I started the company when I was 24. 14 years in, I still feel like I'm pretty young. The thing is, the world doesn't stay static. The world changes around you. People study how you work. They pay attention to what works, what doesn't. They develop new techniques because they're trying to beat you. We're going to be the automation company. That means AI is going to be priorities one, two, and three for the foreseeable future.
Hello and welcome to Between Two COOs. I'm your host, Michael Kanig, and today I'm joined by Wade Foster, co-founder and CEO of Zapier. Wade has been building the same company for 14 years, and that alone makes him an outlier. Most tech startups don't make it past year five, and fewer still grow into a multi-billion dollar business and have taken as little funding as Zapier has, and even fewer do it without losing their identity along the way. Zapier is best known as the workflow automation layer behind modern work. Millions of people use it to connect thousands of tools and those workflows now power billions of actions every month. But what interests me most isn't just what Zapier does. It's how the company has managed to last. Over the years, Wade has led Zapier through multiple technology waves, market shifts, and internal evolutions. AI is the latest one and a big one, but it's not the first stress test. What makes this conversation interesting is how durable companies absorb change without thrashing, breaking trust, or losing coherence. So, this isn't a conversation necessarily about hype or tools or predictions. It's about what Wade has learned by staying in the seat for 14 years. What had to change along the way, what didn't, and how that shows up today in the way that Zapier thinks about workflows, automation, and AI. Wade, thanks for joining me.
Thanks for having me, Michael. At what point did you stop feeling like there were startup problems and start feeling like a long-term stewardship problem had to be addressed?
Well, I'm not sure I've ever felt like the startup problems disappeared. You know, we're still as hungry as ever for making the product better, getting new customers, like that I don't know that that ever goes away. You know, something clicked for me, maybe in the pandemic, partially food, maybe it was because I started having kids. I'm not sure where I started to think about decisions over a much longer time horizon and saying, 'Yeah, I'll tell you another thing that helps. I'm a huge fan of this podcast, Acquired. I don't know if you listen to these guys. Yeah.' And so you just started to listen to those episodes and they do deep dives on just these enduring businesses and there's just some common themes around how principled those companies are run. And I think I was always maybe predisposed to thinking that way when most people are just so unwilling to do so. They're trying to make a quick buck, they're trying to flip the thing, they get tired, they get bored, they get distracted. Like there's just a lot of things that cause people to not go long. And I sort of looked at where I was at and thinking, we're in a really unique spot because I started the company when I was 24. 14 years in, I still feel like I'm pretty young and another 14 years is not insane. That's kind of normal. And it felt like we ought to start looking at certain decisions and thinking about it that way. I can give you a really concrete one. We made a big pricing and packaging change in January of 2024. Very customer-friendly. We gave tons of revenue back to our customers because we just sort of looked at it and said, if we keep going about this the way we're going, maybe we make more money in the short term, but over the next decade, I felt like it would compound a lot better if we gave a lot more value to our customers. We'd earned more market share. And 10 years out from now, we'd be in better shape. Maybe the next two years might not look as good, but I was okay with that because I'm making a decision for 10 years down the line. Before we get into it, follow Between Two Coos in your podcast app right now. Every episode lands in your feed when they drop. The next few are worth catching. We've got the former chief operating officer of Grinder, who took the company public at a $6 billion valuation. Then we have the chief operating officer of Cowboy Space who's building data centers in space and the rockets to get them there. And we also have the COO of Katana, the platform behind more than $3 billion in transactions for SMBs selling physical goods. These are real operators building incredible companies talking about how they actually run things. Hit follow so you don't miss them.
And I'm curious, you've made a lot of decisions and maybe my funding information is incorrect, but you have not raised significant venture capital.
Yeah, we did. We went through YC and then post YC we raised a million dollar and change seed round, and that's the only primary capital that's been raised.
Which is absolutely crazy. And so there's a lot of control still that you and your co-founder have, I imagine. How much of that longer term thinking do you think is a result of having that ability to not necessarily optimize for the next two years? Oh, I need you guys to IPO. I need you to exit. Things like that. Has that played into your ability to think about having an enduring company?
