEAKER_02]: It was our North Star Oscar.
• 10:36
  [SPEAKER_02]: And if someone is waiting to use this for helping their personal life, it has to execute reliably.
• 10:41
  [SPEAKER_02]: And to be honest, I can't tell you how many people.
• 10:44
  [SPEAKER_02]: have come back to me now in 2025 to reach this is that I tried your problems a few years ago and loved the vision but it wasn't totally reliably yet and I returned recently and fully how it is a different all game.
• 10:58
  [SPEAKER_02]: And that's why we waited to really light the full fire, because we knew it wasn't ready for prime time.
• 11:06
  [SPEAKER_02]: Number two, we need is to make sure that we weren't losing money with every incremental tax that you outscore, which really needs first margin, re-keed in our profitability, which is challenging when it comes to a human-enginal loop situation.
• 11:19
  [SPEAKER_02]: And I probably believe that services or human-enginal loop were technically
• 11:23
  [SPEAKER_02]: Canadian services can be done wide, because all this is the whole basis of Sequoia new investment thesis, but you have to know what you're getting yourself into and have the discipline to not treat it like that.
• 11:34
  [SPEAKER_02]: Because it doesn't have a reset of art and other aid.
• 11:37
  [SPEAKER_02]: And so, our roadmap followed those two things, the number one priorities about building reliability, and there were a number of different facets to that, but it was all about building the technology that could not only automate as much as possible and leverage the humans for what humans are really uniquely good at.
• 11:54
  [SPEAKER_02]: But also setting the humans up for success, that's that by the time the work got to them, it was packaged in a way that they didn't need a lot of training or expertise to do well.
• 12:04
  [SPEAKER_02]: They just needed to be as smart, capable of human with good judgment in the real world, and like the average Uber driver partner or door-and-ass friends to cart delivery person.
• 12:14
  [SPEAKER_02]: And so it took a while to get there.
• 12:16
  [SPEAKER_02]: Huge amount of tooling, a core and data and model training and retraining to get there.
• 12:24
  [SPEAKER_02]: And then, we thought, came Groze Barton, and profitability and sustainability, and the really nice thing about this model is that it actually gets more profitable as it gets more delightful, because it means less rework, it needs its faster turnaround time, and those things are a rising tide, but has lifted all boats.
• 12:44
  [SPEAKER_02]: So it's been, it's been a real journey.
• 12:47
  [SPEAKER_02]: But those two things were our only priorities for a long time.
• 12:51
  [SPEAKER_02]: And it was only when we got to 99% reliability and real meaningful gross margin that we even started to look at growth.
• 13:01
  [SPEAKER_02]: And that's really that a relatively new thing.
• 13:05
  [SPEAKER_00]: Okay, let's flip to team then.
• 13:07
  [SPEAKER_00]: So I hear you saying, we a lot.
• 13:08
  [SPEAKER_00]: How did you build that team?
• 13:09
  [SPEAKER_00]: And what do you look for in those people to indicate that they are the winning horses to join you?
• 13:15
  [SPEAKER_02]: I, number one, recruited a lot of people that I had worked with in the past that I trusted deeply.
• 13:21
  [SPEAKER_02]: There are folks on my team from Oscar and Uber and those people hired folks that they could work with in the past, restars, other organizations that they had worked in.
• 13:33
  [SPEAKER_02]: And that really saved us a lot of an onboarding getting to know you time.
• 13:39
  [SPEAKER_02]: You already knew each other's idiosyncrasies and ways of working and what good looks like and we had a common life to come from the experiences that we had shared in the past.
• 13:48
  [SPEAKER_02]: It allowed us to just get to building as quickly as possible.
• 13:51
  [SPEAKER_02]: And it was selfishly so much fun as someone who has worked with thousands of people to really get the gang back together and say, we're some of the best people I've worked with in the past, like, let's see if we can get it on board and have fun while we're building.
• 14:05
  [SPEAKER_02]: Y'all work way too hard not to unlock the people we're building alongside.
• 14:10
  [SPEAKER_02]: So that was my criteria number one.
• 14:13
  [SPEAKER_02]: Someone I really trusted who I knew to be a star from a past environment.
• 14:17
  [SPEAKER_02]: And number two, there was a lot of exploration in the early days of building an AI around what was the right profile of person for this product.
• 14:29
  [SPEAKER_02]: And we talked to everyone under the sun.
• 14:32
  [SPEAKER_02]: AI researchers, folks who had been working in this field forever, data scientists, she learning experts.
