The episode goes live, and the social posts become whatever someone can fit in before the next client deadline. On a small team, the person who handled repurposing last week is on billable work this week, so the next batch comes out in a different style, or it doesn’t come out at all.
Rich Brooks, president of Flyte New Media and host of The Agents of Change podcast for 638 episodes, fixed that by locking his whole post-production checklist inside one Claude project. In this session, he drops in a transcript, types “do your voodoo,” and walks through what comes back: SEO fields, five takeaways, a ranked keyword table, separate posts for Facebook, Instagram, and LinkedIn, a blog draft in his voice, and five ways to sign off the episode. Then he opens Descript and pulls the video clips.
Rich stops writing and starts editing. A blog post that takes him two to three hours from scratch takes him five to 10 minutes to fix, and he spends that saved attention on the things the project still gets wrong.
What you will learn
- Rich drafts a 1,000 to 1,500-word blog post from each transcript in a Claude project, then spends five to 10 minutes editing it instead of the two to three hours he needs to write one himself.
- One set of project instructions returns an SEO title, a meta description, five takeaways, a ranked keyword list, posts for Facebook, Instagram and LinkedIn, the blog draft, and five episode sign-offs.
- Uploading finished outputs you are proud of as knowledge docs shows the project what good looks like, which written instructions alone struggle to do.
- When drafts used “actually” seven times in 2,000 words, Rich capped the word at one use per post, and the project now reports how many it removed.
- Republishing the edited post as a LinkedIn newsletter every Friday takes Rich about 10 minutes and has drawn more than 1,300 subscribers with no promotion.
- A saved Descript template returns 10 candidate 30-second clips per episode, centered on the guest and carrying Rich’s branding; he keeps about three.
Timestamp index
00:00 Two to three hours per blog post, cut to 10 minutes of editing
01:19 Why repurposing broke down whenever client work came first
02:22 Why Rich built his own Claude project instead of using a built-in AI
04:02 Recording in Descript and exporting the transcript
05:21 The instruction list inside the AOC Podcast Producer
06:48 A sign-off generator for a show with no catchphrase
07:49 “Do your voodoo” and uploading the transcript as markdown
08:08 Descript returns 10 clips, Rich keeps the best three
10:06 Platform posts, including no hashtags on Facebook
10:50 Capping “actually” at one use per post
12:29 Republishing the blog post as a LinkedIn newsletter
13:16 Asking whether a fix is a one-off or an instruction update
13:35 Catching a quote credited to the wrong person
16:04 Uploading finished examples as knowledge docs
16:48 Checking for drift after model updates
17:12 The RICCE framework for project instructions
18:29 Why a raw transcript may rank worse than a structured post
19:46 Better output first, time saved second
21:05 What changed on social: LinkedIn and YouTube, not Facebook
22:50 Where to find Rich
One project holds the whole post-production checklist
Rich calls his project the AOC Podcast Producer. The instructions cast Claude as an experienced podcast producer and promoter, then list the jobs in order: an SEO title of 50 to 60 characters ending in “with [guest’s name],” a meta description of 150 to 160 characters, an introductory paragraph in Rich’s voice that he also uses for the episode’s ID3 tags, five takeaways, a keyword list ranked by relevance and estimated search volume, the social posts, a standalone blog post of 750 to 1,500 words, and five sign-offs. Rich shared the full instructions after the recording, and they are in the Blueprint below.
Two details in those instructions do more work than they look like they should. One line tells Claude to produce everything without waiting for confirmation, and to pause only for genuine clarification, so Rich can check its work. The platform rules are exact: no hashtags on Facebook, every Instagram post ends with “link in bio,” and every hashtag set includes #digitalmarketing and #aocp plus two or three episode-specific tags.
Running it is short. Rich exports the episode from Descript, saves the transcript as a markdown file because Claude handles it more cleanly, and uploads it. Without a command, the project sometimes hesitates even though the instructions say what to do, so he now gives it “really sophisticated prompts like, ‘Do your voodoo'”.
