Wade Foster on how Zapier's SDK turns Claude Code and Cursor into 9,000-tool automation engines

strategy of not overreaching and uh it could really work out for them. So it's an exciting it's an exciting time and an exciting story to to follow. Well uh without further ado we have our next guest in the waiting room. Wade Foster from Zapier. Zapier. Wade. How you doing?

Good. How are you guys doing? We're good. Thanks so much for coming back to the show. Uh, how are things going? Give us the general update on, uh, your company and then there's a million different things that we can talk about, but I'd be interested to know uh, what you're learning from your vantage point as CEO of Zapier.

Well, yeah, I, you know, we launched our Zapier SDK last week. That's probably the most exciting news. Congratulations. this uh you can install it inside of you know cursor cloud code any coding agent you might use uh I use it I'm not an engineer and I basically built like an entire AI chief of staff on top of this team thing and uh you know runs in the background I don't have to close my laptop uh or I can close my laptop lid and not worry about uh things breaking so uh it hooks into all 9,000 different tools and I have it running all over the place so that that's probably the exciting news out of Zapier's land. Do you do you feel like uh the the vibe coding boom generally is more heavily tilted towards internal tools, dashboards, reports, chief of staff, B2B applications because it feels like clearly people are consuming a lot of tokens. They're spending a lot on inference. The models clearly work. The coding agents clearly work. At the same time, uh, I open up the app store and I'm not seeing the, "Oh, wow. Somebody vibecoded a Call of Duty competitor that's blowing up on the Steam charts, right?" Like, we're just not in this like in this like everyone's talking about the the the the new app or the new Uber or the new game that's like been created much faster with with a leaner team. But everyone has a story of we pulled forward our roadmap at our company. Yeah, we've definitely had a renaissance of internal tools internally. Uh we our marketing team ran a hackathon recently. It's about 50 people on the marketing team. At the end of the week there was like 80 plus new internal tools. Most of these were like dashboards and you know data visualizations and things like that that in the past marketers like especially are pretty analytical but they may not be like the best at like using SQL or running these things and you know getting access to the data has been clumsy and hard and now it just feels like they've got a jetpack on where they can just go build these like really quick quick simple tools

that allows them to make like hyper specialized hyper customized campaigns uh because they can quickly run an analysis and see oh you know, uh, across the last month of like social posts, these are the ones that took off. Uh, here's the common traits about them. So, let's figure out how to like tweak our next campaign for the next month to be more like this and less like that. Uh, that type of analysis would have been really difficult for a marketer to do solo. Um, you know, in the past, now it's, you know,

less than an hour of work. Yeah, I think Terrence Tao had a the mathematician had a point about this where uh he was he was he was saying that like in his papers he will use AI agents to create new visualizations that would have taken him multiple days to make. Now he but it's not like it's saving him time because in previous papers he just wouldn't have put the v visualization in. And so it's it's this it's this element of like people just doing more with more. But I'm wondering like as you see, you know, so many companies have done this where there's a hackathon for the marketing department or some organization and the the the the prompt can often be like replace yourself. And I'm wondering if you're seeing that actually happen or if it's more like the the competitive dynamic is everyone's getting more leverage, doing more things, and just out of the 20 ideas that they have, they're able to do 15 instead of five. Yeah, I I definitely think it's more the latter right now. Uh, you know, across Zapper, 100% of our employees are using AI daily. I would say this is like individual AI usage. People are individually more productive.

But if you were to ask me, is the company more productive? Institutionally, are we more productive? A lot harder to say yes to that. Yeah.

And I think that this is where like the company's on the cutting edge, this feels like the next thing that everybody's working on is how do we actually accelerate the institution? And this gets harder because you actually have to rethink how the company works. Um to give you an example, we've gone from a world where code was expensive to now code is cheap. Yeah. In the world where code was expensive, you had all these processes up front to talk to customers, to align on road maps, to figure out what the right thing to build was because if you chose the wrong thing, that was a really expensive mistake. And so all these humans and all these processes exist up here. But when code is cheap, you can get rid of all that stuff.

And that takes redesigning jobs or rethinking the human where the human actually exists in the first place.