• 14:41
  [SPEAKER_02]: And we found that a couple things were wildly predictive
• 14:48
  [SPEAKER_02]: Number one, cure hunger and passion for the mission.
• 14:52
  [SPEAKER_02]: Nothing can replace that.
• 14:54
  [SPEAKER_02]: And when you have it, you're gonna be thinking about it in your spare time, you're gonna be reading up on this subject over the weekends.
• 15:00
  [SPEAKER_02]: You're really gonna care about getting this right.
• 15:03
  [SPEAKER_02]: Number two was an innate interest in AI.
• 15:07
  [SPEAKER_02]: The space was moving so fast that just because you had gotten a PhD in AI some many years ago, didn't mean you were up to date on the latest and greatest.
• 15:16
  [SPEAKER_02]: And in order to stay current,
• 15:18
  [SPEAKER_02]: and having people who are drawn to the space and reading up on the white papers and assessing the field and the landscape because they found it fascinating was a better predictor for us than deep experience and knowledge in the space.
• 15:37
  [SPEAKER_02]: The third one was you had been in an environment that combined technology and humans before.
• 15:43
  [SPEAKER_02]: It is a fundamentally different thing to build pure sats or pure software or social network or things like that.
• 15:50
  [SPEAKER_02]: Versus a system like Uber or Door-Ash or Instacord or Amazon or a firm or plated that really bring together technology
• 16:01
  [SPEAKER_02]: A hand-real world logistics and real-world humans and humanity, be in all the wonderful and challenging things that come with that?
• 16:10
  [SPEAKER_02]: Because there are just some works and aspects of that space that create incredible opportunity if you know if you have the pattern recognition for it.
• 16:22
  [SPEAKER_02]: Those are the companies that a lot of our early people have drawn from and combined with hiring people we've worked with in the past to have a passion for what we're building and really love staying abreast of the latest technology in AI has gotten us the winning team.
• 16:41
  [SPEAKER_00]: You mentioned scalability earlier, so I want to double back on that.
• 16:45
  [SPEAKER_00]: How did you approach scalability in the beginning?
• 16:47
  [SPEAKER_00]: And how do you think about that?
• 16:49
  [SPEAKER_00]: And then, have there been interesting areas where you've had to fight scale as you've grown?
• 16:54
  [SPEAKER_02]: Absolutely.
• 16:56
  [SPEAKER_02]: So much of the first few years were building systems, scaling them, and getting to a point where, yep, this is where the next thing has broken.
• 17:06
  [SPEAKER_02]: Right?
• 17:06
  [SPEAKER_02]: So much of business, so much of operations is chasing bottleneck to bottleneck.
• 17:10
  [SPEAKER_02]: If you're doing your job well, you're constantly running into gaining items to the next phase of growth.
• 17:17
  [SPEAKER_02]: I think the best operators, especially in a high growth environment, are the ones who are seeing around corners before the volume or before the bottlenecks arise so that they can get proactive about tackling them.
• 17:33
  [SPEAKER_02]: But in our first few years, it was really that around binding either proactively or reactively in the bottleneck that we're getting in the way of scalability and letting a group of folks off the way list and then they, okay, I'll hear it be found the next one.
• 17:50
  [SPEAKER_02]: And I think a lot of the bottlenecks that we found peeing down to the challenger, what we were building and what makes this space so difficult, frankly,
• 18:01
  [SPEAKER_02]: As much as we all have 70 or 80 things on our to-do list, no two tasks look exactly the same.
• 18:08
  [SPEAKER_02]: There are no two duplicates within our, however many tens of millions of real-world interactions have looked exactly the same.
• 18:18
  [SPEAKER_02]: Even if it's a recurring task that you need, maybe you need your haircut book every month.
• 18:23
  [SPEAKER_02]: It's a new day, it's a new schedule.
• 18:25
  [SPEAKER_02]: Goodness knows something might be going on in town.
• 18:27
  [SPEAKER_02]: The barber might be on vacation.
• 18:29
  [SPEAKER_02]: And you might have had a change in your schedule.
• 18:32
  [SPEAKER_02]: Maybe the barber is retiring.
• 18:33
  [SPEAKER_02]: There are always idiosyncratic things about what you need and what that vendor is able to provide that cause no two work flows to look exactly the same.
• 18:46
  [SPEAKER_02]: And this is why services have not been vertically integrated or scaled outside of a few selecting samples because it's really hard to anticipate and build systems and accommodate all those videos and crises.
