• @jamesclift on Twitter
• Durable

Allen: Welcome to It Shipped That Way, where we talk to product leaders about the lessons they’ve learned helping build great products and teams. I’m Allen Pike, and today’s interview is with James Clift. James is the founder of Durable, which in its first two years has raised over $20 million and he has previously funded many other businesses including VisualCV. Welcome, James.

James: Thanks for having me.

Allen: I’m excited to have you on the show. As you were mentioning beforehand, we’ve crossed indirectly paths. I’ve seen your name and your story go by over the years of building businesses. And so now we get a high bandwidth way to get to know each other under Q&A about some of the many things that you’ve learned building various businesses.

James: Awesome. Yeah, super stoked to get into it.

Allen: Before we start digging into those specific lessons learned, I have some specific areas that I want to dig into that I know you’re going to have some things to share on in terms of having built this many businesses, mindsets for doing so, strategies, as well as some of the hard learned lessons. Because you can’t build a few businesses without breaking a couple of eggs and learning from it. I wanted to give you a chance to summarize the James Cliff story so far, which I am sure over the last couple of years raising funding and building a business you’ve had a lot of chance to tell that story in a pretty concise way, but how do you like to summarize the origin story, so to speak?

James: I like to say I’ve been unemployed my entire career, so the last job I had was a part-time job coaching tennis when I was 16. Really just been an entrepreneur my whole life. I don’t even look at it as entrepreneurship, I’ve been building and owning businesses. Started off with a web design company back in high school, ran a home services business in university, HD window cleaning, still my favorite business, just working with my buddies outside in the summer and made some good cash. And then went back to the web design world after graduating very long, probably three year journey of shipping a bunch of different stuff. We had a hardware product that was a day lighting system for commercial office building, so super hard tech. We’re like, “We won 20 grand in business plan competitions. This is enough to fund the entire company for five years.” Just naive young kids trying to build a real business. That didn’t work out, went back to web design consulting. That was hard. I think we were still in that mode of, “Hey, we’re trying to build a product that scales and that works.” Probably a dozen things that we launched. One of them ended up being this site called VisualCV. That’s the one that worked basically top three resume builder on the internet. That grew very quickly, so zero to about a million in revenue in the first 12 months with three employees, so super profitable. Scaled that for about eight years and then sold it in 2020. Had a brief stint building remote work software pre-COVID, so side project that actually took off pretty quickly. It was like remote work productivity, so kind of in the virtual office category. Lots of great companies signed up. It was like Netflix, BarkBox, SurveyMonkey. Never really hit breakout product market fit, but definitely real traction. Ended up selling that one, when was that? Probably 2021. And then started Durable, and Durable is basically this simple thesis and the convergence of all the things that I believe. But the TLDR’s, it’s way too hard to own a business and to start a business. And I think business is the key to economic freedom for billions of people. Basically, how do we unlock that? And if you look at, I mean labor, it’s like employment or business ownership. Employment I think is going away with how AI is disrupting the entire economy. And that trend was already happening, so business ownership is growing super quickly, especially around the world. Primarily very, very small companies. Like sub 10 employees, usually sub one or sub two employees, or one-person companies. And then we just thought about how do we abstract and automate all the business parts? If you have a skill, you should be able to have a successful, profitable business in a very short amount of time. If you’re really good at landscaping, personal training, consulting, why do you have to learn web design, marketing, customer success, accounting, tax, all this other extraneous things? It’s really no, there’s an economic value for your skill. And you should be paid for that and take the lion’s share of that as opposed to your employers. That’s kind of the core thesis, how do we automate an abstract all parts of running a business so you the individual can just focus on what you’re great at? And AI was the big unlock for that. We had that core thesis, we’re trying to build with much worse models originally, so early to the open AI train. And then kind of GDB-3.5 was the big unlock. I was like, “Wow, this vision’s actually possible now.” I think we were on the path of being kind of a bundled up business in a box with traditional SaaS, which would have been successful. But the opportunity of AI is like, you can actually solve these workflows now, so you don’t have to build this copycat software. You can build brand new user experiences that actually start to solve the problem for the individual. That wasn’t a concise summary, but yeah, Durable has been growing super quickly. Like you mentioned, raised a bunch of cash. Team’s about 20 people now. Businesses scaled a lot in the past 18 months. And yeah, it’s been super intense, a lot of fun and lots of work left to do.

Allen: There’s at least five questions that spring out of that. Thinking a little bit about some of these previous businesses you talked about. You already touched on this a bit, some of the lessons you learned. Like you I also founded, founded, it didn’t even feel like founded, I started doing contract web design for people when I didn’t know what I was doing and I was 19 or whatever like that. And you learn, there’s obviously really obvious ways that you fail when you don’t know not to make a fixed bid on something when you haven’t seen the code yet, and those easy lessons. They’re unpleasant, but they’re obvious. What are some of the harder one or subtler lessons that you learned in your previous iterations having built some of these businesses so far?

