You open a fresh chat, type a prompt you half remember from last week, and get back something that reads like a press release from a company you have never heard of. So you rewrite it. Then the rewritten version sits in that chat, uncopied, while three more days go by.
Loren Bartley, AI marketing educator and founder of Impactiv8, had both halves of that problem and fixed them with the same move. She stopped prompting and started building named skills inside a single Claude project, each one holding a standard rather than an instruction. There are now over 150 of them. In this session, she walks through the five that touch her social content, shows the Notion hub they write into, and stages a week of posts into Agorapulse without leaving the chat.
What changes is where your attention goes. Loren still reads every post and still approves every one before it schedules. She just stopped being the person who copies text out of a chat window and pastes it into a scheduler.
What you will learn
- A skill holds your standards as well as your instructions, so every post runs to the same bar instead of drifting with whatever you happened to type that day.
- A stateless chat has no memory of the positions you have already taken, which produces contradictions across a month of posts that you will not catch because you did not write them.
- Naming each skill after its job turns a list of 150 into something searchable: Audrey Ads Auditor audits ads, Hugo writes hooks, Caprice Caption Composer writes captions.
- Feeding refined versions of your past talks into a knowledge folder, rather than raw transcripts, keeps your ums, tangents and dead Q&A out of the source material your posts get built from.
- Putting one skill between drafting and scheduling catches format errors before they ship, including a post briefed as video that has a still image attached.
- Posts arriving in a scheduler as drafts rather than live posts preserve the human approval step without adding a copy-paste step back in.
Timestamp index
00:00 Why another post stopped feeling like a chore
00:52 The two bottlenecks that had nothing to do with writing
02:17 Why a fresh chat every time produces contradictions
03:41 What a skill holds that a saved prompt does not
05:33 Vague instructions scale into vague output
06:22 Inside the Content Accelerator project and its knowledge files
07:52 The authority folder and its 500 pieces of past content
08:12 Trini rewrites her old talks before they enter the knowledge base
09:29 Constance builds the quarterly content reference guide
10:19 Why Claude is the only LLM running skills inside a project
11:38 Edison turns the strategy into a weekly cycle brief
13:19 Eight minutes of her time produced the whole week’s plan
15:09 Naming 150 skills so she can find the right one
16:22 Connie assigns Caprice, Vivian and Hugo to the batch
17:51 Hubert’s Notion hub replaces posts buried in chats
20:23 What the hub tracks, and the 50,000 chats it replaced
21:13 Agatha stages the approved rows into Agorapulse
23:17 Agatha catches the still image on a video post
24:53 Why the posts land in Agorapulse as drafts
26:22 What Loren does with the time she got back
28:04 Where to get Edison
Neither bottleneck was about writing
Ask most marketers where AI breaks down, and they will point at output quality. Loren points somewhere else. She had two problems, and she is blunt that the second one embarrassed her more.
The first was rewriting. Generic material kept coming back, carrying none of her expertise and none of her voice, and at some point the maths stopped working. “What’s the point of even doing this if I have to rewrite it?” The second was that the content she did finish never shipped. It sat in chats she could not find again, or it waited on a round of copying and pasting she describes as joy-zapping. So it kept getting pushed to later, and later never arrived.
Those are operational failures, not creative ones. That distinction matters, because the fix for bad output is a better prompt, and the fix for content that never leaves the chat window is a system that carries it out the door.
A fresh chat every time produces contradiction, not just repetition
The obvious cost of opening a blank chat for every post is that you repeat yourself. Loren names a worse one. A model with no memory of what you published three weeks ago can take the opposite position this week, and you will not notice, because you are not the one writing.
Her framing of what that costs is worth sitting with. Authority is not one clever post. It is “a body of work that’s clearly going somewhere,” and repetition only helps when it builds on the same concepts instead of firing off opinions that drift. The system has to read what came before.
She pairs this with the reason saved prompts do not solve it. You write a good prompt once; it disappears into the scroll, and next time you cannot remember it or cannot face hunting through a prompt library to find it. So you improvise, your inputs get inconsistent, and your outputs follow. Her summary of where that ends: “AI doesn’t fix the vague brand; it scales it.”
What a skill holds that a prompt does not
A skill carries the instruction, same as a prompt. It also carries the standard, which is the part that survives between sessions.
Loren gives a concrete example of what that buys. A hook lands well, and she realises her audience responds to educational openers. She does not have to remember that for next time, and she does not have to find the prompt where she wrote it down. She opens Hugo, the skill that writes her hooks, and tells him to lean further into educational hooks. Every post after that inherits the change.
