Big Idea

The argument in one line.

A repeatable AI agent team, built from five specific instruction fields per bot, turns raw news into an approved X post, LinkedIn post, and Instagram carousel while a human only approves angles and final drafts.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…

A marketer or founder who publishes regularly and wants an AI system that turns news into drafts instead of writing every post from scratch.
Someone comfortable defining explicit instructions for an AI agent (role, goal, inputs, process, quality bar) rather than one-line prompts.
A solo operator or small team without dedicated writers who wants one repeatable pipeline across X, LinkedIn, and Instagram.

SKIP IF…

You want a fully hands-off, zero-review posting system; this workflow keeps a human approving every angle and every draft.
You don't work in a niche with a steady stream of news or research to react to.

TL;DR

The full version, fast.

Grok Bot can run a five-agent content team, a chief, a researcher, an X writer, a LinkedIn writer, and a slide designer, but the video's real argument is that writing was never the hard part. The harder work is deciding what to cover and finding the angle that stops someone scrolling. Each agent needs five defined fields: role, goal, inputs, process, and quality bar, with explicit rules against 'LinkedIn slop' and invented facts. In a live run, the researcher agent pulls the day's health news, a coordinator proposes angles, and the human approves one before the writer agents draft an X post, LinkedIn post, and Instagram carousel. About 70% of the pipeline is automated; the remaining 30%, taste and final approval, stays human.

Free for members

Chat with this breakdown — free.

Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.

Create a free account →

Chapters

Where the time goes.

00:00 – 01:18

01 · Intro: viral content with Grok Bot

Ben introduces the AI agent content team concept and previews a 70/30 automated-to-human-reviewed split.

01:18 – 02:10

02 · Why writing isn't the hard part

Argues LLMs already write well; the harder problems are picking what to cover and finding a scroll-stopping angle.

02:10 – 03:38

03 · Meet the Grok Bot agent team

Introduces the named agent org chart: Tecumseh (chief), Flock (researcher), Xemingway (X writer), L.I. Lewis (LinkedIn writer), Johnny Ive and SlideMaster (visuals).

03:38 – 04:12

04 · Creating an agent in Grok Bot

Walks through the Grok Bot chat UI: naming a bot, setting its label/role, and writing its description field.

04:12 – 05:01

05 · The 5-part agent instruction framework

Every agent needs role, goal, inputs, process, and quality bar defined explicitly, with examples wherever possible.

05:01 – 07:08

06 · Example LinkedIn writer agent prompt

Reads through a full L.I. Lewis instruction doc: mission, reader, hook rules, structure, style, anti-slop rules, and a trustworthiness/no-hallucination clause.

07:08 – 08:24

07 · Grok Bot interface, marketplace, and pricing

Tours bots, the shared computer each bot can use, the plugin/bot marketplace for borrowing other people's agents, and the $20/month price (free with X Premium+ or Cursor).

08:24 – 08:59

08 · Using AI to write agent prompts

Recommends dumping a rough idea into any bot and asking it to draft and improve the instruction prompt rather than writing it cold.

08:59 – 09:23

09 · Live workflow: finding stories with Flock

With the whole team built, Ben kicks off a live run by asking Flock to pull the best health and science news from the past 24-48 hours.

09:23 – 11:08

10 · Typeless voice dictation (sponsor)

Demonstrates the sponsor tool: press F1 to dictate instructions instead of typing, and highlight-plus-shortcut to summarize, translate, or clean up any on-screen text.

11:08 – 13:23

11 · Story shortlist and angle selection

Flock returns a curated shortlist of news stories; Ben picks one and Tecumseh proposes several angle options for the human to approve before drafting starts.

13:23 – 14:21

12 · Xemingway's X post drafts

Reviews draft X posts, critiques weak hooks for lacking context, and asks for a revision with clearer, higher-stakes hooks.

14:21 – 14:58

13 · LinkedIn post and Instagram carousel

Approves a LinkedIn draft from L.I. Lewis and reviews SlideMaster's tweet-style Instagram carousel, which also opens in Figma.

14:58 – 15:33

14 · Why he doesn't use auto-posting tools

Argues that signing social accounts into third-party auto-posters flags the account as a business account, which he believes gets its reach suppressed.