James: I mean, lots of tactical lessons. I think the most important one, it’s like, what advice would you give your 20-year-old self? And it’s one, it’s going to work out. I think this long-term view of, hey, if you keep at it, you keep shipping, you keep learning, you will be successful over a long enough timeframe. And then how do you do it with less of the stress? I think that was the biggest thing. I mean running a business, there’s this constant stress that happens at whatever scale. It’s like, “Hey, I need to pay rent this month, or I need to pay my employees this month and now we need to grow and provide a return.” All these things just add up both the little things and the small things. I think that time period, especially in that first three years when things were not working out were super stressful and I was really hard on myself. I think it’s just, I mean, I’m still really hard on myself, but it’s getting better. I think it’s just, “Hey, this is going to be hard. You’re choosing to do something that is by nature difficult, but just keep that long-term optimism,” which I always had. But that short term, “Hey, try and enjoy the process more.” I think just generally for anyone that’s starting off on this journey, I have dozens of things that I forget about that we ship. It was months of energy, super excited about, I don’t even know what it is, I can’t even tell you what it was. There’s all these projects I find in my old notes that I was like, “Oh, wow, I did that. That was a dumb idea.” Some was really good too. I was like, “I should have done that idea. That was a really great idea.” But I think it’s just the mindset of, “Hey, this is a long-term game that you’re trying to learn and each level is different.” I think just being able to abstract a bit and not take it so seriously in a way while still working really hard. I think that’s the core. If you’re going to do this, just be fully aware of what you’re signing up for and try and enjoy the process a bit more. I think that’s the more mindset that I think is super important, and that’s the advice I’d give myself back in the day.

Allen: Yeah, I appreciate that, there’s kind of infamously, even though it’s sort of infamous, I think every founder needs to find their own path to it. Need to both be continually and stubbornly optimistic that we’ll find a way and it will eventually work out and that we can overcome the challenges, but also clarity about the challenges are real and they will kill you if you don’t do something about them.

James: Yeah. It’s like long-term, optimistic, short-term pessimistic.

Allen: Clicking into one of those specific stories. I appreciated that you wrote a bit about Holopod, which is the remote work startup that you touched on. Often when people have something that isn’t maybe an on-brand mega hit, it can be hard for us to call attention to it. And you went and did the work of helping other founders and people think about their own situations by writing about it. And something that I found interesting in that story that you wrote is that you had 15% month-on-month growth in that business, which if you’ve never founded a funded startup, then you’d be like, “Okay, I guess it sounds like you succeeded, then 50% month-on-month growth, that would compound really large eventually.” Tell us a little bit about any sort of tactical or reflections that you had from that path in trying to build that business? Which obviously is going to contrast to the really successful ones.

James: Yeah, that’s an interesting case study. I mean, I wrote that right when it was failing. And then I was about to like, “I’m not going to publish this.” I mean, it’s kind of a cop out. I waited until I was “successful” again, then I published it, but at least the data’s there.

Allen: I’ve been guilty of doing the exact same thing.

James: Yeah, exactly.

Allen: Three years ago, here’s how I failed and this is why I’ve now overcome that.

James: I didn’t have the guts to actually publish it at the time. But it was close. But yeah, that was an interesting business. That was a side project basically in this remote work productivity space. The thesis was it’s super annoying to work fully remote, especially with these bombardment of notifications that happens all the time. Still an unsolved problem, but it’s like Slack, Google Calendar. Then, so basically how do you solve that issue of, “Hey, I want to collaborate more closely with my team, but I don’t want to be constantly online and bombarded with notification.” Basically, pretty great app. It automated your Slack status based on your current work state. We detect whether you’re in a meeting out for lunch, available to co-work on your calendar, or actually based on your desktop. If you’re in your terminal, then we detect that, or you’re in Figma, like, “Hey, I’m heads down right now, leave me alone.” Super cool app. I was trying to think the actual timeline, but it was live during VisualCV, so side project I was working on with my friend, and then post exited VisualCV, I was still like, “Okay, I don’t really know what to do.” So, I jumped right into that project, it had decent traction. And then COVID hit and it really blew up. Just like people were looking for one, solving this specific problem, they’re looking for apps that make their offices more productive. Yeah, crazy breakout momentum. Yeah, you’re like, I’m looking at the dashboard. I’m like, I don’t know, Uber signing up, public companies. What else is there? Zapier, so really cool companies that are using the product. But just never hit that. We had some virality within Teams, actually decent, because it was year one of revenue. I think we probably hit 80 or 90K, so pretty decent.

Allen: Year one.

James: Yeah, year one. And then I think the challenge there was both my co-founder, I had to go get a job because I wasn’t generating revenue. I had self-funded it, so I put a significant bit of cash into the business. I was trying to hire people to work on it. Just never really hit our stride on team. I think number one, it was like I’m trying to figure out how to grow the team while growing the business. I think the other core part of that was, just wasn’t a business I cared that much about. I’m not really a productivity nerd. I’m probably the least structured person. Is this actually the app that … It was solved, we liked using it, people really liked the product. But just is this the thing I want to spend a decade on, especially for my next thing? I mean, I had some cash, I had some runway, I had time to figure it out. And I just didn’t want to just jump right into something that I wasn’t totally … I’m not even passionate about, but just like, “Is this a long-term space I want to spend all my time obsessing about?” I just wasn’t getting there. And then on the business side too, it wasn’t like breakout growth. We didn’t have true product market fit with a category that we knew how to scale. And I think a lot of things went wrong there. It’s like you’re an app within Slack is your primary use case, so it’s a passive app in the background. You’re anchored to something that doesn’t monetize already, so now you’re a paid app on top of a freemium app. In terms of raising money too, we actually went through the cycle of raising capital. I don’t know how many investor calls we did, but a lot. And I think we closed 250K commitment on 100 investor calls or something, so terrible outcome there. One lesson there just from the VC standpoint was, that category was already funded. There was already a billion dollars or more of capital deployed against this virtual office thesis, no clear winners. We’re falling up against that just from, how is your business categorized from a fundraising standpoint? That was a no on fundraising, no real breakout traction, some struggles with team, and then, “Do I really want to do this,” is the core question. “Or do I want to spend more time figuring out what the thing I actually want to do is, and is there more upside there?” Yeah, I made the decision to shut down or sell. We found a great buyer for the business. Sideways deal, I mean, we actually got a decent amount of cash out and then just rolled that into the new co, which has been much better.