Mike makes the point that this is not a Claude-only pattern. A custom GPT, a Gem in Gemini and an agent in Copilot all work the same way: build it once, edit it when it drifts, scale from there. The difference he flags comes later, and it is about where the skill can run.
The knowledge layer, and why the transcripts get rewritten first
Loren’s Content Accelerator is a Claude project, and the knowledge sitting in it comes in two parts.
The first is core: brand voice, brand DNA, personal DNA, ideal customer, thought leadership themes covering both what she talks about and what she refuses to, a story bank of the things that are uniquely hers, and examples of how she actually phrases things. The second is a folder holding over 500 pieces of her past work. Evergreen blog posts, with the stale ones retired. Podcast appearances. Coaching sessions and training videos from her academy.
The step most people skip is what happens to that material before it lands in the folder. Loren has a skill called Trini Transcript Transformer. Trini takes a talk or a podcast where slides exist, uses the slides as the structure, blends the two, and strips out the ums, the tangents she went off on, and the Q&A that added nothing. What comes back is a transcript of the presentation she would have given if she had given it perfectly. “That’s what I feed into this folder. Not the raw material, but the refined material.”
Anyone can dump a Descript export into a project folder. The result is a knowledge base that has learned your verbal tics alongside your expertise.
Constance, Edison, Connie: the handoff chain
Three skills run the planning, and each one hands off to the next.
Constance Content Cornerstone runs about once a quarter. She interviews Loren, at length, and produces a content reference guide covering cadence, operating rhythm, weekly content pattern, content mix and ratios, pillar content standards, platform standards, calls to action, offer and theme mapping, promotional defaults, proof and stories, repurposing rules, guardrails and language conventions, content record defaults, and publishing approval benchmarks. Loren updates it quarterly as things change and as she learns what worked.
Edison Editorial Executive takes that guide and runs weekly. She maps the arc of what Loren will say across social and the rest of her marketing for the week ahead, and outputs a cycle brief. That brief gets saved, so the next cycle knows what came before. Edison can plan up to eight weeks out; Loren caps it at one or two, on the grounds that she is not organised enough to plan further.
The cycle brief she pulls up on screen names the current period as replay promotion after her AI Maker class, then answers who this is for, what the calls to action are, what the proof is, what needs to shift in people’s minds, what the arc is, and what the production briefs are. Building it cost her eight minutes of attention while she cleaned her office, spread across roughly half an hour of elapsed time. Loren is direct about what that replaces: “it’s doing what a lot of people just don’t ever get around to doing.”
Connie Content Commander takes Edison’s plan and assigns the work. Caprice Caption Composer writes captions, Vivian Video Verbatim writes the long-form video scripts, Hugo gets pulled in when a hook needs sharpening. Before Connie existed, Loren called each skill herself and wrote posts one at a time, which she describes as part of the nightmare. Connie is the conductor.
Why the skills have names like Audrey Ads Auditor
Loren has over 150 skills and freely calls the names weird. They follow a pattern: alliteration, with the job baked into the name.
She built it that way because she is bad with names. If she wants to audit her ads, she knows it is Audrey Ads Auditor, because the answer is sitting in the name. At 150 skills, that stops being a quirk and starts being the retrieval system.
Mike adds the mechanic that makes it work in practice. Claude knows which skills are installed, so it can often pull the right one into a conversation on its own, and you can always call one by name. No button, no menu, no going into settings to pick a skill first. You say you want Connie and Connie shows up.
He also flags the structural difference. Projects exist across the major tools, called projects in Claude and ChatGPT and notebooks in Gemini and Copilot. As of this recording, Claude is the only one where a custom instruction assistant runs inside one of those containers. That is what lets Loren’s skills read her knowledge folder while they work.
The Notion hub that replaced 50,000 chats
Connie writes the finished batch into a Notion database that another skill, Hubert, set up. Notion connects to Claude over MCP, which took Loren very little effort to configure.
The problem this solves is specific. Content produced in a chat is stuck in a chat, and it is not editable text. Changing one word means a new round trip, which Loren points out is harder than just typing the thing yourself. In the hub, she opens the row and edits the copy in place.
Each row carries the platform variants, so a LinkedIn version and an Instagram version sit together with the alternatives Connie offered. It also tracks format, offer, platform, status, the post URL once it goes live, and the image. “This previously to me was in 50,000 chats all over the place and messing with my head.”
