Zapier is using AI to sell to AI

The agent is actually choosing what products to buy on behalf of a human. And so you're no longer advertising to the human. You are trying to get the agent to say, "Pick me, pick me, pick me."

Our CPTO vibecoded like a meeting recording software. It's a cool proof of concept, but we're not canceling our subscriptions anytime soon to our meeting recording software.

He calls it his war council. What the skill is is if you have a particular problem or decision you're making, you can just say, "Hey, I want to invoke the war council to like advise me on this."

Sure. Okay. Here we go.

Well, so I think this is a new thing that folks on the cutting edge are paying attention to, which is the agent is actually choosing what products to buy on behalf of a human. And so you're no longer advertising to the human. You are trying to get the agent to say, "Pick me, pick me, pick me." And that is often a very different skill set than marketing to a human. You know, with human, there's all sorts of like advertising history around like consumer psychology and like how do you position these things and the way an agent makes a decision is somewhat different than how a human makes a decision. And so, you know, this person is building all sorts of tools to like track like, you know, when does ChatGPT recommend this tool or not, when does like, you know, Claude recommend this or that and why do they recommend it? And then trying to figure out how do you like, you know, structure your content, structure your APIs, structure your CLI tools such that, you know, more often than not it's going to say oh when I have this problem I want to go use that vendor versus that other vendor.

Yeah.

Well, so agents care a ton about like clean documentation, like fast web pages. So, you know, if it's like the thing I'm seeing a lot of folks do is serve up a separate version of their web page that is meant just for the agent. And so that content is like plain text. It's served up like really fast. It's like written like very, it's like very descriptive, almost like mechanical because like that's what the agent prefers. It's like I want to have just like give me just the facts in like an easy to consume way. Very clear, very direct and as a result like it, you know, it's able to like understand what you're offering much better. Whereas like, you know, a human cares a ton about the design, like the visual design and things like that. Agent, it's not so much, right? So it's a different, you just kind of have to like figure out like what is the thing that it's actually optimizing. And I'll be honest like a lot of this is, you know, very much like more art than science right now. Yes, people are building evals and, you know, things like that to kind of track it. But yeah, when you ask like people who are really good at this, their answers are like very handwavy still.

Well, I think just the level of experimentation is just much, much more.

Exactly.

I mean, they're just like running a bunch of queries. They're having their friends run a bunch of queries. They're trying to pay attention to these things. There's like automated techniques that they're using. Yeah. So, there's, you know, it's just like a lot of trial and error.

Uh, you know, those are kind of the big things that I've heard of right now.

Well, our CPTO like vibecoded like a meeting recording software, you know, sort of like a Fathom or a Granola or, you know, a Grain or something like this. And you know, it's a cool proof of concept. But we're not canceling our subscriptions anytime soon to our meeting recording software.

You can't, you can't, like the maintenance on this stuff, like the tokens you burn setting it up, like, you know, at the end of the day like these products are not that expensive and the teams behind them are just putting so much effort in polishing them, making them better, advancing their capabilities and those are all things that we would have to staff engineers to keep working on them. And yes, it's gotten so much cheaper to build things but I think we would rather deploy that engineering power to making our stuff better versus, you know, improving our internal tools.

Yeah. Exactly. It's like, you know, for, I think vibe coding is fantastic for building tools that you can't buy or these like smaller utilities where it's like hey this is core to sort of like a one-off piece of software that I need to use to solve a particular job. Fantastic. We're seeing so much more of that stuff happen. But for like load-bearing infrastructure inside of a company there's not a lot that we're building and replacing. Like there's a lot of really good software out there that we're very happy to continue being customers of.

It's minimal, you know, it's like single-digit percentage. It's so small compared to like headcount.

Yeah. It's a drop in the bucket and another thing to maintain, support, pay attention to, etc. Let another company who cares about that deeply be awesome at that thing.