Allen: I’m curious to follow up on that idea, which I think comes up often when leaders, founders or other leaders are thinking about, “Okay, what’s next for me?” And obviously, if you’re at the point in your career where you have a lot of options, that’s great, but also can be really challenging when you’re like, “Okay, there’s all these different opportunities.” And if you’ve built a skill of seeing them, you’ll end up often kind of assessing and weighing businesses or problem spaces that might really personally excite you that you’re like, “Oh, I’m the person.” You’re saying to this, “Oh, productivity. Maybe I am a big productivity nerd, or maybe I’m really into golf,” or whatever the thing is for an individual that they’re super excited about and that they have this kind of urge to solve. But then there’s also this set of problems where it’s like the economics are pulling you. Where it’s like there is a real reason. There’s my market need and I’m well suited. I really understand the need well because of my previous experience. Or I have good connections and relationships in that space, or I just structurally see it’s an opportunity that’s really just happening right now. And a couple episodes ago we had Ian Crosby on the show who founded Bench. And he made a case for really indexing on the second, going first, really only considering situations where there’s a really strong economic driver of a business. And then basically finding the fun in that, rather than being like, okay, well obviously you can go to the extreme and do the futile thing where like, “Oh, but I just would really love to make finely crafted network or whatever, and I’m going to try and make a VC funded business out of it.” And it’s like, “Well, that just isn’t a thing.” How do you think about that as someone who has been, you just expressed the pain that it can generate when you are looking at, “Oh, is this a 10-year commitment to an idea that maybe I’m not super deeply personally invested in?” How do you think about the relative importance of those two pieces?

James: I think it’s this idea of founder market fit is really, really important, especially at that early stage. Founder market fit is not product market fit, but are you the right person to solve this particular problem? And how do you define that is variable, right? I think there’s this massive Venn diagram of like, “Do I care about this problem? Is it economically feasible? Is there a big market, is it fundable?” And all of those things need to be true in order to be really successful. But it’s like the missionary versus mercenary framework too, like, “Hey, if you’re a missionary, you deeply believe in this vision and you don’t care if it really works or not, but you’re willing to sacrifice everything to solve this problem.” Or you’re a mercenary, which is, “Hey, I’ve identified this economic opportunity, there’s this big gap in data analytics for enterprise software and we’re going to go rebuild Datadog or something.” And there’s no wrong answer. I think it’s like, “Are you good at that? Can you scale that? Do you have the skills to do that,” is an important question? And then on how I looked at it, I just really care about this problem and it is a big market. Luckily enough, it’s like an infinite TAM market I don’t think I would enjoy building, one, I don’t have experience building enterprisey software, more boring software. I think I picked the problem that I care most about and I found a way to reverse engineer the economics so they made sense. And you pivot a lot too. The idea has evolved, but the core thesis has always been there. And I think it really decides what you’re trying to optimize for. There’s lots of ways to win. I think there’s, especially I mean business generally, but in venture there’s no clear formula for success. You don’t know what your breakout hits are going to be. You don’t know how markets are going to shift. Typically, it’s the stuff that looks really weird or off thesis initially like Airbnb that ends up being the biggest, not typically, but often. Or you look at, hey, there’s some formula for building enterprise. We built PeopleSoft and then we go win and build Workday. It was slightly better software, a different sales motion, and then those do it again. Just for me personally, I can’t imagine another business that I would want to build. My eyes glaze over when I look at most software, and I’m just really excited. I love our customers. I think they’re just super entrepreneurial and fun to talk to. There’s so many problems within this space. More it’s just a fun problem space to be in, and the world’s moving so quickly. But I think there’s more risk to that too. We’re picking a hard business. It’s like a prosumer or close to consumer business. There’s economics around churn that are challenging, versus saying, “Hey, here’s a problem. Here’s a vertical, here’s a go-to market. We can build it. There’s no product risk and we’re just going to get after it.” That’s a great strategy too. They’re all good. There’s multiple ways to win. I think it just depends on what you’re trying to optimize for. And I think too, having that margin of safety financially, and some runway allowed me to just like, “Hey, I want to take the biggest swing possible for the biggest possible market outcome business, and see if I can play at that level.” But yeah, multiple ways to look at that.