Mike’s caveat on MCP is worth keeping in view. An MCP connection lets the AI tool do roughly everything you can do in that platform, which is a step past the hook-and-trigger model most of us know from Zapier. That reach is the benefit and also the risk.
Agatha’s job is to tell Loren what she got wrong
Once Loren moves rows to approved, she calls Agatha, her Agorapulse assistant, from the same chat. Agatha could simply push the approved rows into the scheduler. Loren built her to do something else first.
“I believe the human in the loop is so important, and I make a lot of mistakes.” So Agatha fetches the Notion entries, checks what is approved, and reports back before creating anything. In the session, she returns four approved rows: three of them stageable, one that should not be touched, and three things to look at.
The first is the one that lands. The row briefs video in the visual brief, and the only asset attached is a still. Loren checks, finds she attached an image to a post that should have been a video, and says so on camera: “I made a mistake. Look, I attached an image, and I’ve been called up on it, and that’s what I love.”
She resolves it in conversation rather than in a settings panel. She tells Agatha to skip post one, proceed with post three, and leave out the post about Marietta because she had already scheduled that one. Then she hits return.
Most AI assistants are built to agree. This one is built to hold the work back, and that is the part of Loren’s stack that is hardest to copy from a screenshot.
Drafts, not live posts
The posts land in Agorapulse carrying the date Connie assigned from Edison’s brief, and Loren switches them to save as a draft. A human still looks at every one before it goes out.
Mike calls this the best practice in the whole workflow, and the reason is that it costs nothing. The approval step survives without reintroducing the copy-paste work that made Loren stop shipping in the first place.
What the time bought
Loren used to publish almost no video, because she did not have time for it. Vivian Video Verbatim now writes the script of exactly what she will say, and the rest of the video work takes less than it used to, so video is back on the table.
The answer she gives last is the one worth ending on, and it is not about hours. Most people carrying a content calendar know the feeling of another post coming due. Loren says that feeling is gone. She is having conversations with her AI team members; what comes out is better than what she produced before, and the thing she names is energy rather than time.
Loren is giving away Edison
Edison is the skill that turns Loren’s quarterly strategy into a weekly cycle brief, and she is sharing it. Constance normally feeds her the strategy first, but Edison is trained to work without that input. She will just ask you a great many more questions the first time, and store the brief so the next round moves faster.
Get Edison from Loren
Her editorial executive skill, the one that builds the weekly cycle brief. Grab it here.
Grab the Episode 6 Blueprint
The full roster build: the four planning roles, the gate role that checks the work, what each one owns, and the handoffs between them. Download it here.
Connect
CONNECT WITH LOREN BARTLEY
Listen and subscribe
Full episode transcript
Loren Bartley: Most people listening to this will probably think, “Oh, my God, I’ve got to do another post.” I don’t have that feeling anymore. I actually really enjoy it. I’m having conversations with my AI team members, and we’re bouncing ideas off each other.
What’s coming out is better than what I would have produced before. It’s not just the time it’s giving me back; it’s also the energy it’s giving me back to put into my business. And I absolutely love that.
Mike Allton: Welcome back to The Lab. I’m Mike Allton.
Generic prompts produce generic social posts that sound like everybody else’s. Today, Loren Bartley, AI marketing educator and founder of Impactiv8, shows us how to configure modular Claude skills and role-based personas that draft on-brand social content directly into Notion.
Hey, Loren. Welcome to the show.
Loren Bartley: It’s so great to be here. Thanks, Mike.
The Two Bottlenecks
Mike Allton: My pleasure. Now, before you built out custom skills in Claude, what was the biggest bottleneck in your daily content creation?
Loren Bartley: I probably had two major problems, and neither of them was really about creating the content itself.
The first thing I was finding was that, even though I was using AI, I was still rewriting quite a fair bit. There was a lot of generic internet content being put into the drafts. None of my expertise, none of my voice. And when that’s happening, at some point you’ve got to ask yourself, “What’s the point of even doing this if I have to rewrite it?”
So that was one issue I was having. And probably the bigger, more embarrassing one was that I was creating content, but it was just never getting out the door.
It was either lost in chats that I couldn’t find, or I was spending so much time cutting and pasting. That was the job I was left with after AI had helped me create the content. It was a joy-zapping task, and I really didn’t want to do that.
So I’d think, “Oh yeah, I’ll get back to that later. I’ll get back to that later.” And, combined with not knowing where the hell anything was, I was often creating the content, but it just wasn’t getting posted.