The keyword step runs on what the model already knows. Rich considered connecting a keyword tool like Ahrefs and decided it wasn’t worth the tokens. Mike takes the other route on this show and connects Ubersuggest so Claude can pull targeted keywords in a few minutes. Rich also built the first version as a custom GPT in ChatGPT before moving to Claude, and the same setup works in ChatGPT projects or notebooks in Gemini and Copilot.
Give it examples first, then constraints for your own tics
When Rich built the project, he uploaded finished outputs he was proud of as knowledge docs, alongside the transcripts they came from, so Claude could see what good post-production looked like. “I feel that examples are the special sauce when you’re trying to get AI to deliver something in the way it is in your head”.
Then he started catching patterns. An early blog draft used “actually” seven times in about 2,000 words. Rich checked and found he says it about that often, so Claude was matching his voice a little too well. He added a constraint allowing the word once per post, and the project’s final notes now report that it removed the extras. The same double-check step strips em dashes.
He also asks a question most people skip after a bad output: is this a one-off, or should the instructions change? Claude answers, and Rich decides whether to edit the project. Human review still earns its place. In one recent draft, Claude credited a quote to his guest that Rich had said himself.
Rich’s warning for anyone who builds one of these is that it drifts. Models update, so he goes back to the original knowledge docs and instructions every so often to check the output still matches what he built it for. Mike adds the RICCE framework for writing the instructions in the first place: role, instructions, context, constraints, and examples. Examples are the step most people skip, because the AI has no way to know what good looks like to you until you show it.
Descript handles the clips while Claude writes
Rich records in Descript, which captures both sides of the call much like StreamYard or Riverside. While the Claude project runs, he goes back into Descript and asks for clips. A saved template returns 10 candidate 30-second clips, centered on the guest’s face, with his branding and logo in place. He picks about three, tidies them in the text editor, and hands them to his team for the social calendar.
He knows that hand-off could be automated too. “And yes, I’m sure I could automate more of that part too, but baby steps”.
One blog post, published twice
The episode goes live on Wednesday with the blog post on the Agents of Change site, and the full transcript stays on the page for anyone who wants it. On Friday, Rich takes his edited version of the same post and publishes it as a LinkedIn newsletter. He adds a short blurb, uses the image his team made, and tags the right people, in about 10 minutes. The newsletter has more than 1,300 subscribers, and he has never promoted it.
He also has a theory about search, and he labels it a theory. The raw transcript meanders, so Google and AI tools struggle to make sense of it. The structured post built from the same conversation gives them clear takeaways to work with. Rich says he can see an uptick in high-ranking keywords on episodes that got the blog treatment, while being clear he can’t prove the blog post caused it.
Time saved is a bonus in his book. The test he runs on anything he builds with AI is whether the output is better than what he produced alone: “Is it better than it was when I was working by myself?”
Be honest about which channels move
Mike asked what changed on the social side. Rich’s answer was that the feeds haven’t changed much. He counts the podcast and the LinkedIn newsletter as his social media, and those are where he sees results. The Descript clips now run as YouTube Shorts and pull hundreds of views, sometimes a thousand, though he can’t yet tie those views to revenue.
Organic Facebook posts for Agents of Change haven’t gotten traction in years, and he said so on a show about AI for social: “I feel like I’m checking a box”. Rich points out that most of his business is B2B, and a B2C brand may see something different. Mike’s experience matches Rich’s. Agorapulse pushes the reels to Facebook, Instagram, and TikTok in one click, and his own focus stays on LinkedIn and YouTube.
For a social team deciding where repurposing effort goes, that’s a useful split. Let the project produce the posts for every channel, and spend your editing minutes on the channels that return them.
Grab the Episode 8 Blueprint
Rich’s complete AOC Podcast Producer instructions, a version adapted for any show or webinar, and the steps for loading knowledge docs and constraints. Download it here.
Connect
CONNECT WITH RICH BROOKS
Listen and subscribe
Full episode transcript
Rich Brooks: ‘Cause to write a 1,000- or 1,500-word blog post, for me, honestly, sitting down—not even including the research—is at least two to three hours. It takes me about five to 10 minutes to edit this draft, so that’s an incredible time savings right there. Five to 10 minutes each Friday morning, and without any promotion at all, we have over 1,300 subscribers to that email. The goal here is to save me time and to repurpose the content in a way that’s aligned with what my audience needs and coming from my voice.