Allen: There’s something in what you’re saying that I like though, which is, obviously you can’t reduce like, “Okay, what should I do next with the next 10 years of my career into a single sentence aphorism?” And certainly the approach that Ian advocates, which is, find the most profitable thing that you could, the most business value you could possibly return, and then use that as rocket fuel to the moon, which is not his wording. But that’s the way he thinks about it. But you articulated a neighboring potential strategy and potential playbook, which is, find something that you, a problem that you care a lot about and that you find very motivating, and then find articulation of that problem that has a very large market size.

James: And also, what do you have an unfair advantage in, right? I think with Ian in particular has, he’s the right person to build that business because he’s got that … I mean, 0.001% of people have the experience he has in that space. Intuitively you understand that you can go after it, you know how to scale it. But if someone else went and said, “Hey, accounting super profitable, or bookkeeping and we can automate it now,” 90 people are going to have that, 1,000 people. It’s kind of an obvious idea, but then it’s an execution problem. And can they execute? Do they have the relationships? Do they understand the nuances of? All that needs to be true. I think it really depends on this, again, this convergence of your skills experience and your learnings. And if you’re, I know a 19-year-old who just graduated, I don’t know, second year university or whatever, are you going to build that business? Probably not, right? You’re probably going to build something that’s got a higher variability. It probably won’t work your first view, but it might actually be huge and you end up building some crazy crypto gaming thing that’s way bigger than anything that I’ll ever build. I don’t get that stuff. But I do think the passion piece too, it’s not blind. You have to objectively look at every day of your business and whether this idea is still worth it. You see people that spend a decade on something that’s not working, and that’s not fun either, right? It’s certainly not an enjoyable place to be. And I think that it’s more, can you pivot within the problem space enough to see the success that you’re aiming for, or at least have a good shot at it? But yeah, I think all this stuff is complex. You can also just go, “Hey, I’m going to go copy a bunch of series A funded startups and just get a few customers,” because I know do it cheaper with AI or whatever. There’s lots of things. It’s more, what are you personally trying to solve for? Are you trying to solve for money? Are you trying to solve for freedom? Are you trying to solve for impact? Are you trying to solve for ego? I think just that self-reflection beforehand is important, because a lot of people don’t do it. You end up on this path that you’re just like, “Oh, I don’t actually like doing this and it’s challenging and I signed up for this and this seems odd.”

Allen: And that’s why I framed the whole question as something that people in the middle of their career or later in their career often actually learn to think about. Because they end up eight years in to trying to wait to IPO for some company that they have enough shares in that it’s a huge difference about whether or not they leave. And then it’s kind of like, “Ah, next time I’m going to think a bit about whether or not I care about the pet food industry or not before I attach myself to a long-term train.”

James: And also money traction growth solves a lot of that. You can pretend to care about pet food if you’re growing 100% month over month and you’re getting paid a bunch and your stock options, you can win a different way. You get passionate about creative brand, ad optimization. Liquid Death is a water company that just amazing marketing and branding, and that’s probably really fun to be in that business. But there’s slang in tap water. You can break down, you can be very passionate about a certain part of your business too. And I think that’s, again, there’s multiple ways to do this. There’s not that many things that humans need that are totally innovative and crazy. You find a different way to innovate on a problem and enjoy the output. So with Durable, I enjoy obviously the customers, the products, but hey, can we create really great education? Can we create really good marketing videos? Can we try stuff that’s more fun? So, parts of your job can still be really fun, even if the core overarching mission isn’t that exciting to you too.

Allen: Well, there’s something fundamentally fun about growing, or at least to me, maybe it is my bias, but I find there’s something fundamental fun about growing a team, growing a business, solving problems. It’s almost like in some ways like a video game when it’s working well, where it’s like, and now you’ve get to the next level and you have better stats, but also the enemies are harder now.

James: Yeah, exactly.

Allen: Now scaling is a problem and like, oh, now you have to learn how to hire a VP, instead of just knowing how to hire a line manager. And what does that mean? And how do you do that well? Once you’re moving it’s fun. I think often where a lot of companies really struggle is if they end up on a path where they don’t yet have the growth that’s making it compelling, and they haven’t really gotten also what’s driving that founder or that early team to push through to doing the really hard thing, to getting from the zero to one. Once you’re at one plus, then that’s more fun.

James: Yeah, if you’re going to build the diagram, it’s probably business you don’t care that much about, but growth, like okay, it’s not bad. Or it’s because you really care about no growth, that kind of sucks. But you might figure it out and then it’s like business you really care about plus growth is the ultimate. But you also have to earn the right to that too. VisualCV, I’ve never had a job and it was a resume building platform, so not a business that had a lot of founder market fit. I enjoyed the process of building a nice product for millions of people that generated lots of revenue, and it enabled me to, I was trying to solve for financial freedom with that business. That was my goal as a 21-year-old or whatever. It was like, “Hey, I want to be able to travel.” So I lived in Argentina for six months. We had a small team. I had lots of freedom and time, and that was the goal. And I did that. It was actually bigger than I thought it would be, which is awesome. Yeah, that was growth, but product I wasn’t super passionate about. But you got passionate about SEO, about paid marketing, about customer, so building the business part of it. And then I think for Durable, it takes the fun of the business building, but also the fun of the product that I really care about, so just fortunate to do that. And that’s just, you never know what’s going to work too. And I think you want to earn your right to the passion project too, to some degree. I think it’s, maybe lead with economic value first if you’re trying to solve for that. And then again, the crazy ideas can come later, but everyone’s different too. And you can never really reverse engineer these things, especially if the ones that get really big,

Allen: But it’s my job to at least try.

