#2308 $570K/Year in a Month: One Guy + Claude Runs a 30-Client Agency

Andrew Warner (0:00)

This guy reached $570,000 in annual recurring revenue within a month of launching his idea. You see, he realized that he had a system for getting his site to rank in ChatGPT, Google and other discovery tools. So he offered it as a paid service. He signed up dozens of clients, and then he let Claude do most of the work. This is the kind of business that can really take off in an AI driven world. So I asked my friend Josh Pickford to talk about how it's done. Let's get into it.

Presented by Zapier, the AI automation company. Josh, this is the tweet that made me just reach out to you on tech so many times to have you come on here. You are up to, is this still accurate, $474,000 in annual recurring revenue?

Josh Pigford (0:43)

No, it's about 100,000 short now.

Andrew Warner (0:46)

What do you mean? Where is it?

Josh Pigford (0:48)

I mean, it's at like 570,000 something now.

Andrew Warner (0:53)

And this is an idea that wasn't public a month ago.

Josh Pigford (0:57)

Correct, so this is the first tweet. This was on a Friday at almost 5 p.m. On September 25th was what kicked us off.

So it hasn't even been a month yet.

Andrew Warner (1:08)

Summarize this message. What was the offer that you made in that first tweet?

Josh Pigford (1:12)

Yeah, so it's, I've been doing, well, I mean, I've been an entrepreneur for two decades, and so obviously I've done a lot of SEO stuff with different businesses over the years. So I had sort of turned all of this SEO plus AI mention tracking stuff into this whole system that I've been using on my own sites for most of the year. And so I said, hey, let me, I would love to do this for some other companies and just see if there's anything here.

So that was a very much, I'm wrapping up at work for the day, about to go downstairs and eat dinner with the family. And so I just threw this out and said, $5,000 a month, this all happens in Slack. Shoot me a DM if you're interested.

Andrew Warner (1:57)

And the offer is that when somebody is searching for them on OpenAI's different chat apps on Claude, et cetera, that you will help them be the answer that those AI engines offer up, right?

Josh Pigford (2:11)

Correct, that's the basic value proposition.

Andrew Warner (2:15)

How can you do that? What's the approach that you take? Because this is also new and mysterious that nobody has an easy answer, do they?

Josh Pigford (2:23)

Sure, no, no. Anybody who's promising rankings is full of it, but all this stuff is rooted in sort of good SEO, which has been around since the beginning of the internet. And so at the foundation, it's just good SEO practices. And then we're tracking all of both SEO and what would be called like AEO or GEO.

And making changes that address the ways that these different LLMs approach finding information and gearing it towards them, and then tracking what works, what doesn't, and then adjusting. So this isn't like a one and done thing. This is like every single day we're making tweaks to your content and deciding what's sort of most important to do in a given day.

Andrew Warner (3:14)

When you say we, how many people are on the team doing this?

Josh Pigford (3:19)

Me and a bunch of AI bots.

Andrew Warner (3:25)

What's the AI that you're using to run it? What's the overall approach?

Josh Pigford (3:30)

Yeah, so it's Claude at the foundation. That's for pretty much everything. There's a little bit of GPT sprinkled in, around some image generation stuff when that's required, but even that's pretty rare. So it's basically Claude, Opus, at pretty much every step.

Andrew Warner (3:54)

Okay, how do you keep them separated, all these different clients? Is it different Claude chats? Is it something else?

Josh Pigford (3:59)

Yeah, so they are all, from a technical perspective, very siloed, so there's not this bleed between clients.

They've each got their own. So, we operate through GitHub repos, so both their own, and then we also have our own versions that are instances that we're tracking all this stuff, so that, again, there's no bleed between files or anything like that, and then there's different sessions with Claude to do each task, so there's no context bleed or anything like that.

Andrew Warner (4:32)

What do you keep in a repo in GitHub for each client to make sure that you're organized with them?

Josh Pigford (4:38)

Yeah, so that's a lot of its data is the, it's both what would be like a work log, like here are the things that we did, here's all of the research that we've done, so when you produce an article, we might go find 100 different pieces of content to use as reference for producing that content, so we're saving all of that.