15:33 – 16:30

15 · The full newsroom workflow

Recaps the six-stage pipeline end to end: find the story, research and verify, generate angles, send the brief, review the work, ship the package.

16:30 – 17:35

16 · Where AI agent teams go next

Closes on the idea that the workflow still starts manually today but could run continuously, with agents doing more of the work as they improve.

Atomic Insights

Lines worth screenshotting.

Writing is no longer the hard part of content: most LLMs already write well, so the hard part is deciding what to cover and which angle stops someone scrolling.
A good AI agent instruction defines five things: role, goal, inputs, process, and quality bar, not a one-line command like 'write viral LinkedIn posts.'
This creator runs about 70% of his content pipeline fully automated; the other 30% is a human reviewing for taste before anything publishes.
Rewriting a post for a different platform means reconstructing it for that platform's reader, not just reformatting the same words.
An explicit 'avoid LinkedIn slop' instruction bans openers like 'I'm thrilled to announce' and forces specific, evidence-first writing instead.
Strong agent prompts include a trust rule: never invent a number, quote, source, or scientific conclusion, and preserve uncertainty instead of erasing it.
Voice dictation produces better AI outputs than typing, because most people speak faster and more naturally than they type instructions.
Signing an auto-poster tool into your social accounts marks you as a business account, and platforms cut organic reach for business accounts.
Grok Bot costs $20 a month standalone, but comes included with an X Premium+ subscription or a Cursor subscription.
Letting AI draft the first version of an agent's own instructions, then editing that draft, produces a better prompt than writing one from scratch.
Giving each AI agent a name, an avatar, and a personality makes it easier for a human operator to track which bot handles which job.
A researcher agent can run continuously in the background, so new story ideas are already waiting the moment the operator opens their laptop.

Takeaway

Five fields make an AI agent work.

PROMPT ARCHITECTURE

The video's real lesson isn't about virality, it's that a five-field instruction (role, goal, inputs, process, quality bar) is what separates a working AI agent from a generic one.

01Intro: viral content with Grok Bot

The pipeline behind a piece of viral content isn't one AI writer, it's a small team of specialized agents handling research, angle selection, drafting, and visuals separately.
Treating roughly 70% of the process as automatable and reserving 30% for human taste is a workable split for anyone worried AI content will feel generic.

02Why writing isn't the hard part

If your bottleneck is 'what should I write,' the fix isn't a better writing prompt, it's a system for surfacing and ranking timely stories in your niche.
The angle, the specific framing that makes someone stop scrolling, is worth treating as its own deliberate step, not something that happens automatically once you pick a topic.

03Meet the Grok Bot agent team

Splitting a content operation into named roles (researcher, chief, platform-specific writers, visual designer) makes it easier to reason about what's missing than one general-purpose 'content AI.'
You don't need every role on this list: a coordinator, a researcher, and one writer are the minimum viable team; visuals are optional.

04Creating an agent in Grok Bot

An agent's description field is where the real instruction-writing work happens, the name and role are just labels for a human to keep track of the agent.

05The 5-part agent instruction framework

Before writing a single prompt, define role, goal, inputs, process, and quality bar for the agent; skipping any one of these is where vague, generic AI output comes from.
Providing concrete examples in the quality-bar section is one of the highest-leverage additions to any agent prompt.

06Example LinkedIn writer agent prompt

A working platform-specific writer prompt states the platform's actual reader (intelligent, professionally curious) instead of a generic audience description.
Naming the specific phrases to avoid (like 'I'm thrilled to announce') is more effective than a vague instruction to 'sound less like AI.'
A dedicated trust clause, never invent a number, quote, source, or conclusion, and preserve uncertainty, matters more for a content agent than style rules do.

07Grok Bot interface, marketplace, and pricing

A plugin and bot marketplace where other users' agents can be installed and reused means you don't have to build every agent in a new system from scratch.
Whether a paid AI tool is worth it can depend on bundling: a $20/month tool may already be included in a subscription you already pay for.

08Using AI to write agent prompts

Asking an AI to draft and improve its own instruction prompt from a rough idea produces a usable starting point faster than writing the prompt cold.