• 19:01
  [SPEAKER_02]: And that has now changed things to AI, but it doesn't mean that take AI as the shelf and it can just accommodate them out of the gate.
• 19:09
  [SPEAKER_02]: You need to build a set of models that accommodate multi-turns.
• 19:14
  [SPEAKER_02]: And you need to aggravate enough data about the kinds of failure cases that do create bottlenecks, whether it's because the human worker or the vendor,
• 19:25
  [SPEAKER_02]: Certain types of failure modes, they're on vacation, they're busy, they change their prices.
• 19:31
  [SPEAKER_02]: Or because the customer has idiosyncratic needs.
• 19:36
  [SPEAKER_02]: They want to switch their hair dresser.
• 19:38
  [SPEAKER_02]: This time they want to get coloring in addition to a cut.
• 19:41
  [SPEAKER_02]: You're booking your restaurant reservation, but this time there's someone in your party who has a nut allergy.
• 19:45
  [SPEAKER_02]: Or you need to expand the size of your party all of a sudden because you have folks in town.
• 19:50
  [SPEAKER_02]: all of those things.
• 19:52
  [SPEAKER_02]: We need to be distilled down into a set of predictive manageable atomic bite-sized pieces that can then be dealt with by AI or humans reliably.
• 20:04
  [SPEAKER_02]: And every single one of them was the bottleneck.
• 20:07
  [SPEAKER_02]: In Zollary, we did reach that cut of 10 million interaction rate where we had enough data to get predictive and no single type of bottleneck.
• 20:18
  [SPEAKER_02]: was capping our growth at any given moment in time.
• 20:23
  [SPEAKER_00]: Okay, so as you step out on the balcony, and you look across all that you've built thus far with duck build, what are you most proud of?
• 20:30
  [SPEAKER_02]: I think it's the opportunity that we have now created, having followed a mission and a dedication to serving real people,
• 20:46
  [SPEAKER_02]: and being creative and...
• 20:50
  [SPEAKER_02]: flexible in how we do that, meaning there are a few things that were really important to us.
• 20:56
  [SPEAKER_02]: We really wanted to give genuine leverage to everyday people on the hardest last mile task that really dragged them down.
• 21:07
  [SPEAKER_02]: Nobody had figured this out yet and that's what we wanted to solve.
• 21:10
  [SPEAKER_02]: But we were pretty agnostic about how much AI and how much humans in the loop and how much robots
• 21:19
  [SPEAKER_02]: Because we know that if we did this well, we would be making the world better for those people and if you believe that humans are required in the loop for, at least some of the foreseeable future, and that's indeed what we found, then you'd be creating noble work opportunities for those people as well.
• 21:36
  [SPEAKER_02]: Everything else was unfortabate.
• 21:39
  [SPEAKER_02]: Everything else was up for experimentation, and
• 21:43
  [SPEAKER_02]: What my team did was take those perhaps contrarian bets at a moment of time when a lot of people are thinking about AI's job replacement or AI's a standard set of things that does not actually touch the last mile and hit the wall at the point of the last mile and bet on those things.
• 22:01
  [SPEAKER_02]: And frankly, a lot of people told us it was a possible.
• 22:04
  [SPEAKER_02]: Well, first that we were foolish to try to do it.
• 22:07
  [SPEAKER_02]: Or that I've been going to require a ton of blood sweat and tears and they were right about that.
• 22:12
  [SPEAKER_02]: But we stood up with it and we did it smartly and we were incredibly to nature.
• 22:17
  [SPEAKER_02]: And now, at this point, not only have we proven this out on well over 10 million real-world action, but now we're starting to function as the infrastructure layer, such that when those AI assistants in autonomous agents hit a wall, engineers and customers and enterprises are starting to flumb.bell in to be their last mile execution layer.
• 22:42
  [SPEAKER_02]: or some of those large sources of aggregated labor are saying, I want to make sure I have a clear runway for all of my people to keep doing work in the AI future.
• 22:54
  [SPEAKER_02]: As robots and autonomous delivery agents and things like that start to cut into their supply hours, and wow, Duck Bill is actually creating work that they can stop up.
• 23:05
  [SPEAKER_02]: as a result of this infrastructure layer.
• 23:07
  [SPEAKER_02]: And so I think as contrarian as it has felt at various moments in the last several years, this idea of physical AI or AI extending into the physical world is now a place that people are turning and saying, of course this will happen.