That’s where I started developing skills and connectors to fix these problems. And that was an absolute— I hate the word—game changer for me.
Mike Allton: Every once in a while, these clichés ring true. Now, I know a lot of marketers who are using AI, but they’re just opening up a fresh chat every single time and typing in some quick prompts. Why does that approach break down when you need this kind of consistent, high-volume social content?
Loren Bartley: Well, there are a few reasons, but one of the ones I think is probably most important is that it has no context of what’s gone before.
What ends up happening is that you get a lot of repetition, or worse still, you could get contradictions because you’ve got AI creating the content for you, but it’s creating it from scratch every single time.
You might take a position on something in one post, and then three weeks later, you might take a completely opposite position. You might not notice that this is the case because you’re not actually writing the content.
What you really need is a system that reads your past content and builds on it so that you can start demonstrating your authority.
Authority isn’t just one clever post that you write. It’s more of a body of work that’s clearly going somewhere. And yes, repetition is good, but repetition needs to build upon concepts consistently, rather than randomly shooting off varying opinions that are completely different.
So that’s one of the issues.
What a Skill Holds That a Prompt Does Not
Loren Bartley: And then another issue that I find is that if you’re typing out a prompt over and over, you put that prompt there once and then it’s lost. Next time, you can’t quite remember it, or you’ve got to go searching for this prompt in some prompt library that you’ve got somewhere that you probably can’t be bothered finding to do the work.
And so then what happens is you start getting inconsistencies in the inputs that you’re putting in yourself, and therefore that’s going to again lead to inconsistencies in the outputs that are coming out.
If you can develop a skill instead, a skill is not only the instructions of what you want to do, in the same way that a prompt is, but it also holds your standards and makes sure that what you’re creating is running the same way every time, to the same standards. That makes a massive difference in the outputs that you’re able to produce over time.
And so, just to give you an example of this, if I do a post and I find that the hook lands really well, and I’m like, “Oh, people are really intrigued by these educational-type hooks,” then I don’t have to remember that for next time that I’m going to do a post.
I can either go and create a skill that does that, or if I’ve already got a skill, which I do—Hugo, he writes my hooks—I can go and update Hugo to lean more into putting more educational hooks into the content that’s being created.
So this makes a big difference when you’re working at volume because, you know, you might have one or two posts that are kind of varying. That can be a problem when you get to volume because then what happens is you start scaling those things that are not particularly your opinions, directions, or information that you’re trying to get across, and it’s harder to catch at volume as well.
So I suppose it comes back to, you know, what is a bit of a cliché again: AI doesn’t fix the vague brand; it scales it. So if you’re giving it vague instructions, you’re going to get vague outputs at volume. And that’s kind of what happens when people are just quickly firing off a prompt here and there.
Not to mention all the extra work that they’ve got to do if they’re only doing these one at a time.
Mike Allton: Right? Now, I love this so much because you’re absolutely right. If you take the time to develop a skill, or if you’re using ChatGPT, it can be a custom GPT, or a Gem in Gemini, or an agent in Copilot. They all function very similarly.
But you take that time to develop it, and you find that it’s putting in too many emojis or it’s doing something else that you don’t want it to do, just go edit that skill, Gem, or custom GPT and make sure that it doesn’t do that anymore. Then you can scale up, just as you said.
Inside the Content Accelerator Project
Mike Allton: So let’s go ahead and dive into your workspace. Can you pull up your screen and kind of show us how you’ve got your custom Claude project and skills configured?
Loren Bartley: Yeah, definitely. This is my Content Accelerator, and I’ve created this as a specific project for creating my social media content. But not just my social media content—it also does my blogs and my emails as well, all in the one sweep. But I’m going to focus on the social media side of things today.
So, in terms of what’s in here, the first thing is the knowledge side of things. You’ll notice on the side here, in the project itself, I’ve got my Impactiv8 Knowledge, which is my core knowledge.
So this is my brand voice, my brand DNA, my personal DNA, which is more about me because there’s a lot of me in the content that I create. My ideal customer, my thought leadership themes—so the things I do talk about and things I don’t talk about—and the balance of what I want my content to be created around in that regard.
Things like my story bank, so the things that are really uniquely me and my stories, so they can be weaved in there. And some examples of how I actually say things, so that’s picked up in the content that I’m creating as well.
So that’s one side of the knowledge that I have in here. But the really important thing is the Authority Content Knowledge. And this is when we’re talking about social media content creation, so I don’t sound like everyone else and just, you know, say things the same way as others, and have the content sounding like just generic information that could have been pulled off the internet.