09Live workflow: finding stories with Flock

A single, simple prompt to a well-instructed research agent ('pull the best news from the last 24-48 hours for X industry') can replace a recurring manual research task.

10Typeless voice dictation (sponsor)

Speaking instructions instead of typing them tends to produce longer, more detailed inputs, and more detailed inputs produce better AI outputs.
A highlight-and-ask shortcut for summarizing or explaining on-screen text is a small habit that compounds across a lot of daily reading and research.

11Story shortlist and angle selection

Having a research agent return several ranked story options, not just one, keeps a human genuinely in the loop on which story is worth covering at all.
Generating multiple angle options and having a human pick one before any drafting starts prevents the team from writing a full post around the wrong framing.

12Xemingway's X post drafts

Reviewing a draft for whether the hook actually explains the stakes, not just whether it reads smoothly, is a specific and teachable editorial skill.
It's fine to trust one AI model for research and coordination while preferring a different model for the actual writing, if that model is genuinely the stronger writer.

13LinkedIn post and Instagram carousel

The same underlying story needs a distinct draft, not just a shorter or reformatted version, for each platform and format it goes out on.

14Why he doesn't use auto-posting tools

Copy-pasting AI-drafted posts by hand into each platform, instead of using a tool that logs into your accounts to auto-post, is a deliberate reach-protection choice, not laziness.

15The full newsroom workflow

Naming and diagramming your own workflow (find, research, angle, brief, review, ship) makes it obvious which steps are automatable and which need a human.

16Where AI agent teams go next

A workflow that's manually triggered today can be redesigned to run continuously later, once the underlying agents are reliable enough to trust with less oversight.

Glossary

Terms worth knowing.

Grok Bot: An AI agent-building platform where each 'bot' is a persistent assistant with its own name, instructions, and memory, and bots can message each other to hand off work.
Agent: A single AI assistant configured for one job (researcher, writer, editor) that receives input from other agents and produces a defined output.
LinkedIn slop: Generic, hype-filled AI writing patterns, like 'I'm thrilled to announce', that make a post read as obviously AI-generated and get skipped by readers.
Angle: The specific narrative framing chosen for a story before writing begins, the detail that makes a reader stop scrolling instead of skimming past.
Human-in-the-loop: A workflow step where a person reviews or approves AI output before it moves to the next stage or gets published.
Typeless: A voice dictation tool that turns spoken words into text, used here as a faster way to give AI agents detailed instructions.

Resources

Things they pointed at.

05:02linkFree Grok Bot agent prompts (Science Based AI) ↗

08:11productCursor

09:23toolTypeless ↗

13:10toolChatGPT

Quotables

Lines you could clip.

01:18

“Writing isn't the hardest part of content anymore.”

short, counterintuitive claim that reframes the whole video→ TikTok hook↗ Tweet quote

01:44

“What's the angle that makes someone stop scrolling?”

clean, quotable definition of the real skill being taught→ IG reel cold open↗ Tweet quote

05:57

“Never default to, you know, I'm thrilled to announce this.”

specific, funny anti-pattern callout anyone writing AI copy will recognize→ newsletter pull-quote↗ Tweet quote

15:17

“They will nuke your reach because they want you to spend more on advertising to promote your posts.”

sharp, contrarian claim about platform incentives→ TikTok hook↗ Tweet quote

The Script

Word for word.

Read-along

Don't just watch it. Burn it in.

See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.

metaphoranalogy

In this video, I'm going to show you how to use Grokbot to create viral content on X and LinkedIn. I've built an AI agent content team whose job it is to find interesting stories and news, figure out which of those would make for good content, give me unique angles for each one, and then the ones that I approve, it turns into viral content across X, LinkedIn, and visuals that we'll use as like Instagram carousels.

I'd say about 70 % of this is fully automated and the other 30 % is a human in the loop for taste. Now, you could automate the whole thing or pretty much the whole thing if you want to. I just think quality is really going to suffer and you're going to get slop.

And here at Science Based AI, we have a no slop policy. But I'll show you everything. So if you want to go that route, you definitely can.