Well, I'll tell you when I built, so a guy who runs all our social stuff built a skill that's going viral inside the company right now. It's something you've been able to do for a while. So it's not actually new but the way in which he deployed it is somewhat novel and he calls it his war council. And so what this skill is is, it's for a decision you're making. So say like you're trying to decide on hiring a new person or say you're trying to decide whether you should enter a market or not or buy a company or not or, you know, just think of any like type of decision you might make on a day-to-day basis no matter how big or how little. And so, you know, you can give it all the like, you know, research you've done on the decision and then you can just say, "Hey, I want to invoke the war council to like advise me on this." And so, what it does is it spins up sub-agents that represent personas and they're standing members of the council. So, like I have the ruthless CFO, the wartime operator, the compassionate, you know, customer support person. And then for each scenario, it will also spin up dynamic personas. So if you ask it about hiring, it will spin up like a hiring expert. It will spin up a domain expert for the role if you're asking about like a product experience. It'll spin up a CTO. It'll spin up like a Steve Jobs like design visionary type person. And then all these like sub-agents go out and like debate and critique the problem and they come back and they give each one gives like their response to it. And then there's a master agent like the head of the council that sort of summarizes it and says here's what I think it should do. And then it says if I had a thousand dollars like here's what I would bet, you know, that the answer happens to be. And again like you could have done this like any time in the last couple years you could have spun it up but like the new skills, like techniques, the ability to launch sub-agents, the ability to think for longer makes the experience so much more fun. And so like, you know, I was like telling my wife I was like Chelsea like can I show you my war council? It's just, I don't know like I feel like a kid playing video games again.

Well so I'll tell you how I built it. My, the guy that's run social for us. We catch up 15 minutes every other week and he's always like building crazy stuff. And so I'm always asking him like, "Tell me what new thing you built." And he's like, "Wait, I got to tell you about my war council." And so, you know, he explains what it is to me. And I was like, "That's incredible." And so here's how I built it. I got off the call and I have Cursor hooked up to Granola. And so the meeting was recorded with Granola. And so I said, "Hey, Granola, go fetch that meeting I just had with Ryan." And in it, he describes the war council he built. Build that for me. And it just went and it just did it.

It just did it. Yeah. And like, you know, I popped open the skill and looked at it and I was like, "All right, looks good to me." Like, I didn't have any other critiques, but it was like it was perfect. And that's all I had to do is like I never actually saw what he specifically built. I literally just had him describe it to me and then I just told Cursor, I'm like, "Hey, can you take the notes and just do the same thing, please?" And the way that you do it is you do it as a skill, not as like just software.

I think skills are, the inspiration behind skills is way more useful. They're not hard to build at all. Like they're just a document with instructions and guides to them. They kind of remind me of, like this, remember like, I mean shoot this would have been like 15 years ago. There was all these like WordPress theming businesses, right? Where, you know, you could go find like a skin for your WordPress site and people would sell them for like, you know, buy this one for $10, buy this one for $15, things like that. I could see skills kind of having like a business model like that where it's like you could buy, you know, buy one for five bucks, 10 bucks. There would be a bunch of free skills out there. You know, it feels kind of like a very commodity thing at a certain point in time. But there is value in just seeing what other skills people have built because, you know, like the creativity of these users is what's so cool is you're like, "Oh, wow. I like the way you did that. I want to use that, too."

Well, I love using them for hiring decisions. Because it is very critical and so, you know, I find that humans when you go through hiring processes have all sorts of like biases built into them. I think one, you find this, it's hard for humans to say no to like candidates that are good. It gets harder to say no as the interview goes along. So like two people before you said yes, now are you going to be the person that says no? So you get more of that. And also humans I find are not often as like critical and so, you know, they'll ask like a surface level question and then kind of move on to the next thing and they don't necessarily dig as much to be like is this person really, really good. And so as a result when you score against the rubric I find that human graders on average tend to be a little softer. Now that's not universally true. I've seen like human evaluators be like very, very strict too. But yeah, I think the war council is like, it plays its role and it tells you exactly what it sees. It's not influenced by, you know, these other things. It just calls it like it sees it. And as a result, I find it to be a really helpful tool at keeping the panel honest, keeping me honest on a decision where I can find myself sort of like feeling one way or the other. And then you talk to it and it like really grounds a decision and you're like then you have to think about it. Like if it disagrees with you, you're like, hm, why does it disagree with me?

Uh let's see. I've got, what was a recent one that I was working with it on? I had, oh I was reviewing a bunch of sales data and I was trying to figure out like, some of the stuff was good, some of the stuff we were doing wasn't so good and I was like trying to figure out what kind of feedback I needed to give. And so I first like went back and forth with it without invoking the skill, just like pulling in a bunch of contacts, loading in a bunch of data. And then finally I was like, okay, I want the war council to advise on this. Like what is, like where are we winning and where are we losing on this? Like what do you see? And like what kind of feedback do you, you know, needs to be given? And so it like quickly summarized it was like here's the three things that are like most important. You're killing the first two. The third is like a blood bath. You need to fix it. And so you're like, oh, okay. Like it put language to a thing I was maybe feeling but couldn't quite articulate yet.