I mean, I think it has. You know, obviously there's plenty of venture-backed companies, even public companies that have been able to think long term, but I also know tons of peers that are in positions where they're having to make short-term decisions. They're trying to get the next quarter. They're trying to get to the next fundraising mile. They're trying to get to certain things. And you can see them struggling with certain product decisions or certain sales or marketing decisions or certain pricing and packaging. Packaging is always one of these where you're like, I can make more money just by raising prices on existing customers and doing some things that maybe don't feel crazy customer friendly and I'll make more money, but is it going to be the right thing in the long haul? You can see everyone struggles with that. Everyone struggles with those decisions. I mean, even we struggle with them. But I just think because we don't have that same board member or venture person breathing over our shoulders saying, 'Hey, you got to do it,' we probably have a little more fortitude to say, 'Hey, let's play. Let's go long here.'
Yeah, I love that. Now, founders often talk about those near-death moments with fundraising scares or traction cliffs. Was there ever a moment where Zapier wasn't necessarily about to die financially, but you felt like it might lose its way?
I mean, right now, I think is a really interesting time, honestly. You know, we called a Code Red in spring of 2023 when GPT4 came out in part because of the delta in improvement between 3.5 and 4 was so big and it was only six months between those two releases. So we looked at that trend line, two trend lines I guess, and said, 'Wow, if this is where the world is heading, there's an opportunity here and there's a risk here.' And so we wanted to just jump on it. We called the code red, had the company, paused everything in the company for a week, ran a hackathon, and said, 'Hey, everybody needs to be building with these APIs. They need to be playing with these tools. They need to be understanding the art of possible here because I think this is what the next chapter of automation looks like. And if we're going to be the automation company, that means AI is going to be priorities one, two, and three for the foreseeable future.' So that was one time where it was like, what's this going to look like? And honestly, I still think we figured some things out, but I don't know that anyone is out there feeling like they figured it all out. I mean, you even look at the big dogs, you know, the Open AIs, Anthropics, the Googles of the world. It feels like they're all just passing the baton around on who's in the lead, who's doing the best stuff. One moment you think it's OpenAI, then it's Google, then it's Anthropic, and you're just like you're not really sure who's the bell of the ball at this moment.
Yeah, so interesting. And especially I'm sure there were signals, there were other moments within the company and building where you had all of these different tech moments happen, mobile, etc., really basic ones that you have weathered. What feels different because from your perspective and vantage point of having that endurance, what feels different about this?
Well, I think automation like AI just is a new way to automate and so it just hits at the core mission and premise of the company in a way that, you know, say mobile didn't as much. With the mobile wave, we were so young that it just didn't much matter one way or the other. Most of the automation work we were doing was for white collar knowledge workers who were mostly sitting in front of desktops and laptops and just logging into SaaS products and things like that. So mobile was there and people did stuff, but it was just not the predominant form factor of how white collar knowledge workers were automating their day to day. Whereas with AI, this is just an entirely new way that you could deliver automation, and so that hits at our doorstep in a way that maybe some of these other shifts haven't been.
Well, with a lot of those other shifts too, you end up having to say no to a lot of things. Has there been anything that you have said no to with AI?
I mean, there's definitely things that we haven't done. I think our philosophy at the outset of this was try everything. We don't actually know what we're doing. There's so much to learn. There's so much to figure out. People can build stuff so quickly now. And I think that we get critiqued about that sometimes internally. We're like, what is the thing that actually is going to be the one? But I do think we're just at a moment where you need to have this experimental mindset. You need to have this beginner's mind. And I don't feel like we're at this waypoint where it's like, we 1,000% know the winner. Say no to everything else, focus on this one thing, and milk it for all it's worth. I just don't know that we're at that part of the story yet.
Yeah. And I think as I was prepping for this, I'd listened to a bunch of your podcast and in one you said, and I'll have to drop it in the notes or something, but you said we still don't have God in a box yet, which I loved. But you struck on something interesting, which is the workplace headspace, the growth mindset. Now, a lot of examples that we talk about come from tech companies that already live in software and workflows. A lot of our listeners aren't necessarily within tech itself. So for companies that aren't necessarily tech-native, maybe older organizations, industrial companies, professional services, what's your advice for getting AI into the workplace headspace before rolling out tools or automation? Because you can't necessarily have a hackathon. What does a hackathon even mean in some of those companies?