I’ve got this folder full of over 500 pieces of content that I’ve created in the past. So I’m talking about all the blog posts that I’ve written that are still relevant, like the evergreen ones. I have retired quite a few of them. Any podcasts that I might have been on, a tidied-up transcript of that, and any of the coaching that I’ve done.
And in all of them, you have—you give away so much knowledge in there, which is kind of captured behind a paywall, but is obviously stuff that you could potentially be putting out there to demonstrate your expertise a lot more broadly. So that’s embedded in there as well.
And I have a really cool tool that I use, and her name’s Trini Transcript Transformer. And what Trini does is she takes my presentations or my podcasts I’ve delivered where there’s slides associated with them. She then uses the slides as the format and effectively blends the two together, gets rid of all my ums and ahs, and maybe the side quest I just went off on—a tangent that wasn’t relevant to the topic—or the Q&A that was asked that didn’t really add value to that presentation.
And then she gives me back this transcript of what would have been the perfect presentation I could have given on that topic. And that’s what I feed into this folder—not the raw material, but the refined material.
So that’s one part: the knowledge. And it’s a really important part. And the other part is the skills. And so, in this particular Content Accelerator, there’s a few skills that I use.
The first one is—and I’m going to trip myself up on all their names, because you might have realized I like to use a bit of alliteration there.
Constance, Edison, and the Weekly Cycle Brief
Loren Bartley: And actually, let me show you all my skills, because this is going to blow your mind a little bit. Here she is: Constance, Content Cornerstone. I really only use her probably once a quarter.
What Constance gives me at the end of this is, she interviews me, asks me all this information about—and as you can see, it’s quite extensive—and she comes back with a strategy for me in terms of how often I want to be posting during non-launch periods, the framework around what channels I want to be on, all that kind of stuff.
So she produces this output that really becomes the cornerstone framework that I use from a strategy perspective with all of my social posts. And I go and update that probably once a quarter, as things change and I learn from what’s worked and what hasn’t worked.
Mike Allton: I want to interject real quick for everybody who’s listening. Number one, if you haven’t been listening to this show for a while, or maybe this is your first episode, I want to encourage you, if you’re listening on Apple or Spotify or someplace where you can’t see the video, hop over to our YouTube channel because Loren’s sharing her screen.
I’m going to have her share the anatomy of one of her skills here in a second, and you’re not going to be able to see that. It’s going to be harder for you to follow along.
The second thing I want to clarify real quick, again, for everybody watching, because a moment ago we were talking about skills, and then you shifted into the project. I want to make sure folks understand the difference.
Projects in Claude are like digital binders or notebooks where you can insert files, you can add additional instructions, and that sort of thing. And that same functionality exists in Gemini with notebooks, in Copilot with notebooks, and they’re called projects also in ChatGPT.
But there’s one huge fundamental difference that Loren’s talking about here that, to my knowledge, none of the other systems do. Because I said you could do a skill as a custom GPT or an agent or a Gem, right? None of those others work inside of their project or notebooks.
Claude is the only LLM as of this recording that I’m aware of that allows you to use their custom instruction assistant, which they call a skill, inside of one of these digital notebooks.
So that’s what you’ve done here: you’ve set up one or more projects, and you’re using multiple skills inside them. So definitely tell us more about Edison.
Loren Bartley: She is probably one of my favorite skills when it comes to social media creation. And Edison is my editorial executive, so Edison Editorial Executive, and I work with her usually on a weekly basis.
And what we do is we work out the arc of what I’m going to be talking about on social media and across all of my marketing during the week ahead.
So for Edison, basically what I’ve done with her is—she is a—I will show you this because I think it’s important. I’m—because I didn’t mention this before, but I quickly found what Connie gave me.
So Connie gave me this Content Reference Guide as the first step. She basically spits out a document that has the scope and review of my cadence, my operating rhythm, my weekly content pattern, my content mix and ratios, my pillar content standards, my platform standards, my calls to action, the offer and theme mapping, promotional defaults, proof and stories, repurposing rules, guardrails and language conventions, content record defaults, and publishing approval benchmarks.
That’s what the first one does from Constance, which is pretty extensive, and then she hands that off to Edison.
And so what Edison then does is output for me a cycle brief. And as I mentioned before, I’ll do this, let’s say, in a weekly cycle, but she’s trained to be able to do this from, like, if anything, up to eight weeks is what she can produce if you wanted to do that much.