• 23:23
  [SPEAKER_02]: And I think because we made a few gut seed bets and stuck with it through thick and thin, we are able to meet this moment over the only ones.
• 23:35
  [SPEAKER_00]: OK, Megan, let's flip the script a little bit.
• 23:38
  [SPEAKER_00]: Tell me about a mistake you made and how you and your team responded to it.
• 23:44
  [SPEAKER_02]: There were here full of bets that were just not built up the pace that this moment in time requires.
• 23:55
  [SPEAKER_02]: Mediawee set out with a six-week long road map.
• 23:59
  [SPEAKER_02]: Even a three-week long road map.
• 24:01
  [SPEAKER_02]: And midway through that build, we would get new information, or we would see a new capability hit the market that would teach us something about how we should be building and how we should be approaching the problem.
• 24:15
  [SPEAKER_02]: And then we were a moment in time where we were midstream on a build.
• 24:20
  [SPEAKER_02]: Nobody wants to tell a team.
• 24:23
  [SPEAKER_02]: You've been building this thing for however many weeks.
• 24:25
  [SPEAKER_02]: It'd be months in some cases, and we're gonna throw it out and do something differently.
• 24:30
  [SPEAKER_02]: It also meant that by the time you were done building it, it was pretty expensive because it had taken several people however many weeks, if not months, to get out into the wild.
• 24:40
  [SPEAKER_02]: And so it kind of cost of ditching those things as they didn't work.
• 24:44
  [SPEAKER_02]: And I think the tooling has gotten better and enabled us to ships so much faster.
• 24:50
  [SPEAKER_02]: But there are also some, I don't know.
• 24:53
  [SPEAKER_02]: mentality and processing culture changes that needed to happen to say we're not even going to deal with spreads.
• 25:01
  [SPEAKER_02]: We are truly going to be shipping every day.
• 25:05
  [SPEAKER_02]: We are going to build a team of agents that do our coding for us so that we can think the big thoughts and put things into the wild and try things out that are scrappy and quick and allow us to learn.
• 25:16
  [SPEAKER_02]: And then if it works, build on it, double-down, do the full-fidelity version of that MVP feature, but do it with the knowledge that we got it out into the real world and it worked or did it.
• 25:27
  [SPEAKER_02]: And that was a huge oomock for us, where, you know, even if we were building faster, then some of us had ever built in the past, it wasn't fast enough for this moment in time.
• 25:37
  [SPEAKER_02]: And I think one of the biggest unlocks there was a big film that I just felt in my gut we really needed to do.
• 25:44
  [SPEAKER_02]: I'll be totally candid with you.
• 25:46
  [SPEAKER_02]: I wasn't 100% sold on it, but I knew it would take us 36 hours.
• 25:53
  [SPEAKER_02]: One of the most exciting side effects that I was looking forward to of this build was that it was going to be a tricky one.
• 26:01
  [SPEAKER_02]: And if we did it, it would show us we can build hard things fast.
• 26:06
  [SPEAKER_02]: And so I pushed the team to do it.
• 26:08
  [SPEAKER_02]: They did it.
• 26:09
  [SPEAKER_02]: They did it better than I ever could have expected our dream.
• 26:13
  [SPEAKER_02]: It got near immediate uptake.
• 26:15
  [SPEAKER_02]: And it really proved to us just how fast and aggressively we can go.
• 26:22
  [SPEAKER_02]: And I have seen a market increase in pace from the team since then, it's amazing what proving to yourself that you can do the hard things can do for your speed and confidence and wobble of Russian.
• 26:37
  [SPEAKER_02]: How much we all need that right now in this moment in time.
• 26:44
  [SPEAKER_00]: Okay, Megan, let's move forward then.
• 26:45
  [SPEAKER_00]: What does the future look like for Duck Bill?
• 26:49
  [SPEAKER_00]: Where the
• 26:53
  [SPEAKER_02]: So I'm talking about where we started, which was bringing this product to consumers.
• 26:59
  [SPEAKER_02]: It was really important to me that we'd build the product in such a way that consumers loved it and were willing to pay for it.
• 27:06
  [SPEAKER_02]: As a consumer person at my core, that to me is the surest sign of product market set and of a product that people really love.
• 27:16
  [SPEAKER_02]: Uber famously resisted offering a B2B product for a long time because
• 27:21
  [SPEAKER_02]: They knew it would be a distraction the minute that we started selling to corporate travel departments or HRTs.