Mike Allton: Welcome back to the lab. I’m Mike Allton. Most social teams treat a podcast as one post and then stop. Today, Rich Brooks, president of Flyte New Media, shows us the Claude project that turns one recording into an entire week of content. Hey, Rich. Welcome to the show.
Rich Brooks: Hey, Mike, thanks for having me. I’m excited to be in the lab with you.
Why Repurposing Kept Breaking Down
Mike Allton: My pleasure. Most of us, as I said, hit record, do this episode thing, then we hit publish, and then social becomes whatever we can squeeze in before we move on to the next thing. And when I say most of us, I’m really talking about myself here. But before you built this pipeline, is that pretty much where you were at, too?
Rich Brooks: Yeah. I mean, obviously, I run a small agency, as you know, and I have people on my team who can take on some of this work. So it was a combination of me doing some work and them doing some work, but if client work came up, we were always going to do client work before taking care of ourselves.
So there was a lot of inconsistency, or somebody was busy one week, and someone else took it over, but they didn’t necessarily do it the same way. I just wanted to streamline a lot of these processes, lock them down, and see how much of it I could take off the plates of the people doing billable work in my office so we could get things done.
Mike Allton: Right. Right. Now, I should say for those of you watching and listening, Rich and I are going to be talking about podcasts specifically, but if you don’t have a podcast, you’re probably doing webinars or other kinds of live-streaming videos, maybe. There are a lot of different ways to create video assets that would still fall into this bucket of the problem that we’re trying to solve.
Now, Rich, what made you decide to build your own Claude project for this instead of just letting a scheduling tool’s built-in AI take a swing at it?
Rich Brooks: I guess part of it is because I like building things—I want to see how everything works, and building things is one of the best ways of doing it. But I also wanted some consistency, and I knew if I just used regular prompts or even used the AI tools that somebody else had built, I wouldn’t have nearly as much control over it.
I have a certain style of doing things, and one of the things that I’ve really gotten in tune with over the last year is the fact that AI is an incredibly powerful tool, but it can also separate us from the people we’re trying to talk to. AI slop—especially on the visual side, but even on the writing side—can sometimes create a valley or a chasm between us and our audience.
So what I wanted to do is build the tools myself so that, A, there was a human in the loop, but also so that it would be most aligned with who I am and what I’m doing. Hopefully, even though I still have ways that I could make this more efficient, I’ve found the balance between efficiency and remaining human—remaining Rich Brooks when I’m creating this content.
Mike Allton: Love it. Love it. So I want you to dive in and start showing how you do this. But before you get into that, let me just caveat for the audience: as we’re showing you Claude, that’s the tool that Rich uses, and that’s the tool I happen to use, but I don’t think too much of what we’re about to go through is exclusive to Claude. You have projects in Claude similar to how you have projects in ChatGPT, or notebooks in Gemini or Copilot. They all function fairly similarly, and if Rich starts to go into something that I think might be exclusive to Claude, I’ll call that out for you.
So Rich, walk us through where this actually starts. You’ve got a finished recording, and you’re in Descript. What do you do from there?
From Descript Recording to Claude Project
Rich Brooks: Well, to your point, I actually did the first version of this in ChatGPT as a custom GPT, and then I started to make the switch over to Claude just for a number of personal reasons. I liked it better. There are slight features that are different in each one, but what we’re talking about today should really work for anyone.
So, I had recently been recording in Zoom using Fathom to get a transcript. I’ve also got a human assistant who works with me who does a lot of this transcribing, but even with her, I’ve been coaching her on how to leverage AI more efficiently so we can get the best work—because I always think the best work comes from using AI along with a human being.
So really, where this comes from or the process is: I’ll get on a call with a great guest like you, Mike, and these days I’m recording in Descript. Descript has something—I think it’s called Rooms—but basically, it works like Riverside or some of these other tools, or StreamYard for example, where it’s recording on both sides and then creates a finished product. Using Descript, I can export into an audio file, which is what I’m doing. So I export this, I clean it up a little bit, and then I’ve got the transcript.