I’m not a fan of that. I’m kind of like one or two weeks max because, you know, things—I’m not that organized to be planned any further in advance.
But what she’ll basically do is come out and produce me this cycle brief, and that cycle brief gets saved so that in the future, we know what’s gone before. And it’s based on what Constance outputted that I just showed before.
And so basically here she’s going to say, you know, what’s going on at the moment. So I just had my AI Maker Class that I do every couple of months last week, and so I got her to do this on Friday.
She put this together for me, and this is basically coming up with the plan as to what we’ll do. So, you know, what period are we in at the moment? And so we’re in the replay promotion period. And, you know, what we’re building towards.
And then, you know, she’s interviewed me. So this whole chat basically, you know, doesn’t go on that long, which is a good thing because, you know, it doesn’t take a lot of my time. It took me, in total, to do this about eight minutes of my time whilst I cleaned my office. So I was… It probably took about, I don’t know, half an hour all up, but I was off doing other things.
And then it basically came through and, you know, who’s this for? What are the calls to action? What’s the proof? What needs to shift in people’s minds? What’s the arc of what we’re trying to achieve during this period? You know, what are the production briefs?
And then, so this is through a conversation with me. It’s doing what a lot of people just don’t ever get around to doing. They’ll just fire off these random posts, and they won’t look at the strategy.
And this is the really important part of this: it’s allowed me to get more strategic in what I’m doing by layering in, you know, an extra team member in Edison who does this bit for me.
So she’s probably, I think, one of the most valuable skills that I’ve got, and I absolutely love her. And then she hands it off to Connie.
Naming 150 Skills
Loren Bartley: And so I showed you my skills before. Here they are. There are over 150 of them that I’ve got, and they’ve all got weird-ass names.
Mike Allton: Well, it makes me feel better because I thought I had too many, so that’s good.
Loren Bartley: No, no. And it makes sense to me because—
Mike Allton: Yeah…
Loren Bartley: I’m one of those people who’s hopeless at remembering anyone’s names, and I kind of wish that everyone was named this way because it makes it so much easier for me to know. So, you know, if I want to audit my ads, I know it’s Audrey Ads Auditor that I need to bring onto my team at that point in time because it makes sense. It’s in her name.
Mike Allton: And the nice thing about Claude, for those watching, is Claude is aware of the skills that are installed. And when you start having a conversation with Claude, it can sometimes know exactly which skill to bring into the conversation depending on what you have installed, or you can reference it by name.
You don’t have to click a magic button. You don’t have to go into Customize and choose a skill to use it. You can just say, “I want to bring in Connie or Edison,” or whatever you’re calling your skills at that moment, and it automatically brings it into the conversation.
So what do we do next in your workflow when it comes to actually creating social media copy?
Loren Bartley: Yeah. So the next bit, now that we have the plan, is to say, “Okay, now we need to put the team to work.” So, of those 150 skills, which ones are going to do what?
Initially, what I used to do is I would go and call each of the skills as I thought I needed to use them, and I would be creating the posts one by one. And it was just, you know, part of the nightmare.
And so what I did is I created a new skill called Connie Content Commander. And what she does is she coordinates the weekly batch of content based on what Edison has said that we’re going to be doing, which is—Edison came up with that plan in conjunction with me.
And then what Connie does is she says, “Okay, we’re going to get Caprice Caption Composer to write the captions. We’ll get Vivian Video Verbatim to write the long-form video scripts,” and then she coordinates that.
I don’t even have to think about who should be there. As you said, the way that skills are created for us means we don’t even have to go and call the skill just because of the thing that we’re wanting to do. Claude’s smart enough to then go, “Okay, this is the skill. I need Hugo because we need a better hook for this.”
So she’s basically my conductor, really.
Mike Allton: Hmm…
Loren Bartley: That commands all of the team members to actually go and do the work. And the way that she then presents this to me is that I have used skills again.
The Notion Content Hub
Loren Bartley: I have this skill called Hubert Hub or whatever. It’s not Creator. It must be another H word. But anyway, Hubert has set up for me this really, really cool—and it might overwhelm a few people—but let me just show you this Content Hub.
I had this problem where things were in random chats. And not only were they in random chats, but they’re in chats as uneditable content. And if you have to go back and say, “Oh, change this word, change this word, do this, do this,” it’s so much harder than just typing it yourself.
So one of the things that Connie does is, once she’s got that information back from the team and she’s compiled the content together, it basically gets put into my Content Hub, where I can then go and edit the content.