And to be clear, I'm not just talking about this as some random guy like this is literally my job. I'm the head of marketing at a biotech startup called Stack Health. And I use AI workflows like this every single day to make sure that we are producing very high quality content that helps us stay.

relevant in the conversation of what's going on in the science and health world we've already grown to tens of thousands of followers using this approach and then in my free time i break it all down for you guys so i will show you how grokbot works what the agent team looks like how to create these agents yourself and then we'll walk through a whole workflow that whole process you can see it live step by step now that you have the big picture Let's jump in.

So the thing is, writing isn't the hardest part of content anymore. Most LLMs are pretty good at writing now. They still write in kind of like an AI sounding way, but if you know how to prompt it and you're down to do a little bit of editing, you can get around that.

So the harder questions have become these things like, what should I talk about? What's happening right now? Especially if you work in an industry or an area where there's a lot of news and stuff like that, which stories actually have potential to go viral and be interesting.

And then This is probably the most important one. What's the angle that makes someone stop scrolling?

So at Stack Health, we work in the peptides and personalized medicine space that also touches biotech, longevity, emerging science. And so for us, something interesting is happening basically every day. There's a ton of good content.

We just have to figure out how to figure out what's best and then turn that into our own stuff. And so for me, the goal isn't really an AI writer, because again, we kind of already have that. It's more so an AI newsroom.

So now I want to show you at a high level what the AI agent team is going to look like. These are all of my AI agents. I personally give them funny names and I'll explain what each one means.

You can name them whatever you want. So the chief is Tecumseh. That's like a famous Native American chief.

He oversees everything. So he coordinates with all of the other agents. And if you've never used Grokbot before, you can see literally like group chats that they will have.

Then we've got a researcher. This is named Flock because he watches everything all the time. So he's scanning for news, scanning for X.

for like new viral posts and things going on. Then we've got Shemingway, which is our writer, but again for X. So he is specifically writing tweets for X.

We've got our LinkedIn writer, Eli Lewis, which is like C .S. Lewis. And he basically takes like whatever the idea is that we've talked about over here and rewrites it for LinkedIn because you want to write differently for LinkedIn versus for X.

Then we've got Johnny Ive. He is our slide designer in Figma for like these beautiful, very visually striking slides. And I'll show you some.

images of what those look like. And then we've got SlideMaster, who is our guy who builds this certain tweet -style carousel for Instagram. And so these will all coordinate.

Not every single thing that we do will need each of these, but a lot of them will. The key ones that you're going to need for this workflow are a chief is very important, something that watches for news and updates, and then a writer. If you don't want to do visuals, you don't need those, but the writer will definitely be key.

Now, before we actually go and build these in Grokbot, I have started a new Grokbot here. So you can see I don't have everything built out, but I'll show you. But before we do that, I want to talk about what makes for a good agent in Grokbot.

So if we're creating an agent, you just go up here and you hit create new bot. But I'm going to make this L .I. Lewis.

If you go up here and you click his name, you will see. So stuff pop up. So we've got the name that's already in there.

A label is basically like his role. Think of it as what team you would be on in a company. So he is going to be a writer.

Now, a description. This is where you give the instruction. And so you don't want to just say something like you are a great LinkedIn writer.

Write viral LinkedIn posts. That doesn't work so well. Instead, there's a process to it.

And here's what the process looks like. What every agent needs is five things. And those are going to be, first of all, the role.

What exactly is this agent responsible for? Second is the goal. What outcome are you optimizing for?

Inputs. What information should you expect to receive? This is especially important when we're building these agentic workflows because it's going to receive information from someone else.

So like Eli Lewis, who we were just talking about, he's going to receive information from Flock. And so he needs to know what that is and what to do with it. Four is the process.

How should it think through the task before producing output? And then five is the quality bar. Like what does great work look like here?

Whenever possible, try to give examples. That's very helpful. So a good prompt for Eli Lewis would be something like this.

And if you guys want to steal these, I will give you all of these prompts linked below. And so you can just download them and take them as well. So I'll read through the key things.

You're Eli Lewis, the LinkedIn writer for Stack Health. Turn approved ideas and research into exceptional LinkedIn posts that earn attention, teach the reader something useful, and build Stack Health's reputation as a trusted, thoughtful voice. So that's key.

We're giving it the goal. Do not merely rewrite an ex -post with more words. Reconstruct it for LinkedIn.