Well, yeah. I mean, obviously, if you're doing sort of white collar work, digital work, etc., AI is going to be so critical. If you're moving steel beams around or in the physical world, AI is going to be a part of that, but it's going to be more of a supporting role versus maybe the central cast. I still do think that if you're in one of these organizations, you need to be the type of leader that's figuring out what's possible with this technology. You need to put your hands on it, you need to figure it out yourself because if done well, I do think you can use it to disrupt old school industries. And if you do that right, you can eliminate waste from that business, you can improve your margin profile, you can lower costs for customers. Those are all things that can disrupt your business. In fact, there's like a playbook in Silicon Valley that is starting to pop up, which is you take these really smart AI-pilled founders and you pair them up with a domain expert who knows an old school industry like the back of your hand, and that founding pair can come in and create the new version of the premier company in this old school space and deliver it at a much better experience. And so if you're in one of those industries, I think you want to do that yourself instead of letting somebody else go build the new company.
Now, hopping out of AI, and we'll come back to it, but hopping out of AI, I want to return to your learnings over these last 14 years. You know, we talk a lot about these zero to one and hypergrowth companies. Most of the company building though lives in that long middle. What problem took you the longest to realize was actually maybe stifling your growth?
So, one of the things I wish we would have done sooner is that we would have focused more on mid-market enterprise. I think we had an opportunity probably 2017, 2018 where we could have been very aggressive to go chase that and we didn't do it. We stayed kind of in our PLG group roots, self-serve, much more of that kind of consumerization of business software approach. And it took us until probably 2020, 2021 to sort of wake up and say, 'Oh wow, there's a big opportunity here. We're kind of sleeping. We're being a little dogmatic, honestly being chicken about the opportunity here.' Let's go get after it. We just lost valuable time by not saying, 'Hey, let's go get after it.'
So, that's one that comes to mind. That's a difficult decision and transition to make, especially when you've been doing something for so long. How do you go about changing that mindset, that behavior of well, let's not be chicken? And just out of curiosity, why were you chicken? What was holding you back?
Well, I think there's a certain comfort in the business. I'm good at this. It's familiar. A good analogy would be Bill Belichick and the Patriots. Iconic coach, arguably the best of all time, but by the time he left the Patriots, he hadn't updated his play style and he wasn't winning as much. Now you see him at North Carolina his first year. I think this is something all leaders struggle with. You come up in an era and there are things you're good at and you know you can win this way. But the world doesn't stay static. People study how you work, pay attention to what works, develop new techniques because they're trying to beat you. If you're fortunate, you can do that for a decade, but eventually there's an expiration period on how you go about that work. So it's important for a leader to always be looking for an edge, continue innovating, thinking about the next act. How did we learn to stop being chicken? It was just recognizing that fact and going, there is a huge opportunity and we're being comfortable. Why are we being comfortable? Let's go push and challenge ourselves. It's more fun to work that way. So let's get out there and do that. Over those 14 years, as a CEO, as an executive, you have grown and matured. I'm sure you've had incredible mentors and learned a lot of things the hard way. What parts of early Wade had to go away for Zapier to keep scaling? The first one I can remember is I started the company when I was 24, young enough to think I'm smarter than everybody. But starting a company has a way of making you realize you're not that smart because you make so many choices and you're wrong so many times. The best way to be right a lot is when you're wrong, just change your mind. So there was a humbling I had to go through to realize that my job is not to be the smartest person in the room. My job is to figure out what are the smart ideas and put more arrows behind them. So that was one of the first things I had to learn: to be a lot more open-minded about what I know and what I don't know. So it sounds like also there's a certain degree of judgment that develops over time to be able to look at those ideas that maybe weren't great and change them. That is a profound shift. There's a certain part of judgment that comes with knowledge and experience and allows you to evaluate which ideas to pursue and also know when to recognize when something's not working. What does that look like in your world?