Mike Allton: You’ve introduced a new tool here.
Loren Bartley: We are. This is Notion now.
Mike Allton: Notion, yeah. And this is connected directly to Claude via MCP?
Loren Bartley: Yes. So that was super, super easy to do.
Mike Allton: And because Claude has this MCP connection, you don’t need to know what MCP stands for. Just understand that when it’s an MCP connection, the AI tool can do pretty much everything that you can do.
It’s a lot more flexible than what a lot of us have been used to over the years with, like, a Zapier connection, where you might have a hook and a couple of trigger options, and that’s all the tool can do.
You let Claude in via MCP, it can do pretty much everything that you can do, which has its own caveats and risks, so be aware of that. But it’s pretty amazing because then you can tell Claude, “Hey, we have this new piece of content. Push it into the Content Hub.” I don’t know if that’s exactly the words that you use, but that’s basically what—
Loren Bartley: Yeah, totally. I also have skills where I don’t even have to say that. So, using Connie as the example, Connie knows that once the content has gone through all of the checks it needs to go through to get to the standard that I want it to be before it ends up in here, she presents it to me in the chat with a link to where I will find it in the Content Hub.
So then I can go into there and start doing my editing in there, which is the really good thing about it because I often have variants in terms of how I might want to post to Facebook or Instagram. In here, there’s a LinkedIn version, there’s an Instagram version, and then she’s also given me some alternative options that I might want to choose.
So I can come in here, and I can add an image URL or whatever else I want to do in here to change this up before I go and schedule the content.
So this becomes my workplace. This also gives me a whole lot of other information.
So, because I’d already told Edison that I wanted to post to Facebook, Instagram, and LinkedIn, it’s come in here. It says what the format is, it shows what the offer is, it shows the platform that I’m—it shows the status. If it’s gone live, the post URL can get put in here. I can put my images in.
This is what is currently in production, to some extent, for this current campaign.
And this previously, to me, was in 50,000 chats all over the place and messing with my head. So now I know I just have to come here. I just have to open up the post that I want to work on.
So this one’s in draft. I go in here, I make the changes that I want to make, and then, effectively, the next thing I do is I can then go and schedule them.
Staging Into Agorapulse
Loren Bartley: And so that allows it then to be written to my scheduling tool of choice. Do you want to see how I do that?
Mike Allton: Please do, yes.
Loren Bartley: Okay. So, you’d see I’ve moved these to approved, and it means that I’ve looked at them, I’m happy with them, and I’ve approved them to go to my scheduler.
And so basically what I would do then is, in the chat I was in—which would usually be the chat with Connie, where we’ve come up with the plan, we’ve chatted backwards and forwards, and then I’ve gone in and made my edits, and I’m totally happy with them—in that same chat, it’s as easy as going, “Run,” or I can actually do a forward slash and then call Agatha in if I want.
Or I could have said, “Can you now post these to Agorapulse for me?” for example.
And so what Agatha’s done—she’s my Agorapulse assistant—is she’s now going to go and put these into Agorapulse, with Facebook, Instagram, LinkedIn, whichever variations are going to get put into that tool. But she’s going to do a bit more than that.
Loren Bartley: And you don’t have to build it this way, but I always believe the human in the loop is so important. And I make a lot of mistakes.
So I’ve actually built Agatha so that Agatha doesn’t just go, “Yeah, cool, I’m going to take it and I’m going to put it in there.” She comes back and tells me what I’ve done wrong, which is pretty cool.
So, like, she may come back and say to me, “Oh, the image description…” because one of the things that Connie coordinates is that it tells me this is the type of image that would go really well with this post.
She may come back and say, “Oh, it said that it should be an image of Lizzie, who’s giving the testimonial, but it’s a picture of you at a computer, putting your headphones on.”
And I’m saying this one because she actually gave me this exact feedback once. And so in that case, I can go, “Oh, no, that’s fine. You know, that’s what I wanted,” or I can go, “Oh, my God, yes, I’ve attached the wrong image.”
So I like putting those checks and balances in.
Loren Bartley: And so basically what’s going to happen here is she’s fetching the Notion entries, looking to see what’s approved and then she’s, you know, here we go.
So she’s come back: “Four approved social rows in the hub, three are stageable, one shouldn’t be touched.”
Here we go. So she’s found a reason, and then down here she’ll give me some feedback.
“Three things before I create anything. Post one: format versus media. The row says video in the visual brief, but the only asset is a still.”