Assume the readers intelligent health conscious etc etc talks about hooks very important here and then structure so a strong post will usually have it'll hook the reader in with a core idea it'll quickly establish what happened or the problem provide the most interesting evidence or details then explain why this matters beyond the immediate story and with a memorable insight prediction question or implication now style conversational intelligent confident i want to read through all of that and then this is key avoiding linked in slop.

So never default to, you know, I'm thrilled to announce this. Here's the thing. This changes everything.

The future is here. Then our prompt needs to make sure that It is trustworthy and accurate.

So we're essentially having it double check that it hasn't hallucinated anything. So we're never inventing a number, quote, source, personal experience, medical implication, or scientific conclusion. If it contains uncertainty, preserve it without destroying the momentum of the post.

So here's the workflow. Here's how it operates. And then the output.

So I would just come up here. I will copy the whole thing. I will go back to Li Lewis, and then I will give it this description.

Now it auto saves. Boom, now we have Eli Lewis. Also, if you want to customize your bots, you can click up here and then just click this, and it will give you a bunch of options.

So I like, for a writer, I'm going to make him look like this guy, and we're going to make him red. One of the things that I really like about Grokbot is it helps to anthropomorphize. your agents, which some people think is bad, but I think it's nice because you feel like they are an actual team member.

So you get to choose what they look like and their name and stuff like that and makes it a little bit easier to kind of keep the context of what each agent is supposed to do tied to them. Now, I want to teach you a little bit about Grokbot so that you know how to use it. This is actually coming from later, so you can see here I actually have my bots built out already, but we'll cut back to building.

Real quick, key things that you need to know. Over here are going to be your bots. These can all coordinate with each other, talk with each other.

You can start chats where they can talk to each other and you can like see them. It's like a group chat with your friends. Over here you have a computer.

So every single bot has a computer that it can use. And now it's a super basic computer, but we can open Google Chrome and then sign in so it can use the computer to do whatever it wants. Now we are going to use this for stuff later on, but for right now that's all you need to know.

Then down here we have the marketplace where you can connect to all sorts of apps. your bots can use. So look through and find the ones that are most helpful for you.

I really like having Notion, having Gmail in here is great, Google Drive as well. If you want people also essentially open source their bots so you can borrow their bots and use them yourself. I'm not going to talk about that too much more here, but just know that is something you can do.

So there's other people build really cool bots so you can install it. Other than that, it's got all the basic stuff down here, all the settings, usage, etc. Grokbot does cost $20 a month.

It's kind of their minimum plan. If you pay for one of the X premiums, though, that also works, or if you pay for Cursor as well. So now that you've seen how we build one bot, we build the others in pretty much the exact same way.

So I'm going to speed through and do this. The important thing to note is that if you have a rough idea of what you want the bot to do, what I would suggest is... Throw that into a chat.

You can just use any of the bots here and ask it to write and improve a prompt that you can give as instructions to that bot. So you have an idea that's, you know, maybe a couple sentences long. Give it to AI.

AI will improve it, make it better. You can tweak it if needed and then use that as the instruction. So that's basically what I was doing with Eli Lewis before.

So now that we have all of the agents built out, our team is here. The Avengers have assembled. We're good to go.

I want to walk through a real workflow with you guys. So we're going to do that. And it's all going to start with Flock.

Again, this is the one that watches for news. So I'm just going to talk to Flock. Hey, Flock, can you pull the latest and best news in the science and health world over the past 24 to 48 hours?

Pull some of the best ideas that you think would make good content for Stack Health and then give them to me. And by the way, if you notice how I am talking to my computer instead of typing everything out, the tool that I'm using to do this is called Typeless. It is this little guy down here.

You just press F1 to talk and then it will listen to everything you say and turn it into like actually super high quality text. Like it's way better than the normal text to speech that you would get. They're actually the sponsor of today's video.

I've been using them for a while now. Love it. Mainly because I get so much better output when using...

text -to -speech with typeless because the thing is if you have to type all of your instructions out it gets kind of long and tedious like i'm not the fastest typer and most people can speak way faster than they can type and so i will end up giving worse inputs and therefore i will get worse outputs and so everything that i do downstream is worse if i'm typing instead of talking so that's why i use typeless all the time mainly to get better outputs but it can do a bunch of other really cool stuff too So for example, if you are reading some article and you need something summarized or translated, you can just highlight text and then press FN plus space.