James: Yeah, exactly. Yeah.

Allen: Let’s shift gears from the … And probably did it backwards. But let’s shift gears from the big picture, like philosophy of and company stuff to some of the tactics, and actually what goes into some of this stuff. And I know you’ve spent some time talking to and advising some startups and seen not just the path of your businesses, but some others. Are there any things that you’ve seen either in some of the combination of experiences that you’ve seen that are tactics or approaches that don’t necessarily come naturally to teams, founders and leaders that you find yourself advocating for people to put more attention into, or to get more serious about that have come out of what you’ve seen?

James: Yeah, I think a lot of people end up doing a lot of fake work when they first start a project. And it’s almost just like they want this academic exercise of starting a business. It’s the, “Hey, what’s the right answer to that?” Probably what we’re doing now, what’s the right answer here? And it’s that kind of, especially for type A people that are really successful in school and work, it’s like, “Hey, there needs to be a right answer here. What should I do? I just want to do that.” So you end up like, “Oh, I should,” I don’t know, “optimize my ads, optimize my website. I should,” I don’t know, “make sure companies form.” You have a lot of this extraneous work, but what really matters is, are you building an amazing product and do you have customers that really like it? So, just getting down to that specific problem is actually surprising how easy it is to forget when there’s all this other stuff going on, because it’s all new and you’re trying to figure all this out. So, it’s like the majoring and minor things, like I am running some non-materially amount of paid ads to my landing page to try to optimize my conversion rate. It’s like you have no data, you can’t get data from that, right? You’re just trying to smoke test it to prove to yourself it’s a good idea. I think it’s that truly, go manually recruit your first batch of customers, work really, really closely with them, and be in constant communication with them to solve a real problem, and talk to them as much as you possibly can. And build something really great too. I think the bar for products is a lot higher than it used to be. It depends on the market, but you can’t launch MVPs anymore. They have to be fully baked.

Allen: Or at least the V has gotten more strict. You still could want to launch a minimum viable product, but what’s viable is now way higher.

James: Yeah, I call it a minimum valuable product. So, is this actually useful for somebody? So that’s, one is just like, “Hey, are you actually focusing on the right things and have you scheduled your time and your data to understand what those are?” And typically it’s just write code, talk to users. I think YC has got good advice there. But I would also say spending more time on the, is this a good business, before you actually write any code? So that zero to one, or the negative one to zero stage. I mean, my first investors HealthPark Commons, they have a program called the Founder Fellowship that, I mean, they write you a 500K cheque. They say, “Hey, spend as much time as you want to figure out what the actual business is.” Spending more time in that customer validation, it’s talking to people, it’s like coming up with ideas, it’s testing them, it’s launching them. Just take that period, spend more time in that period and try and prove stuff out faster. I mean, don’t get totally married to your idea. I think that’s truly the most important phase of a new startup is, what are you actually going to do and is there market demand for this? I think with that program, I think it was 15 people in that cohort. But just seeing the iteration there, what we thought were great ideas a month later, “Hey, no one actually cared about this.” And we talked to 50 customers. We actually built something we tested. Getting really good at rapidly iterating on ideas, then actually being able to just kill them. We saw teams that had ideas that seemed great, didn’t really work, then they had another idea, and then two years later they’ve raised $50 million and they’re growing. Really, it’s so interesting how the right thing at the right time just grows so much quicker than the non-right thing at the right time. So more time developing that muscle of what’s actually a good idea and what’s going to grow and how those product market customers is their tailwinds to the market. Getting good at that exercise I think is infinitely valuable. And that’s also some of the harder thing to do, because it doesn’t feel productive sometimes too. Like, “Hey, I should be building something.” But if it’s the wrong thing, you can spend a lot of time in that kind of going down the wrong path.

Allen: And it’s also one of the challenges, I’ve been through that a bunch running for many years of services business where we were building products for clients, but then also building our own experimental products like, “Oh, maybe this can be its own business, maybe this can be its own business.” And it can be difficult for the morale of the team if you’re like the founder who’s like, “Yeah, no, kill it.” “Okay, yeah, okay, no, but this is really good. Let’s build this really well. I’m really fast.” “Oh, I know, kill it.” And you’re doing that over and over again, let alone your own morale. But that’s part of the secret, I guess, is to be able to keep doing that.

James: And everything that you do, I think it’s the realization that if you choose to do it, it’s going to be … Going to take five times longer, be 100 times harder than you think it is from that early stage. It’s like, “Oh, this is a great idea. We’ve got four people really like it.” And then a year later like, “Holy shit, that was a lot of work and now seven people like it.” It’s just that it’s just harder. And I think the overhead, you have to learn how to … Again, you go-to market, your marketing, your brand, all those things have to come together. And I think in that stage too, it’s like having a small team, if not just the co-founders is the right size for that zero to one. We’re trying to figure out what to do phase or negative one to zero-

Allen: Partly because it’s so much easier to turn the ship when the ship is deliver to the two of you and are both fully bought in and then you and your co-founder, maybe there’s six or four of you, it’s way easier to be in a room and be like, “All right, this isn’t working. Okay, let’s get excited about the next thing.” Then it’s like, “Okay, well, how are we going to break it to our ninth hire or a 15th person or whatever, who their entire career was in this market, which we’ve now decided that we’re not going to be in.”