Okay, so I attached an image to something that should have been a video. So if I go back in here, I should have one that says video, video, video. Yes.
Loren Bartley: I made a mistake. Look, I attached an image, and I’ve been called up on it, and that’s what I love. I love that I’ve been called up on it.
So I might just have a conversation with it here and say, “Don’t post one. That was my mistake. Go ahead with post three, and let’s not put the one about Marietta in there either because I didn’t realize I’d already scheduled her. Let’s just say…” And then I can just hit return, and now it’s going to go and post those things into Agorapulse.
And this, sometimes depending upon how much content you’re creating, it might take a few minutes. And then we’ll come back to see if it’s actually pushed across.
But what it’s basically done is it’s posted it across here into Agorapulse, and then all I need to do is change that to “Save as draft.” It’s already got the date because I had that in Notion as to when it was going to be scheduled.
Mike Allton: Which I didn’t come up with.
Loren Bartley: Yeah. That’s basically what Connie came up with based on the brief that Edison had given her.
So that, to me, is super, super, super, super cool.
And then if I come back here and just refresh this now, I’m going to see those new posts will be in here. But I can then go and do the things that I just showed you, where I can make further edits if I see the need, or I can just go and post them now if I wanted to post straight away, or let them be scheduled at the time.
So, because I canceled the other two, only one post across the different platforms came in here.
But if I had a week’s worth of content, then a week’s worth of content would have populated in here.
I don’t know about you, Mike, but do you think that would be super useful for a lot of people out there?
Mike Allton: Oh my gosh, yes.
So let me just recap what we all just saw, because some of you might be thinking, “Wow, I don’t believe any of that.”
So first of all, she’s got Agorapulse as her social media scheduler connected via MCP, just like Notion, which means Claude can operate Agorapulse and pretty much do everything that we can do.
I love that you showed a best practice there, Loren, which is to have those social media posts show up inside of Agorapulse as drafts, still requiring that human to look at it.
Because everything that you’ve been talking about is essentially your digital crew. I love that you named all of these AI agents because they really are filling seats in your organization.
It’s not taking anybody’s job. This is work that you had to do before, but now you don’t have to, because you have these different AI assistants who are doing the jobs for you.
The Time and Energy Returned
Mike Allton: So my last question for you, Loren: now that you’ve got all these different AI crew members who are helping you deliberate on strategy, come up with content, push the content into Notion where you can edit it, and then pull it out of Notion and push it into Agorapulse for you, what have you been able to do with your time that you now have back?
Loren Bartley: Oh, well, to be able to put out more video myself, because I just didn’t have time for it. But I’ve got time back now, and I’ve got Vivian Video, who will write a verbatim script of what I’m to say in my video. And then I can put that out there.
And I can do that now because I have more time to do that, and because the video creation side of things, which I haven’t even touched here, is not taking as long as it used to take when I didn’t have these skills to call upon.
So that’s one side of it. Obviously, I can then go and focus on the other parts of my business that really need me.
Loren Bartley: More touch points with the customers, developing new products, and the stuff that really we should be spending time in our business doing.
Most people listening to this will probably have some feeling of, “Oh, my God, I’ve got to do another post.” I don’t have that anymore. I actually really enjoy it because I’m having conversations with my AI team members, and we’re bouncing ideas off each other.
What’s coming out is better than what I would have produced before. So I’m just feeling it’s not so much just the time that it’s giving me back; it’s also the energy it’s giving me back to put into my business, and I absolutely love that.
Mike Allton: This has been so cool. I know you’ve got a wonderful thing that you want to share with everybody who’s been watching, so why don’t you tell them about that, and then where they can go if they’ve got questions?
Loren Bartley: Yeah. So, I spoke a lot about Edison. She is one of my favorite team members. And if anyone wants to get their hands on Edison, then I would send you to aimegasuccess.com/academy.
So I’m happy to share that with everyone because she’s been that game changer for me and my business, and I’d love that to be the same for others.
Mike Allton: I love that. That is so amazing. Thank you so much, Loren. And folks, that’s a wrap for today’s session.
If you want all of the prompts, the tool sequences, and all the step-by-step blueprints that we broke down in this episode, as with every single link, hit the link right in the description below, and of course, Loren’s links and everything down there with it as well.
Mike Allton: And if you found this technical breakdown valuable, please make sure to subscribe to our YouTube channel. Leave us a review on Apple Podcasts, if that’s your preference, and it really helps other serious operators find the show.
I’ll see you next Monday.
Mike Allton: 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.