You can ask it anything. So I could say, hey, this is pretty confusing. Like, I don't really understand Newton's second law of fluid motion.

Can you explain that to me like I am five? And then it'll give you this little guy here. So imagine you're pushing a toy across the floor.

You know, you guys get the picture. It can translate stuff. It also can clean up the text a little bit.

So if you're rambling and using a lot of ums and ahs and likes and stuff like that, it can clean it up for you if you want. And the way I see it, if you're going to be using AI, you want to get as good of output as possible. And so therefore, using a text -to -speech tool like Typeless is extremely impactful.

Would highly recommend it. It's free to start, but if you want a premium subscription, you can get $5 off with the link in the description. So you can see it's a pretty simple prompt, but because I gave Flock such good instructions initially, it understands what to do.

It's going to go on X, it's going to pull the news, it's going to scan for all of that stuff and give me stories that I want. So you can see now it's finished. It's given me a curated shortlist from the last 48 hours.

And so we're going to run through them. She opened one confounding, just lost another legal fight. Some of glue tied in kids six under.

Yeah, I think this is going to be the one we could go with. So we're going to run with this one. If you're not in the health world, it doesn't matter the story.

Just know that it found some good ones. And we're now going to turn this into content. Hey, flock, this is good.

Let's run with number two. Some of glue tied in kids six under 12. So the step one.

Coordinate with Tecumseh. He's going to kick this off. I want him to figure out the best angles for this.

And then I will approve what we run. And let's make a post on X. Let's make a post on LinkedIn.

And then let's send it over to Slidemaster too because he's going to come up with a tweet carousel that we can use as well. Now you can see here. that he is looping in Tecumseh.

And then it also knows because I'm asking about stuff for X and for LinkedIn, it's going to hand it off to Schemingway and to Eli Lewis. Obviously SlideMaster too, but that one I called out explicitly. Now, if we go over to Tecumseh here, you can see he got a message from Flock.

So if I click this, it's going to, you can see their actual chat. So here you can see the agents talking to each other. It's pretty cool.

It's saying like, Ben wants you to kick off this next Stack Health piece. Here's your job, et cetera, et cetera. You don't ever have to look at these, but it's good to know that you have.

Cool. So real quick, this story is about a GLP -1 trial for children, and it's saying that people who were on the GLP -1 lost significantly more weight. So both sides of the trial had diet and exercise, but just like exercising more didn't actually help them lose weight.

So I think that is pretty interesting. That's option A, which is what he was suggesting right here. I like A, so now he's giving me this little thing here, and I'm just gonna select A.

So you can see he's drafting briefs for Schemingway, for Eli Lewis, and SlideMaster now. You can see over on the side, they're moving and bouncing around. That's because they're getting info and they're taking action.

So now we've got options. So now we've got options from Schemingway. And here's the one part where I got to be honest, I honestly don't think that Grok is the best writer out there of the LLMs.

I will typically use ChachiBT for writing. So that is one part where... It's not always perfect.

Now, it's really nice to have everything within GrokBots. Sometimes it works too, but you could occasionally, if you really don't like how it's written, give this to ChatGBT, ask to write it. It's just that takes you out of the workflow.

It's data, but if you really want a high -quality post, that might be the way to go. These are good. I like the content, but your hooks are not that interesting.

They don't really explain what's going on. I feel like people need more context. The idea is there's a trial of...

GLP -1s in young people, and it had pretty shocking results. So that's more of the hook that we need to run with. Can you come up with more hooks like that?

Make sure it's very clear to people what's going on. So then if we move down, we can see we've got a LinkedIn draft by Lewis. Now...

you don't see it here because li lewis sent it in the chat so you have to go up here and click and we'll go to li lewis okay so we've got the post and then i personally after reading through i like hook number five now i'm going to hop over to slide master and I didn't fully explain this before, but SlideMaster is building this specific type of Instagram carousel.

And I'll pull it up here. It's like this tweet style thing where you just see essentially all the information, but like in a carousel where you slide through. I'm sure you guys have seen these on Instagram.

It also opens it in Figma. So now we have our content. We've got X, we've got LinkedIn, we've got...