Now let me show you: this is my Agents of Change (AOC)—I came up with the acronym first before the politician from New York. I told you I wasn’t going to talk politics; that’s as close as I’ll come.
Mike Allton: That’s it.
Rich Brooks: So I call this the AOC, Agents Of Change podcast producer, and I created a series of instructions. I won’t bore you with all the details, but it basically breaks down some of the things that are important to me as somebody who loves SEO, by the way.
This has evolved over time, but the most current version basically takes the transcript and does a number of things with it. It identifies what are likely the most important keywords from an SEO standpoint—and it doesn’t actually go and use something like Ahrefs to do the keyword research (I could do this, but I didn’t think it was worth the tokens); it uses its basic model knowledge to assume what the most important keywords would be.
Based on that, it writes an SEO-enhanced title, meta description, and goes through a number of other steps. It drafts a blog post—I like to use the word “draft”—in my voice from my perspective of what I learned from speaking with my guest. It writes up all the descriptions for social media and follows my brand voice, so it talks like me throughout. It creates the ID3 tags, which—if you don’t understand podcasting—are just some of the meta information that goes along with a podcast, in the show notes and things like that.
I also have this brand, The Agents of Change, so one of the things that we talk about is how to be different and how to be an agent of change. One of the things that I always struggled with is having a tagline or a catchphrase at the end of each episode. Honestly, I’m 638 episodes in, and I still don’t have one! So recently I added an instruction: “I need you to create five different versions of a sign-off pulling from the content that we talked about today, ending with the phrase ‘agent of change.'”
That’s one of the steps it takes, too, and honestly, that’s the step I love the most. I’m not saying that I use it verbatim—sometimes I do, and other times I’m like, “That’s good, but I’m going to tweak it here and just add something in that’s relevant to that.” That gives me a starting point. As I said, I prefer to think of AI as drafting content, not writing a final version.
All that’s in there, maybe a few other things as well. Basically, all I need to do with the new episode is take that transcript, upload it, and drop it into the Agents of Change Podcast producer. If I don’t give it a command, sometimes it gets a little confused, or it’s like, “I think I know what to do,” even though it’s in the instructions.
Do Your Voodoo: Running the Project and Pulling Clips
Rich Brooks: So now I do things like really sophisticated prompts like, “Do your voodoo,” and that seems to be enough to get it done, right? So I upload, and I usually upload as an MD file, a markdown file, just because it’s easier and AI seems to be able to deal with it better.
So basically, I do this. While it’s starting, I actually go back into Descript. I use Descript, actually—one of the things I love about Descript is creating content for reels. So I’ll just type in… You know, a recent one that I did with Andy Crestodina was on GEO. So I just go in, I say, “Give me five clips.” I’ve got the template already set up, and it pulls in 10 potential 30-second clips that it finds to be really relevant. It centers on Andy’s face. It’s got my branding and logo in there. I go through and find, like, the three that I think are best. Sometimes I’ll edit it using the text editor over there, clean it up a little bit. I hand it off to my team, and it gets into our social media calendar.
And yes, I’m sure I could automate more of that part, too, but baby steps.
Oh, one other thing is I always like in my emails that go out to my subscribers a few of my favorite takeaways from the episode. So I always ask Claude to take the five best ones, and then from those five, I choose three to drop into the email as kind of a teaser for each episode.
And as I said, there are more automations that can be done. Where it is right now is where I’m comfortable. Maybe, you know, in a few months from now, I will have automated more and more of these processes.
So basically, I drop the transcript in, it goes up here, Claude does its voodoo, and basically gives me a summary. It sometimes double-checks things for me, as I introduced her as something like Friedman rather than Friedlander in this particular case, and AI’s like, “Just make sure you got her spelling right, ’cause, you know, you botched the intro there, Rich.”
And then it’s got all this other information, and then it kind of goes through, and it just shows me all its work, right? It pulls out the SEO title for me, so here’s How to Grow on Substack for Business with Tracy Friedlander, and this is the most recent episode we have.