If you're in YC, you no longer if you were in YC and you were about to talk to your customer, just cancel the call. Just don't you don't need to do it anymore. It's all good. That's a old No, I'm kidding. No, it's a it's it's a very good point. It's like, hey, show the customer like that's the thing. It's like you don't need to show

Yeah. Yeah, it's like a non- tech founder in the past would like show up and like give them a survey or like ask them generically about the problem or something like that and you would get some signal from that. But now a non-technical founder can literally show up and say like

would like can you just use this?

Yep.

Yeah. And see what happened and that gives you a lot more information in the past. And so it's just a much different like learning cycle. And I think comp like companies that are, you know, not in YC, companies that have been around for a while have all these systems and processes that have been built up that have to be torn down and reassembled to work in the AI area. And that's where I think the institutional AI really kicks in.

Yeah. How have you been grappling with the idea of work slop? There was this a great cartoon early on where uh you know the the the joke was take these three bullet points and turn it into an email. And then on the other side, somebody says, "Take this email and turn it into three bullet points." And people should have just been sending three bullet points to each other the entire time. And I feel like we are hyper in that world where I can send you a, you know, a full essay, a deep research report, a book if I want, a dashboard, a visualization, a 100page deck. And uh at the same time, sometimes, you know, the human intuition of knowing, okay, this was a good post, let's just do more of that might be the right intuition. So, how have you grappled with uh you know fine-tuning the team on okay it's great everyone should be using AI but let's stay away from work slop

yeah you my uh co-founder Brian says you can delegate the work but you can't delegate the accountability and so you know if you're slinging this stuff you got to stand behind it's okay to present AI author work I do that often often times I will say hey I was going back and forth with my AI friends

here is what I found I have read this and I actually like stand behind it and agree with this,

but if you're just going to like, you know, throw a random prompt in and then pass it off as like your own work.

Yeah.

That ain't that ain't cool. And so, like, we kind of try and coach people on that the kind of etiquette around that kind of stuff. Um

Yeah.

Yeah.

I I had that early on. Uh I I was working with somebody and they sent me something that would had clearly been, you know, hydrated by an LLM and I was like, you you could just send me the prompt because I can actually imagine it exactly what it was. And like, yeah, if there's a fact that we need to look up, we can use it for knowledge retrieval. But, uh, I I I actually don't need all of the hydration. You can just use this on the prompt.

You guys you guys are remote company. How do do you think that gives you an advantage in the in the AI era? Do you know what what are the ways?

I think the thing the biggest advantage we have is every last bit of work exhaust is documented. So, all of the stuff is inside of Slack. All of our meetings are recording. uh you know every last inch of work that happens there is a like a written trace of that and so we can put chat bots on top of that we can put LLMs on top of that and that creates a whole bunch of institutional knowledge that accelerates the work so a new person coming in can literally figure out is there a standard operating procedure for this is there a skill for this is there whatever to do this thing uh and you don't have to go like chase people down in offices and tap on shoulders and sort of hope that the campfire wisdom like finds you. So, I do think remote companies have a big advantage because they do tend to have so much of the work um has a digital exhaust.

Yeah. Uh how how are you thinking about like uh the the volume of content? Are is it just like stuff everything in one large context window? Are there like pro are there agents that are going around and like creating little rollups of things and there's processes for teams to boil things down? because even even just the explosion of like uh you know Zoom call recording and note-taking can produce like so many transcripts and I've been sort of shocked by the slow speed of just search in every tool that I use because there's so many more words that when I search for receipt in my email or something I get you know every single email possible because they all have so much text now

so what we have done internally is we've built a company brain out that has uh canonical data. So it exists at a couple different layers. There's a set of company data. This is like our company strategy, our company values, our ideal customer profile, things like that that is curated by me and a handful of other people that sort of say, "Hey, this is the truth in these areas."

Then there's team data and so every team internally has a similar version that kind of cascades down. And then finally, folks have individual content that they've sort of used to supplement that that's private to them. So anytime someone is like talking to their AI, they have this you know source of truth that has been curated then you know folks are know that they can pull in other additional content at any given time. And so for me you know if I wanted to pull in a meeting transcript or a slack thread or something like that I would point the SDK at both those documents and say hey I want you to go review those against the backdrop of this company brain. And so it sort of helps you manage the context window because you can take, you know, bite-sized chunks here and here, you know, have the like company brain as like the backdrop to all that stuff and get better um insights versus just saying look at all of Slack or look at all the meeting notes. Um which that would just not be very effective because there is a lot of noise inside of that.