James: Exactly, right? Super tough.

Allen: Pulling on the thread of timing and the right times for things can make such a huge difference. I want to ask a little bit about what you’ve been learning building on LLMs. You mentioned that you’d started very early on, which obviously the capabilities from GPT-3 to 3.5 to 4, and now supposedly five is coming before the end of the year, is changing the underlying capabilities and the market and the relative competition and the ability to build things at such a pace that it’s really fascinating to me. And that’s something that I’ve been talking to a lot of people about recently. And I’d be curious to learn from you. What are some of the things that you’ve seen building tool so durable? Obviously maybe not obviously for folks who haven’t seen the product yet. But started as this website builder tool where LLMs could create something that, building something where you generate a website that’s actually coherent and relevant to the person who’s trying to get this business started two years ago is not a thing that you could just build, and then now it is a thing that you can build. But what have you been learning is now one of the teams, which there isn’t that many yet, that have kind of scale, you’ve had millions of websites created using these tools. What have you been learning about that transition from, “It’s maybe easy to make a demo harder to scale product built on these LLMs,” and what are the of early lessons learned in really harnessing these tools?

James: When we first launched our … The goal … We had a website product before that we were again toying around with LLMs early or mid 2022. Is that, yeah. Oh man, that was a busy year last year. And then, yeah, the timelines always get blurred in startups. And we had a website probably like, “Hey, what we learned from our customers?” Like, “Hey, no one can actually, as simple as our product was, they can’t complete their website.” They don’t have the time, they can’t figure it out. Then we thought, “Hey, how can AI build your website for you? Can LLMs now select your colors, choose the template, write the copy, just make something a lot faster than what the customer would do?” We’re like, “Yeah, let’s build that and see if it goes.” And this was October 2022. OpenAI was still in beta. We built this product, we shared with a few people and instantly there’s like, “Holy shit.” There was that real visceral moment we saw in-person from people. We went to people in coffee shops, showed them on their phone. And they’re like, their eyes light up. They’re like, “Oh, John, come over. Look at this.” I think it was just like, “Okay, we might have something here.” We launched it. I think we had a thousand sites built our first day. I’m like, okay, there’s something here. And then all of a sudden we have 10,000 sites, then 100,000 sites, and then about a month and a half later and one, we hit our beta quotas very quickly. We’re begging Sam Altman to up our limits. We’re like, “Oh shit.” Weren’t ready for this much growth. And then ChatGPT launched about a month and a half later. And that’s when the world just totally switched to AI is everything, AI first. And we were before, we basically got on the wave right before it broke. We were able to ride that tailwind. And then business has just grown exponentially since then. So, very good timing. I think on the product specific side, I look at AI as this massive enabler of workflow. Previous and current versions of software, the customer or the user still has to do the work. You have a spreadsheet, you got into the data, you have a website, you have to choose the pictures. But with AI, you now have like a human-enabled intelligence, hopefully more than a human at some point, that’s the best web designer in the world that now works for you 24/7. They’re optimizing your website, they’re optimizing your SEO, they’re helping you pick design patterns, and then you take that concept and bring it to marketing. Okay, now you’ve got an amazing marketer that works for you. You’ve got an amazing bookkeeper, accountant. How do you just bring this human level or better intelligence to an individual who could not afford those people, and give them leverage? As a solo business, now you have the power of 10 employees working for you 24/7 to optimize your business. That’s really where we want to go to. We’re certainly not there in the path yet. We’re building some stuff that’s getting reasonably close to the basics of those workflows. But that’s how we think about it is, “Hey, we’ve got this super powerful system that to some degree negates the purpose of a UI. If our customers should … If they don’t want to, they should never have to log into the product. They should get a text message once a day that’s like, “Hey, you’ve got four jobs booked tomorrow. We sent four invoices for you, and your site got 20 leads in the last week.” That’s a super cool user experience. That’s like, Hey, you never have to do anything besides show up and do the work and get paid for it. That’s really where we want to go with the product. And I think, yeah, it’s been certainly a wild time to build on a technology that’s evolving so quickly. I think it’s the like, oh, is GDB-4 going to kill a bunch of businesses? Is GDB-5 going to sell when video comes out? Okay, all these AI video companies are going to have a hard time. But I think with us in terms of differentiation and longevity and durability, it’s this, “Hey, we know who our customers are. We know there’s billions of them, and we’re not trying to build base level models right now.” We’re not competing with OpenAI, we’re not competing on the core frameworks. We’re trying to apply these models, optimize them, build really specific workflows for a customer base in the most effective way possible. Really how have we become a reasonably deep technology companies? We need to do that, but also how do we apply frontier technology to real people in the most effective way possible? And again, a broad enough problem space to work with I think is really important, and a big vision.