The meta description, and again, these are all based on instructions that I gave to Claude or to this AOC podcast producer, to get the most relevant content out there. Introductory paragraph—this is actually the paragraph I use in the ID3 tags. My five key takeaways, such as “Discovery is Substack’s superpower.”
It shows me the prioritized SEO keyword table based on the searches and the relevance, and from that—and the instruction goes into more detail—it then writes the blog post in my voice and for our social media, pulling from that content. It gives me my social media post for Facebook, where I tell it I never want hashtags, ’cause I’m a big fan of no hashtags on Facebook. It gives me the Instagram one, which often has more emojis built into it. It gives me the one for LinkedIn, and although we’re not currently posting on X for this particular thing, it does give me the X/Twitter export as well. And then it gives me the blog post, which has all been optimized for the keywords that it already identified.
Editing the Draft and Republishing It on LinkedIn
Rich Brooks: So what I do at this point is I read through this, because even though it’s been trained on my voice, it doesn’t always sound like me. I go in and tweak things. Like, one of the first things I noticed is it says “actually” about seven times in a 2,000-word blog post. Now, it turns out I actually say “actually” about seven times, so I had to correct that myself. But I fixed it. I said, “You’re only allowed to use ‘actually’ once.” So often when I look at the final notes, it’ll say, “Double-checked, removed all but one of the ‘actuallys.'”
So it’s learning. It breaks down all the big things that we talked about during the episode—my takeaways. It gives her, Tracy in this case, the credit for these things, and there are things in here where I don’t like the way they sound, so I’ll tweak something and all that sort of stuff. Like I said, this is a draft.
To write a 1,000- or 1,500-word blog post for me, honestly, sitting down—not even including the research—is at least two to three hours. That’s just how I write. I’m meticulous. I go back and forth. I read things out loud. I try them in different places. It takes me about five to 10 minutes to edit this draft, so that’s an incredible time savings right there.
This is the whole blog post, and there it is. So I clean it up, and I end up handing this off to my team with the full package. They’re the ones who actually put it on social media. They’re the ones who put this up on the Agents of Change website, along with all the links, everything Tracy shared with us, and obviously deal with some of the podcast stuff, too.
The show goes live on Wednesday. This same blog post I repurpose for LinkedIn. I basically take the revised version, and I publish it as a LinkedIn newsletter on Friday. So again, just squeezing as much juice out of this lemon as I possibly can. It takes me, on average, 10 minutes to take this blog post, put it on LinkedIn, grab the image that my team has made, publish it, add a little blurb at the top of it, and make sure I’ve tagged all the right people in it.
Five to 10 minutes each Friday morning, and without any promotion at all, we have over 1,300 subscribers to that email, introducing me—and in this case, Tracy—to a brand-new audience who might not be following me otherwise.
Then you can see it’s also got my sign-off. So there’s the inspirational one, the practical or tactical one, the contrarian version, the thematic one, and then the long game one as well. Like I said, if I don’t use them verbatim, at least it gives me a jumping-off point.
Sometimes I’ve told it that I’m not happy with something, and sometimes I’ll ask, “Is this a one-off? Do we need to update your instructions, or is this a situation where it’s a one-off where we don’t need to worry about it?” And it will provide feedback.
Here’s one thing: in the blog post, I noticed that it attributed a quote to Tracy that actually I had said. I was like, “We have to fix this.” And it was like, “Oh, I made a mistake.” So again, having a human in the loop is still important.
Then Tracy mentioned she was a concert French horn player. She mentioned that, so I used AI to go find me a copy of that and embedded that in the show notes as well.
Then I didn’t like the takeaways, so I said, “Focus on Substack,” and it gave me five new ones. Again, it wasn’t perfect. I found something else, and so back and forth. It even told me why my idea was better than its, and I’m like, “I agree with you.”
So again, it’s not perfect, but it’s not meant to be. The goal here is to save me time and to repurpose the content in a way that’s aligned with what my audience needs and coming from my voice. I spend a lot of time setting this up so I don’t have to spend a lot of time each week working through it.