Yeah. Have you seen more uh have you seen acceleration either at the lower end below your average customer size or or bigger companies on boarding? Like I I imagine that uh there are a lot of tailwinds right now for you. Uh but what's uh where are you seeing faster adoption or more interest?

You know, I think small companies tend to just move faster. They just have less bottlenecks on this type of stuff. Um that said, I have been somewhat surprised uh over the last couple years how fast enterprises have gotten onto this stuff. Um you know, they still have more procurement hurdles and um you know uh uh boxes to check to really take off here, but they're not asleep at the wheel, which is kind of the meme is that ah you know these enterprises move slow and don't make decisions da da da da. Um yeah, I think they just have more things that they considerations that they have to to do. But um I I do think smaller companies generally just are moving a lot faster on this stuff.

Are you are you feeling CPU poor? We've been seeing like with the with the explosion of agents and just more work being done on the internet. I could imagine a lot of your customers submitting way more requests. There's an explosion in demand. And maybe there's two ways. One is just scale up your servers, but the other one is maybe uh work on optimizations and rewrites. Now the company's been optimizing for over a decade now. So I imagine things are pretty optimized. But how have you been processing just uh increased demand relative to uh you know just uh infrastructure?

I think the most interesting place we've seen this is new usage patterns. So agents you know interact with APIs different than how like a human might set something up. Uh and you're also able to deploy them faster and at a larger scale. And so, you know, when we're hitting, you know, like I don't know, like let's use like Slack for example, like the way you might, you know, set up automation across your Slack channels like looks a lot different than it did preai because you have new use cases and new patterns that you want to use that are more um uh just want more data. Like it just wants to look at more stuff. And so I I do think that uh that's probably the the most interesting change is the the usage patterns are are are different in the AI world.

Yeah. for some of the more like basic workflows, what is the what is the pitch for using uh Zapier instead of vibe coding something that takes my LinkedIn post and emails it to me or dumps it in my Slack because that feels like something that you could effectively oneshot, but then you get caught in who's going to maintenance that or who are you serving that? What happens when the API changes? like uh I feel like there's a lot of uh excitement about oh I can vibe code this in an hour uh and maybe people aren't thinking about what it takes to maintain a service but how are you how are you uh reiterating the benefits of being on Zapier

I think the the fact that we maintain all this stuff is a huge advantage uh the other thing is it's storing your tokens very safely a lot of these vibe coding tools will say hey just copy and paste your API key here and it goes into a plain text file uh that might get shipped up to GitHub and might get shipped somewhere else that's like crap, we're in trouble. And so, you know, Zapier just tends to be a safer place to do this. Plus, you know, we have all the vibe coding uh capabilities as well, too. So, you know, hey, take take my leads from this service and add them to that and d, you know, Zapier is able to do uh a lot of that as well. So, tends to be a little bit, you know, safer, more reliable place to to run these types of automations.

Yeah. uh timelines around fully autonomous companies where you can have an agent that you just kind of

prompt it to uh you know this has been an idea that's been around for a while. You know, you just say like figure out a way to make money selling software and it just goes off and it maybe makes some software and figures how out how to market it. Uh I this already kind of exists in kind of the trading world, hedge fund world where they're creating algorithms that just go out and uh figure out ways to make money. But uh on the software side, I'm curious given everything you know about, you know, automating business processes when you think it's really possible. I I to your point I do think for some very simple things like it is pretty close trading you know whether you're actually good at it or not TBD but like pretty dang close there.

Yeah.

Uh I also have a former Zappier employee this guy Nat Elias who made it has an open cl Yeah. he's an openclaw that you know started making money on courses uh and you know basically was yeah openclaw was just like figuring out how to go do it uh and so uh you know it's like pretty close uh so you're starting to see

there's also use cases like like I think of like a simple app in the Shopify app store where you could just say like go compete with this app and then it's like pretty obvious where you need to advertise it and you can figure out the equation between like you could just make the make a clone of the app and then figure out the equation of like what is the minimum amount that I need to charge in order to be able to spend enough on user acquisition to like eke out some like very thin margin that covers the cost of inference.