• 27:27
  [SPEAKER_02]: We really wanted to build something the consumer loved and don't built it took the same approach.
• 27:33
  [SPEAKER_02]: And now that we have it, that next phase really is.
• 27:37
  [SPEAKER_02]: making it easy for all of those builders and engineers and founders who are either running or building their own companies.
• 27:49
  [SPEAKER_02]: To plug Duck Bill in is that last mile execution layer.
• 27:52
  [SPEAKER_02]: To plug us in as the infrastructure layer, we're physical execution.
• 27:58
  [SPEAKER_02]: And I think of it not to similar from stripe, stripe had a incredible timing at in the early days of the e-commerce film, expecting the both supply and demand would proliferate in e-cars, but that building the payments infrastructure would not be worth folks' time.
• 28:18
  [SPEAKER_02]: And we've become out of the same way.
• 28:20
  [SPEAKER_02]: The foundation models we see as increasingly commoditizing labor we see as only growing in this environment, the demand for good work.
• 28:30
  [SPEAKER_02]: And we believe that the demand and supply for services delivered through AI, through AI agents or autonomous assistance, is only going to boom, we're seeing it already.
• 28:43
  [SPEAKER_02]: And so what we've built is that infrastructure layer that any builder can take off the shelf to facilitate.
• 28:49
  [SPEAKER_02]: the distilling of that last mile services down into something that is reliable and sustainable.
• 28:56
  [SPEAKER_02]: Whether they're bringing their own agent, either bringing their own workers, but Duck Bill is the infrastructure layer that connects them and facilitates it all.
• 29:05
  [SPEAKER_02]: I think about the third growth horizon, really, as being all about our data.
• 29:10
  [SPEAKER_02]: And so given that we've done a well over 10 million real-world actions today, we have the world's largest database of real-world data.
• 29:19
  [SPEAKER_02]: What it takes to hand the baton into the offline world to complete physical execution.
• 29:24
  [SPEAKER_02]: And so we had a full heat map across the country at Gassen supply and demand where there's enormous amounts of desire from consumers and willingness to pay for services, but the supply is not meeting the moment.
• 29:37
  [SPEAKER_02]: Think about how hard it is to find plumbers or land scapears in certain markets.
• 29:44
  [SPEAKER_02]: not every plumbing or landscaping job should be done by AI plus human in the loop but a lot of it can be.
• 29:51
  [SPEAKER_02]: In ways that are higher reliability, higher quality, more consistent and create more lucrative labor for the person doing the work.
• 30:01
  [SPEAKER_02]: And so...
• 30:02
  [SPEAKER_02]: We continue to build this database, we also are pushing further into the last mile and directing our technology and human and low loop functionality towards more and more services.
• 30:16
  [SPEAKER_02]: Right now, we do the complex phone calls and the offline transactions and basic errands and maintenance.
• 30:22
  [SPEAKER_02]: But you can see that going into more complex maintenance.
• 30:26
  [SPEAKER_02]: more skilled, artisanal services over time.
• 30:31
  [SPEAKER_02]: And that is where I think the opportunity for this is truly endless.
• 30:35
  [SPEAKER_02]: Services are a multi-trillion dollar industry, and I think this is how we start to bring them into the 21st century in an AI account.
• 30:44
  [SPEAKER_00]: Awesome.
• 30:45
  [SPEAKER_00]: The future is bright.
• 30:46
  [SPEAKER_00]: Let's wish to you, Megan, who influences the way that you work.
• 30:50
  [SPEAKER_00]: They're a person or many persons or something.
• 30:52
  [SPEAKER_00]: You look up to and why.
• 30:54
  [SPEAKER_02]: I remember my mom telling me when I was a leader in a school play in seventh grade that the best leaders bring out the best in the people around them, and that really that's what leadership is all about.
• 31:08
  [SPEAKER_02]: And if fundamentally changed how I thought about leadership, I think prior to that is a kid I thought about leaders is people who are at the center of things and it was all about them.
• 31:19
  [SPEAKER_02]: And this really challenged me to think about my role here is to help bring out the best in everybody around me.
• 31:27
  [SPEAKER_02]: And to ensure that your people are setting their own high bar, even when you're not looking.
• 31:33
  [SPEAKER_02]: And I think if you can combine those two things up high standards and high support, that's where excellence becomes really sustained.
• 31:39
  [SPEAKER_02]: Where you have people who believe they can do the impossible, and then they go do it over and over again.