Knowledge Docs, Drift and the RICCE Framework
Mike Allton: That is the exact right philosophy. I’ve been doing the exact same thing with Claude for this show, for my own personal show. So now with this show, for instance—and you’re going to get this when we publish it—Claude will take the transcript and create a whole collection of documents for me in a Google Drive folder, including the email that I’m going to send to you, social media posts for you, social media posts for me, for Agorapulse, and the initial blog post that goes out with this.
Because you and I had a similar conversation months ago when you were just starting to experiment with posting a second article, I took that and was inspired by it. So that’s what we’re doing now at Agorapulse with this show: every interview generates two blog posts. What else? A newsletter, but I ship all those as separate documents, including the playbook.
So Claude takes that. I’ll copy and paste the prompts, or it’ll pull it out of the transcript and create this really gorgeous playbook, this downloadable PDF. It does all of that, and the other thing that I would suggest for you guys listening and for you, Rich: I connected Neil Patel’s Ubersuggest as my SEO tool. So it can hit that for just a few minutes, grab targeted keywords, pull that in, and make sure everything’s kosher in terms of who we’re talking to.
Rich Brooks: I think what I love about this entire process is you’re starting with a transcript. You’re starting with a transcript of your exact words, your guest’s exact words. So, generally speaking, I feel like it’s really easy for the AI to come up with content that is in the guest’s voice or your voice, cause it has it right there in front of it.
And I think that we didn’t—we talked about the instructions, but what we didn’t really get into is some of the knowledge docs that I first put up there. So when I was first starting this, I feel that examples are the special sauce when you’re trying to get AI to deliver something in the way it is in your head.
I uploaded a number of—not just the transcripts, but the finished product—and shared ones that I was especially proud of with Claude when I was building this, so it would know what a good transcript or good post-production content looked like. So that was one of the ways that I got started, and then over time, I’ve added additional instructions to improve not just the quality, but the consistency of the outputs as well.
And I think this is something that you can’t set and forget. First, there’s a certain period of iteration that needs to happen during the creation period, but also as the models change and other things evolve, it’s important to every once in a while go back and look at those original knowledge docs, look at the instructions, and make sure that there hasn’t been some drift over time, possibly because of a model update or something like that.
Mike Allton: Love that. One of the things I often teach people who maybe are not familiar with setting up instructions for an AI, whether it’s a custom GPT or a project instruction, something like that, is to use the RICCE framework—R-I-C-C-E—where you start with the Role: whatever this person is. In this case, “You, the AI, are a 10-year veteran at editing podcasts and know the difference between a show that’s snoozing and a show that’s electrifying the audience.” That’s your role.
Then, of course, you give it the Instructions for what you want it to do. But you also give it Context, which is a lot of the knowledge files that you just mentioned. You give it Constraints like, “Hey, if I use ‘actually’ seven times, pare that down to just one or two.”
And then the E is Examples. That’s the most important part—that’s the part that most people skip, because the AI doesn’t know what “good” looks like to you. You’ve got to show it. So I love that you brought that up.
Quality Over Time Saved
Mike Allton: You’ve got this running. How has your actual week changed? What used to eat up that time before Claude and Descript were doing the heavy lifting?
Rich Brooks: I would say that it’s less about how much time I’m saving—although I am—it’s more about improving both the quality and the quantity of the outputs.
In the past, we shared the transcript, but we never shared my takeaways. I never wrote a blog post until I could get AI to draft it for me. And one of the things that I’ve noticed is—and I don’t understand why this is, because I’m like, I put the full transcript up there, why doesn’t it rank as well as my blogs on the Flyte blog? My theory, cause I have no proof of this, is that the transcript is exactly what we said, so it meanders and jumps all over the place. Even if there’s enough good content in there, it’s not really structured in a way that Google can make a lot of sense of it, right? Even AI might struggle with it.
But when I turn it into that blog post that breaks down the real big takeaways in a structured format like we’ve done, then all of a sudden, it’s much easier for these machines to understand it. I will say that our search traffic—and I can’t say that it’s directly related to this, but honestly, you can kind of see the uptick—we’re getting so many more high-priced, high-ranking keywords for different podcasts that all were done using this methodology with this blog post style approach.
We still have the full transcript in there for people who want it, but I feel that this is a much better outcome for Flyte, for Agents of Change, and for any of our listeners out there who want to check that out. So it’s been a big win in that area.