Totally. The thing I don't know is if it's truly going to be a zero person company or not. Like it does still feel like you need a human to like supervise this thing to like you know when it goes off the ropes to actually have the initial idea. Uh, so you know, I don't think we're at this dream where you like poke the

Yeah, there's always gonna be Yeah, there's always going to be qualifiers. Even the one person one billion dollar company that was in the New York Times, it's like okay, like he hired his brother and then also he had a bunch of contractors. And so there's never like even if you do create like an autonomous business like if it's working at all then suddenly there becomes an economic incentive to maybe hey maybe I should spend an hour on it and then like well is it still really autonomous if you spend an hour a week on it.

Yeah. Well, and that's the thing I think is also going to make it hard for these things to be crazy pervasive is that if someone is being that successful, there's another person that would look at the thing you're doing and say, "Well, I can do that, too, and let me go compete against your margins."

Yeah.

How do you think about uh terminology? Jordy always gives this example of a company that had workflows and they just rebranded it agents and they didn't really change anything under the hood. uh and it feels like there's immense pressure from investors and marketers to just use the latest lingo. Uh and sometimes that's useful because the capabilities change and so the nomenclature should change. But we're in this weird continuum where there's there's workflows, AI workflows, agentic workflows, full AI agents, like are these meaningful definitions? How have you been processing all of this since you're sort of in all categories? We definitely have a a slide that explains the difference between all three of the things you just said.

Yeah.

Practically speaking though, they're all agents. You know, you know, I you know, I'm sure like somebody who is, you know, uh uh like a master wordsmith would like, you know, try and debate the finer nuances it. But what I see in the market is practically what people are saying is an agent is a thing that does automation for me. Yeah. Whether it's purely purely deterministic or whether it's purely agentic, it's kind of an agent at the end of the day. Now, you know, deterministic workflows have some advantages and disadvantages. Agentic workflows have some advantages and disadvantages. As you start to go build these in production use cases, you probably should understand the difference between these things, but practically speaking, we see customers showing up and they haven't really figured that stuff out yet. They're just like, I have a problem and I want that problem solved.

Yeah. Uh what about uh vibe coding? Like how how broad do you think this trend goes? I saw someone

uh I think it was who was it downtown Josh Brown joking about how uh there he's getting sent like a new thing every single day. Uh

he gave the example of like somebody that makes an app that like displays their Spotify playlist nicely and he's like you're not my child. Like I'm not going to spend time on this. I don't I don't owe you any time. Um but of course

Yeah. I I'm just wondering how how you're processing like it it feels like this year vibe coding is going very mainstream. A lot of people are toying with it. Uh what do you think retention looks like? What do you think the knock-on implications are of this? Uh any like security concerns? Just how are you processing this idea that like the number of people that will have created software will probably like 10x this year?

I think it's very exciting. uh you have all these folks who have had ideas or coming up with ideas for problems to solve etc. But in the past those ideas have kind of been locked away because they can only implement like a small piece of it and now with vibe coding it feels like you can do a lot more stuff and so I I think that's why we've seen such a surge is that everyone feels like they can get these creative ideas out of themselves. I do think as an industry there's going to have to be a lot more stuff built out to make these things work in all the various um you know use cases. I think one of the places Zapier is like very helpful is uh you know connecting to all your data sources and doing it in a secure way like we'll handle all your authentications clearly and so that's a way that we can sort of uh you know play a really important role in how people go build out um you know these these infrastructure and it can move less from vibe into more trusted uh you know infrastructure.

Yeah. Yeah. Makes a lot of sense. Uh Jordy, anything else?

No, this was great.

Well, thank you so much for taking the time to join us. Have a great rest of your week and we'll talk to you soon. Wade,

good to see you.

Have a good one. Uh before we bring in our next guest, Kevin Walsh, who is the new uh Fed chair nominee, uh had a financial disclosure and has a ton of startup investments. He's in SpaceX and a whole bunch of other companies. uh zero he's a one man DC machine

brave the the browser cognition

uh IP the tech enabled patent law firm

uh dydx

protocol

he's in a lot of stuff uh I think some of this is I think some of this is through funds that he's invested in but still it's uh it's a whole bunch of uh it's a whole bunch of companies uh and it's interesting to to dig into he also owns a a horse racing stable, something like that. Did you see this? Uh, it also includes his role as a general partner in Vicarage Stable, a horse racing operation.