One other thing you could add is you can use a tool that allows you to auto post content onto these platforms. I would not recommend that because oftentimes if you sign in to your social media with one of those tools, the platforms know that you're more of like a business creator and they. typically, at least this is the theory, they don't say this, they will nuke your reach because they want you to spend more on advertising to promote your posts.

So that's why a lot of the best content creators like agency businesses, they never sign into those tools. That's why I would just take this, copy and paste it over to the platform. And oftentimes I think it's helpful to visualize things with clear steps.

So I want to show you guys what this looks like. So this is running a story through the newsroom. Finds a story.

Research and verify. Flock or Tecumseh will do that. Generate angles.

I choose the best ones. Then you send the brief over to Schemingway, Eli Lewis, Slidemaster, Johnny Ive, if you're going to use them. You review the work, give feedback, tweak it, and then there is the final package, which is like the actual post.

Now, just to make it really clear exactly what we did, see this all laid out. This was the flow. We start with the internet for stories.

Flock finds those, sends it to Tecumseh. Tecumseh gives me options and angles. I choose what works best.

Then we send that off to Schemingway, Eli Lewis, or Slidemaster. Then you get all the info back. This is the human review.

I look at it. I say, I don't really like this hook. This doesn't make sense.

Approve. Done. And then I would copy and paste it into those platforms and post it.

Now, I think this is very interesting for many reasons, but one of them is that right now I still initiate a lot of this workflow manually. but it doesn't have to be that way and it will continue to evolve. So if you want to burn more credits, I can just have flock running all of the time, continuously watching for stories.

And as the bots get better and better, they can take over more of the work. So as soon as I open my computer, I get the little notification. Hey, there's a viral story.

I already have like five content ideas just sitting there that are pretty good. So I just pick one, the rest of the team goes to work and I have a bunch of really good content. So for me, Grokbot is especially cool because it's the best way I've seen of actually using agents and visualizing agents as a team instead of just a one -off run.

Now I see how they all work together and each one has its place that helps to create a team that does what I want. So we're slowly moving up the skill ladder from just chatbots and asking it to write something to now a team of agents running, producing high quality content. I hope this was helpful.

Let me know if you have any questions. Try to answer them in the comments below. I'll see you next.

The Hook

The bait, then the rug-pull.

Ben, head of marketing at biotech startup Stack Health, opens by promising a full AI content team, then pivots almost immediately: the real bottleneck was never generating text, it's picking the right story and the angle that earns attention.

Frameworks

Named ideas worth stealing.

04:21list

The 5-Part Agent Instruction Framework

Role
Goal
Inputs
Process
Quality Bar

Every Grok Bot agent's instructions should define what it's responsible for, what outcome it optimizes for, what information it expects to receive, how it should think through the task, and what 'great work' looks like, ideally with examples.

Steal forany AI agent prompt, not just content writers

02:11model

The Agent Team (Org Chart)

Tecumseh (chief)
Flock (researcher)
Xemingway (X writer)
L.I. Lewis (LinkedIn writer)
Johnny Ive (visual designer)
SlideMaster (carousel builder)

A named team of specialized bots, one coordinator plus one bot per platform or output type, that hand work to each other instead of one bot doing everything.

Steal forstructuring any multi-platform content operation

15:44model

Newsroom Workflow (Story-to-Package Pipeline)

Find the Story
Research + Verify
Generate Angles
Send the Brief
Review the Work
Ship the Package

The six-stage pipeline that turns a raw news item into an approved X post, LinkedIn post, and Instagram carousel, with a human choosing the angle and reviewing final drafts.

Steal forany recurring content pipeline that mixes automation with editorial judgment

CTA Breakdown

How they asked for the click.

VERBAL ASK

09:23link

“Check out the sponsor of today's video, Typeless, and get $5 off with that link.”

Delivered organically mid-tutorial while actually demonstrating the tool live (dictating instructions, highlighting text to summarize), rather than as a separate pre-roll ad; the discount link is repeated in the video description.

MENTIONED ON CAMERA

05:02linkFree Grok Bot agent prompts (Science Based AI) ↗

09:23toolTypeless ↗

FROM THE DESCRIPTION

AFFILIATECommission earned if you click.