Has it saved me time? Absolutely, it saved me time. But for me, that’s the gravy. The most important thing for me in anything I do with AI, at least related to business, is: Am I improving the final output? If I get it done faster, great, but that’s not my number one goal. For me, it’s always, Is it better than it was when I was working by myself? And I would have to say that based on that decision-making, it definitely has, and it’s well worth investing a little bit of time upfront to create a tool—really an app—that does this work for me.
Mike Allton: Yeah, and I loved your LinkedIn newsletter example, cause I know at least from my own experience with my own podcast, if I had to write the LinkedIn newsletter myself, I wouldn’t do it. I wouldn’t get to it. Even if I’m copying and pasting a newsletter or blog post that I’d already written, I’d forget; I wouldn’t do it.
What I’ve done is my Claude instance writes the newsletter as though it’s from Claude, as though it’s from my AI chief of staff. So it’s fully transparent that the AI chief of staff wrote it—it’s its opinion on the interview and the blog post. And because Claude has access to Chrome, it pushes it to my LinkedIn. I just have to review it and then hit publish or schedule, whatever we want to do.
So bottom line, Rich: what’s actually different on the social side of your business now that it runs without you sitting down and having to write every single post by hand?
Rich Brooks: Mike, if I’m being completely honest, it hasn’t changed the social aspect. I mean, LinkedIn obviously—talking about getting 1,300 new subscribers—this has also freed us up to start doing, now that we’ve got the video clips, we’re putting them up on YouTube. And the shorts we’re doing on YouTube, we’re getting hundreds, sometimes a thousand views on these clips.
I don’t exactly know how that translates into real attention or real money, if I’m being honest. I put more emphasis on the podcast as social media and the LinkedIn newsletter as social media. In that case, it’s been fantastic.
As far as, like, our Facebook posts from Agents of Change—the organic posts—they haven’t gotten traction in years, and I feel like I’m checking a box. I know that’s probably—on this show, I shouldn’t say it, but I’m being honest: I don’t think anybody cares about Agents of Change podcast on Facebook, even if they’re following us. The algorithm’s keeping them from seeing it, so I have to find other ways to reach them. And the newsletter itself, and then getting subscribers to the podcast—those are the ways that I’m really moving the needle.
Mike Allton: Yeah.
Rich Brooks: Which I think—your mileage may vary. You know, like especially if you’re in B2C, you may have a much different experience than I do, but I would say the majority of our business is B2B, so for me, it’s primarily about LinkedIn and YouTube.
Mike Allton: That’s exactly where I was going. Yeah, I don’t expect to get much in—I don’t get much from Facebook or Instagram or TikTok. You know, but that’s all right. It’s a click of a button in Agorapulse to push the reels to those channels and, you know, if something happens, great. But I’m focused on LinkedIn and YouTube similar to you, and I love that you’re seeing better results from there.
Where to Find Rich Brooks
Mike Allton: You know, more and more views on YouTube will absolutely help you. Rich, this has been so cool. Thank you for sharing this behind-the-scenes look at your Claude setup and what you’re doing. For folks who have got more questions for you directly, where would you like them to go?
Rich Brooks: If you want to talk to me, my favorite platform is LinkedIn. I am at The Rich Brooks there. I’m pretty easy to find if you look it up that way. If you like podcasts—and I’m guessing you do—check out the Agents of Change podcast. We talk about all aspects of digital marketing there.
Mike Allton: Fantastic. Thanks, Rich. That’s a wrap for today’s lab session. If you want the exact system prompts, tool sequences, and the step-by-step blueprints that we broke down in this episode, hit the link down in the description below to grab the free episode download. And if you found this technical breakdown valuable, make sure to subscribe on YouTube or leave us a review on Apple Podcasts.
Let me know what you think—it helps other folks find us. See you next Monday.
That’s how real practitioners are winning with AI on social. The AI Social Playbook is brought to you by Agorapulse and archie.app. Make sure to subscribe to the channel on YouTube so you don’t miss a single hands-on session, and grab the link in the description for the complete episode blueprint. See you in the lab!