I think the more you do this, you have the benefit of pattern matching. You've seen a lot of different stuff. You can pull ideas from different industries and experiences. It gives you a way of thinking about the problem in a particular way. That can be a valuable asset because you can draw on that experience. But you also have to be careful because that experience can fool you when the playing field shifts. Think about the age of AI. Maybe the way teams work is totally different now. Am I running a playbook that is dated? You have to constantly keep yourself in check. It is helpful to spend time with younger founders, folks earlier in their career, to see what they're doing that's different. They may not have all the experience, but they come from a different environment and a different way of working that feels native to them. I think about when we started, we were building cloud integrations and would get quizzical looks about why we didn't have more on-prem integrations. We thought no one works that way, but clearly many people do. That shows that young folks have a view on the world that seems so obvious, and I want to understand that better. So you have to figure out how to use your experience and pattern matching well, but also continually update your priors and get exposure to things you may be discounting. Before we get going, a quick note for listeners. This episode gets into a lot of the stuff I'm writing about in the Between Two COOs newsletter. The podcast is mostly me interviewing people like Wade and pulling on how they think. The newsletter is where I'll add the practical layer around it. Guest takeaways, operating patterns, and some of my own notes from the COO seat, especially with conversations like this where we're talking about long-term company building, automation, AI, and how operators decide what actually deserves attention. So go to between2cco.com or b2co.com and subscribe. All right. One of the interesting things that we saw and we're going to shift gears a little bit here. One of the interesting things that I saw in terms of the AI transformation, and let's kind of move past the buzzword, but nonetheless, AI sits within the chief people officer role for you now. So it's the chief people and AI officer. Tell me about that. Most of the time we see it sitting with tech. Sometimes we see it sitting with operations people. Tell me about that decision.
Sure. So yeah, Brandon took that role for us at the tail end of last year. He's our chief people officer now and also AI transformation. We started this journey in spring of 2023. So it took us a while before we said we want to put a person in that role. Even early on there were suggestions maybe we need a centralized person, but we chose not to. In hindsight, that was correct. Had we named somebody early on, we probably would not have done as good a job of making sure the whole company understood they have to figure out AI. When you name a person, it becomes that person's job. Early on it was important to recognize that all of you have to figure this out. We need to raise the floor for everybody. This can't be just one person's job. All of your jobs are changing. By running hackathons, doing show and tell, having people come to all hands and show what they're building, we built a culture where everybody was chipping in and getting excited. First slowly, then as the flywheel started to spend more energy, it got to a point where it was clear everybody is stoked about this. No one is against AI. But the problem became different: we have all these ideas but some of the bigger ones involve org change, role changes, complex problems. As an individual IC, I can't just decide to redo how hiring happens or the entire customer journey. It's not a one-person job. At that time, Brandon had done such a good job reskilling the people team and using AI. It felt obvious we need somebody to help wrangle the change management, the logistics, the organization around AI. It wasn't a technology problem; it was an organization and operationalization problem, people psychology issues. Brandon had already done a great job with the people.
Team. Why not just ask him to do that for everybody else? And so that's kind of how we got there and why it made sense for us to go that way. I think if you have a similar setup as us, that could be the right choice. But if you have different problems, it might not be that your head of people should be the one doing this. It might be that there's somebody else in the company. I think every company is kind of going through their own journey on this stuff and they have to think through what are the bottlenecks they're facing and how to think through how to go solve those.
I've actually seen it now within some other areas. Mainly because I'm starting to see HR teams be much more experimental with different software and options that are out there versus other teams, for instance.
It kind of makes sense. There are very real problems that are showing up on HR's doorstep that are directly related to AI. The most obvious is recruiting. Recruiting functions everywhere are having a heck of a time right now because they're just getting AI slop as applicants.
Like so much of that stuff. And I understand how the situation got this way because as a candidate, you're just so used to applying for jobs and not getting responses. So you're like, well, I'll just use AI to help me apply to more jobs, increase my surface area. And so the whole system just kind of gets borked and you're like, 'Well, what the heck are we going to do about this?' There are very real problems that are on the recruiting teams right front and center for them where they're going, 'We have to figure this out. This is not a luxury that we can just go figure out how to dabble and mess around with AI. We have no choice but to think through a new way to solve this problem.' And so I do think that's partially why HR teams are getting into the game. It's like they have no choice but to get in the game.