Allen: And as we talked about a bit, you’ve set yourself, I think up in what seems in a really good way in terms of like, okay, you have this problem space. Obviously, I say obviously, you had to actually make the case, but obviously potentially a billion-dollar business, you’re addressing something that’s vaguely if you squinted at Shopify sized in what you’re trying to do. And in some ways there’s that perfect fit in between like, “Okay, you’re trying to solve a problem that’s really big and there’s a whole bunch of different ways you could tackle it, so let’s all go attack this problem.” The business side of it makes sense. From that actual implementation side of as you’ve been building so far, have there been any surprises or things around the actual nuts and bolts of building on this stuff that have changed your way of thinking about some of these tools in their current state? Or is it still just like, “Ah, there’s so many problems at hand.” And there’s so much new stuff happening that you’re just sort of in the process of harnessing it rather than like, “Oh, we thought that this, but it turned out that that path is kind of a bit of a dead end for now and we’ve been more prioritizing and building things with this way in terms of either workflows or things like that?”

James: Yeah, I think tactically the tooling is still very new. With cloud, I don’t know, there’s multiple products. There’s been, I don’t know, 10 or 15, probably 20 years of R&D there. The costs have gone down. I think with LLMs, you’re still building off of brand new tooling. We had to build our own hosting, analytics, latency is a big issue. Kind of the same scaling issues you see with every technology. LLMs have the same thing. Understanding what queries fail, why do they take so long? Is Azure better than OpenAI for hosting those models? Do we experiment with open source? We’ve got this framework now where we can really quickly test different models, fall back to. Building the infra I think was harder than … Not harder than you expect, but quite challenging to build proper infra at our scale. Just from a product standpoint there. And then, how do we apply this stuff as it changes really quickly? OpenAI was launching a video model, can we leverage that to create amazing video for our customers? Or voice is really exciting, so can you automatically answer the phone for everybody and schedule their jobs for them? And how can we build that? Is the tech ready too? Because a lot of this stuff, it feels really great in a demo. And then when it comes to brass tacks, like, “Yeah, it’s not quite there.” There’s some latency. Or you can’t build the proper workflows yet. It’s really this iteration towards is this ready for production? And can we win the market in our category for that specific product category, is one way we look at it. And then on the customer learning side, what’s been surprising is how much they trust the AI already.

Allen: Interesting.

James: And trust in the way that they just want the problem solved for them. They’re like, “Oh, this copy is really good. I don’t know if I could write better copy.” Or some iterations of that, or this web design is really good. It’s really not a question of how do we get more people to use our products? How do we solve the core problem in a much faster, better way? With websites, it’s not, “Hey, we got to make our website editor easier. It’s like, no, the website output has to be better. We have to be optimizing that site. We have to be getting the preferences from the customer in a different way. Maybe that’s a phone call, a conversation, like a survey. But just take that data and apply it as opposed to, again, expecting the user to go and do the work. That’s been surprising how much they just want to be solved for them. And how much they actually, not, trust is the wrong word, it’s just believe in the technology. Being able to solve that already, I think is really interesting. And there’s a ton of opportunity there.

Allen: Do you think of, and this is just sort of a broad product development question for those of us that are building and exploring, what can be built with these tools? Do you think of the chat interface as the default go to, okay, we’re going to evolve these products and what they can do, and obviously if you’re generating a website and then you’re going to want to refine the website. Does your mind and your exploration so far lean towards, “Well, we’re going to have people refine that website by talking to a chatbot,” where you’re like, “Oh, can you zhoosh it up or can you make it sound more corporate?” Or can we talk more about my free trial or whatever the things are that … Or are you finding at least with the level of the tool so far that more kind of bespoke UX workflows that have a little bit more structure to them, is kind of the more powerful approach?

James: Yes. Still, TBD, I’m not long-term bullish on chat as the core UX for a lot of this stuff. Primarily because for chat to work you need to know what to ask. And for a lot of our customers, this is not their core expertise. They don’t know what SEO is really. They’re not going to ask like, “Hey, how do I optimize my SEO?” I think we need to set the bar higher for being the expert here. The way we break it down is reactive and proactive. Reactive workflows would be, “Hey, something’s not working, or I want to do this.” They’re actually asking the question, and that’s kind of a chat or a question response. Like, “Hey, I want to add a new page.” “Okay, sweet, what’s the page? Sweet. It’s added.” But proactive would be, “Hey, we’ve identified that you don’t have individual pages for your services. If you want to optimize for SEO in Vancouver for plumbing, you should have individual pages. We’ve already built these pages, we’ve already optimized. Do you want to press go?” And that’s a yes no answer. I think that’s a mind blowingly cool experience for someone is like, “Hey, wow, this is actually, I wouldn’t have thought of that. It’s smarter than me at SEO, obviously.” And it’s just this nexus of things. If you imagine, well, let’s talk about website just, it’s obviously our core product right now. But there’s 10 things that every amazing website has. We feed the model, those 10 things, and then we need to figure out how to get the data to do those 10 things for the customer. And I think it’s not necessarily a chat. It’s like we send them a text message at the right time asking for photos. Realize they don’t have a logo, so we send them 12 logo options, and they can pick the right one. It’s just more like … And then what order of those things too. Can the LLM actually figure out … Okay like people if they like photos more? It’s easier to get photos than it is to get a logo, so we ask for the photos first, and they’re really excited about that. It’s like, “How precise do you have to model out and schedule these workflows?” There’s ways of doing that are very traditional. If no photos, then send texts for photos. The way we’re trying to design that is more, “Hey, AI web designer, here’s the things that we need to get from our customers in a way that doesn’t annoy them in the most effectively possible. Can you figure it out?” Which is a different and a much harder problem, but it’s pretty cool, right? You’re learning what language works and the end goal for … I mean, websites again, is this idea of what is the perfect website? And how do we iterate towards that on every data point that we have? That’s both output, it’s customer input, it’s customer feedback, it’s conversion optimization. Can you build a model that’s self-perpetuating there, and then start feeding that model in a way that’s really effective? That’s hard. It’s not the way that’s actually going to drive the most customer value immediately. But that’s the long-term R&D that I think if you can apply that concept to every product within our platform, then it just gets better and better over time, and it’s self-perpetuating. And then as these models get smarter, those problems get easier to solve. GDB-4 is, it’s okay at analyzing websites and providing feedback. It’s probably a 12-year-old designing websites like, “Oh, this is cool.” But GDB-5-