• 31:45
  [SPEAKER_02]: That's what I had, Uber and Oscar, and it's what we're doing at Deckville.
• 31:49
  [SPEAKER_02]: And there's no greater place to be as a teammate, but also as just a human, I think to have that confidence building of knowing you can build something really hard and change people lives as a result.
• 32:06
  [SPEAKER_00]: Okay, last question Megan, and it will be interesting to hear if your answer is an echo of what you just said or if there's something else you want to focus on, but I'm curious.
• 32:14
  [SPEAKER_00]: So you're getting on a plane and you're sitting next to a young entrepreneur who's built the next big thing.
• 32:20
  [SPEAKER_00]: They're jazzed about it, they can't wait to show it off to the world and can't wait to show it off to you right there on the plane.
• 32:24
  [SPEAKER_00]: What advice do you give that person having gone down this road a bit?
• 32:28
  [SPEAKER_02]: I remember sitting in a business school classroom listening to Kevin Scherer, who was the outgoing CEO of AMGEN.
• 32:36
  [SPEAKER_02]: Talk about his advice to young business leaders and entrepreneurs, and he said, it's really important to take care of yourself and sleep well and eat well and exercise and spend time at the people you love.
• 32:50
  [SPEAKER_02]: And I remember chuffling and saying, yeah, when I, too, am a retiring CEO and I have generated
• 32:56
  [SPEAKER_02]: thousand next shareholder return, then I'll find time to sleep well and spend time with people I love and all that good stuff.
• 33:03
  [SPEAKER_02]: And just a few weeks later, I was starting at Uber, and I was working probably like similarly very long hours as I was working in my days in private equity, for example.
• 33:15
  [SPEAKER_02]: But it was a much different kind of work.
• 33:17
  [SPEAKER_02]: We were building.
• 33:18
  [SPEAKER_02]: There was a lot of people leadership.
• 33:21
  [SPEAKER_02]: We were having to be very aggressive in building, but also deal with some really tough issues as we went.
• 33:28
  [SPEAKER_02]: Thus those bottlenecks and be really thoughtful about how we build to do it in a high integrity way.
• 33:34
  [SPEAKER_02]: And that was draining to my energy in ways that sitting per 16 hours a day behind a spreadsheet weren't.
• 33:43
  [SPEAKER_02]: And I realized that advice that Kevin Share was giving was so precious that I really had to start investing in my own well-being if I wanted to show up for the people around me.
• 33:58
  [SPEAKER_02]: and they were looking to me for energy.
• 34:00
  [SPEAKER_02]: They were looking to me for a looking feel.
• 34:04
  [SPEAKER_02]: And so, not every one of us can invest every day of eight hours of sleep and eating freed-alls meals and spending time with our loved ones, like we're building companies that's really hard.
• 34:13
  [SPEAKER_02]: But if I pick a couple priority things that I know are essential to my mental health and my well-being,
• 34:21
  [SPEAKER_02]: I've found that they provide enough of a cornerstone to keep you going.
• 34:25
  [SPEAKER_02]: And I often find when I'm feeling particularly anxious or lost, it's because I haven't been investing in those couple of things.
• 34:32
  [SPEAKER_02]: And I need to get back to them.
• 34:33
  [SPEAKER_02]: So maybe six hours of sleep and night is my German a mom or having something to eat, bring me all the things that's not pop chips.
• 34:41
  [SPEAKER_02]: And so I've invested in those things to make them work and I've never regretted it.
• 34:48
  [SPEAKER_00]: That's fantastic advice.
• 34:49
  [SPEAKER_00]: Well, Megan, thank you for being on the show today, and thank you for telling the creation story of Duck Bill.
• 34:55
  [SPEAKER_02]: It is such a pleasure, Noah.
• 34:56
  [SPEAKER_02]: Thank you so much for having me and for all that you do and bringing these stories to the world.
• 35:02
  [SPEAKER_00]: And this concludes another chapter of Code Story.
• 35:10
  [SPEAKER_00]: code story is hosted and produced by Noah Labhart.
• 35:13
  [SPEAKER_00]: Be sure to subscribe on Apple podcast, Spotify, or the podcasting app at your choice.
• 35:19
  [SPEAKER_00]: And when you get a chance, leave us a review.
• 35:21
  [SPEAKER_00]: Both things help us out tremendously.
• 35:26
  [SPEAKER_00]: And thanks again for listening.