Yeah. So interesting. Let's talk about problems to solve. How do you decide which problems actually deserve AI versus those that are better solved with just straight automation or better process design? How should companies, how should operators think about that?
It's a good question. So my learning over the last couple years is that if you can solve the problem with plain old deterministic software, you probably should. The reason why is you're just going to get better reliability, lower costs. For anything that's even remotely mission critical at all, it's probably just a better way to do things. That said, there's a lot of problems that you can't solve that way. And so that's where it gets really interesting. I think a mental model that we had, especially early on, but it's still true today, is we often just looked back at our 10 years of running the company and said, what problems have we always faced that we've just never been able to solve? We just never been able to crack this nut. And can we solve it now? Is this actually possible because we have a new tool? That's been a very helpful mental model, looking backwards. Now, as we've gone forward, you start to recognize places all over the place. You just look at how much of your workday involves working with unstructured data. And that's where AI really thrives. You're dealing with big blobs of text and how can you go wrangle that into a workflow, into a process. And then I often find you're starting to build out these hybrid workflows where you've got AI as a piece of the solution that's tied up with deterministic software or code or something like that that is running a very particular way.
Yeah, that's interesting. And you've been very clear about workflows and I'm using agents in the marketing term sense which is I think...
Yeah. Everything's an agent right now. They're really like assistants, right? But nonetheless, I've stopped fighting it. When it first started coming out, I'm like, 'No, no, no, that's not a...' but in the marketing sense because that is now the verbiage here. Where do you draw that boundary between deterministic workflows and maybe a little bit along the agentic judgment? Where do you draw that today?
Well, yeah, I think it's pretty obvious when you start building these tools. You've got a decision to make. In the deterministic world, if you want those decisions to be made, you have to articulate exactly the decisions. If this criteria is met, then go down this route. If this criteria is met, go down this route. If this criteria is met, go down this route. It works when the world is constrained to a small set of forks in the road where you're like, 'Okay, we're going to do this.' But there's a lot of decisions where you're not able to quickly articulate all the different forks in the road. And so when you get to that, you might say, 'Okay, I want to put an AI step in the middle and I want to have it make the decision for which fork in the road to go down.' And I'm going to give it a set of examples where I'm going to say, 'Hey, if you see something that looks kind of like this, this, or this, I want you to go down this route. If it's more like this, this, or this, I want you to go down that route. And if it looks like this, this, or this, I actually want you to check in with me first before you decide.' And that's helpful because what AI can do is it can pattern match. It can go, 'Hm, okay, you didn't actually give me an example that looked like this, but based on what you told me, it very clearly should be down this route, so I'm going to send it here. Or actually it should be down this very clearly should sit here or actually I don't actually know, it's kind of like both of these things, great, I'm going to check with you.' So that's like the decision tree is a very helpful one. There's cases where you're dealing with just large blobs of text, that's a use case that deterministic software really struggles with. You're seeing more and more workflows where people will go look at the last month's customer support tickets and categorize them in a particular way and tag all the keywords and summarize them by themes. That type of work was painstaking before. And candidly, most people just simply didn't do it. So that's a really interesting one. If you want to go generate emails or memos or documents or things like that, in the past the way you would see folks generate emails is they'd be using a mail merge style system. 'Hi, blank. It's nice seeing you today. I reached out because of blank.' We've all gotten these emails. They're horrific. You instantly archive these things. Now, if you give enough contacts to an AI, you can actually get it to generate an email that's pretty dang good. Now, a lot of folks still don't give it enough context, so you still get kind of AI slop style emails on the other side of it. But if you give it enough context, it's better than 80% of sales reps sending emails. I don't know, a pretty good chunk. It does a pretty good job. And those are the types of use cases again that were just not really possible, or if they were, you did a very poor version of them. So to me, a lot of the 'how do you decide when to use AI?' is just looking at the use case in front of you and it's like, well, there's no other way, I have to use it to do this effectively.