Allen: Kind of like a 12-year-old that’s read a lot of books.

James: Exactly. But GDB-5 might be amazing. It might just be, oh wow, it’s actually providing amazing feedback on conversion optimization. And we can take that and optimize our sites with that. I think it’s just being ready for the next spikes and how stuff is going to evolve. I think that’s one way we’re looking at it. Yeah, it’s going to be okay right now. But the next model is going to be like, “Oh, wow, this workflow that we thought was not solvable or not that great.” Is like, “Wow, this is done.” We’ve figured out how to automatically optimize marketing campaigns for plumbers in Idaho. Now, you can’t do that right now with GDB-4, but maybe five solves that when the agent model is more effective and it’s smart enough to know, “Hey, you should go post on Craigslist and Nextdoor and Thumbtack.” All this stuff just, I think it gets really cool and it’s going to get even cooler very quickly.

Allen: It’s interesting that you frame it that way, because where my mind goes, not having immersed myself into this business the way that you have. Is that, well, I need to become a world expert in how to optimize this Craigslist ads for a plumber in Idaho, so that I can build some sort of hybrid prompting rag, as they call it. Prepping the agent so that it has the context that understands that. And then also some UI that coaxes and encourages the customers that like, “Hey, it’s worth actually doing this.” Help them. Obviously, the plumber is not going to become a world expert in this, but help them understand, “Hey, if you get a sense of how you differentiate yourself from the other plumbers, that will help you a lot. Keep an eye on that and then we’ll do this other piece.” Whereas your mind is, at least sounds like, less focused on, how do we do the customer education, and maybe it’s a little bit mix, just because it’s a bit of a boil the ocean problem. Because how to make the best Craigslist ad for a plumber in Ohio is totally different than how to make the best slogan, front page hero photo for a home chef in Vancouver. I don’t know. I find that interesting.

James: Yeah, I think the core, and I think it’s specifically with our customer base, right? It’s like if you are good at that job, right? We’re constantly building a profile of like, “Hey, how can we just, so again, you’re the plumber in Idaho, how do you think about differentiation?” Do you have really good Google reviews? Are we pulling data from the Google reviews? Like, oh, they’re always reliable, they’re always on time, they’re really friendly, they do a great job. They’re really good at a specific thing. So you’re building this understanding of your customer, that allows you to customize more and more as it goes. You’re basically building this model again. And I think the metaphor that’s helpful for us is if this person was to hire for these functions within their business, what would they want to know? You’d want to know, obviously what you’re good at, what your differentiation is. But the thing with services specifically, it’s kind of like, are you good at this? And do you show up? It’s not E-commerce where there’s a ton of complexity around inventory, supply chain. It is a business where if there is market demand and you can deliver that skill, we can help you with the rest of those things. And the bar’s pretty low. Not for service quality, but for marketing, for follow up, for answering your phone. AI can do that a lot better than these people who are super strapped and going between jobs trying to answer. Text message, it’s a really great solution for this category of business. Primarily because they don’t have the resources to hire the individuals, and they don’t have the time to figure out the tooling. This is where AI, I think really shines is, “Can we be better than our customers at those specific functions, so they can focus on what they’re really great at?” And then over time, we build data sets and profiles that again, like, “Hey, we learn more about how you want to communicate with your customers.” Your voice and tone, or your style, and then that applies to the entire platform. So your ads look a little different, your copy looks a little different, your emails look a little different, your text messages look a little different and just gets way more personal over time. I think that’s where it gets very exciting and very self-fulfilling in a way that’s going to be pretty sweet.

Allen: Well, I love the idea of a world where all the various plumbers and service providers that I engage with over the course of the year, have an assistant that actually answers when it’s like, are you on your way? Or-

James: A hundred percent.

Allen: … are we on for this? Or how much is this going to cost? And all that kind of stuff.

James: Yeah.

Allen: I’m looking forward to seeing how this builds out. Thanks so much for coming on the show, James. Where can people go to find more about you and the work that you’re doing?

James: Twitter probably, @jamescliff, unfortunately my most active platform.

Allen: All right, so we’ll link that up. And of course, Durable as well in the show notes. It Shipped That Way is brought to you by Steamclock Software. If you’re a growing business and your customers need a really nice mobile app, get in touch with Steamclock. That’s it for today. You can give us feedback, follow us on social media, or rate the show by going to Itshipped.fm/contact. Until next time, keep shipping.