I Built a Marketing Team with Clawdbot (free template + files)
By 9x
Summary
Topics Covered
- AI Marketing Team Runs Competitor Analysis
- Head of Marketing Generates Full Campaign Plans
- Creative Director Builds Ready-to-Use Assets
- Performance Marketer Publishes Meta Campaigns
- Full Pipeline Costs Under $24 on Claude
Full Transcript
I built an entire marketing team that can research my competitors, plan a full campaign, create the ads, build the landing pages, and even publish everything to Meta. And the whole thing runs on Claudebot. Here's what the whole
system looks like. I've got four roles.
an ads analyst that does all the competitor research. A head of marketing
competitor research. A head of marketing that takes that research and plans the campaign. A creative director that
campaign. A creative director that produces the actual assets, the ad images, the scripts, the landing pages, and a performance marketer that pushes everything to Meta, creating the campaigns, the adets, the ads.
Everything is there in draft ready for you to review. Now, as you can see, there are quite a few different skills that we've created for our marketing team. Every single one of these is just
team. Every single one of these is just a markdown file. There's no code, no NAM workflows, no complicated automations, just plain written instructions that tell Claudebot exactly what to do. For
context, Claudebot is a new tool that gives Claude or any AI model actual access to your computer instead of copy pasting back and forth. It just does the work directly, opens your files, browses
the web, uses your tools. It's basically
like having a fullon employee. At the
end of the video, I'll show you exactly how to download all of these skills for the complete OpenCL marketing team for free. So you can set this up on your
free. So you can set this up on your Claudebot as well. But let's start with a quick demo of what this actually produces. And then I'll break down how
produces. And then I'll break down how each step works. And I'll also cover pricing and what it actually costs to run the whole system. Let's get into it.
So this whole thing starts by messaging our Open Claw or Claudebot here in Telegram or whatever messaging app you have this connected to. My Claudebot's
name is Otto, so I'll probably be referring to him as Otto for the rest of the video. So I call the ads analyst
the video. So I call the ads analyst command. So orchestrate competitive ad
command. So orchestrate competitive ad research. extract ads from meta ads
research. extract ads from meta ads library and generates a new strategy report. And then all I need to give it
report. And then all I need to give it is the link from the meta ads library of whatever advertiser competitor we want it to analyze. So in my case I want to analyze the ads from Eric partaker who
is a executive CEO uh coach has quite the following and is currently running ads on meta. So all I'm going to do is just copy this URL and then with the ads analyst command send this over to auto.
So auto confirms it's getting to work with the ads analyst pipeline. It's
already extracted the ad account ID from Eric based on the URL.
Now Otto's been running for about 10 minutes and now has given me the confirmation that the ad analysis is complete. And he sent me the full
complete. And he sent me the full results of his analysis in a zip folder.
So let's check that one out. And inside
that folder, we get as separate files every single ad that Eric is currently running uh on his channel. And you can see that it's also visited every landing page that the ads are leading to. But
more importantly, we get this full analysis HTML. Let's open that one up.
analysis HTML. Let's open that one up.
So, here's that analysis, and the results are pretty impressive. We have
every single ad grouped by landing page.
So, in this case, there is uh a quiz as one of the landing pages. It breaks down the full strategy on the landing page.
So, here we have the flow, quiz start, email capture, personalized report, nurture sequence, classic lead magnet funnel with gamification, some of the strengths and areas where it could improve. And then we have every single
improve. And then we have every single ad that is leading to this funnel. For
video ads, >> get this thing called the scale score.
>> We can easily play them here. Um, it's
also for video giving us a breakdown of what's included in the script, the hook, and that the emotion that the ad is playing on. And for the image ads, the
playing on. And for the image ads, the analysis is much more focused on design and messaging. So, with that analysis
and messaging. So, with that analysis now complete, we're going to hand over to our head of marketing, who's going to take those insights and plan a better campaign for us. And we can even see Otto's prompting us here, if you want to hand this over to your head of
marketing, just use the head of marketing uh tool and then put in your website. So, I'm just going to put in
website. So, I'm just going to put in our 9x website and hit enter. And then
we can see auto gets to work. Now, the
first thing our head of marketing does is looks at our own website. So, it can understand exactly what products we're offering, how we communicate, and also our style guide. So after a quick analysis, you can see it tells us
exactly what pages it already visited as well as the style guide elements that it extracted and asks, "Hey, are there any other pages that you want me to check before generating the brand bible?" So I just told it that I also want it to
visit our Luma page where we run free AI workshops every week. So once our head of marketing understands exactly what products we offer, has a full understanding of our brand, it sends over its campaign proposal. So here we
have the 9X campaign proposal, four landing pages, 11 ad creatives, seven scripts, and three funnel stages. This
is a full and comprehensive dock of our campaign strategy. This includes landing
campaign strategy. This includes landing page briefings, image and video ad specs, video scripts, and even budget allocation. With that campaign proposal
allocation. With that campaign proposal ready to go, the head of marketing then hands off that campaign proposal to the creative director who gets to work actually creating the assets. So you can
see here it created two landing pages.
One for our 9x live workshops and another for an AI readiness quiz. And
this landing page it created is actually pretty spot on. I was really annoyed at myself that we don't actually have one like this. I love the tagline learn AI
like this. I love the tagline learn AI live every week. And using our Luma calendar, it's pulled in our next workshop. And we also have our upcoming
workshop. And we also have our upcoming sessions here as well. A bit of social proof. 53 plus workshops hosted, 10K
proof. 53 plus workshops hosted, 10K operators joined, 4.9 average rating.
This is absolutely amazing. And we also have a landing page for our AI readiness assessment. Start the quiz. Find out in
assessment. Start the quiz. Find out in two minutes. Is your team actually ready
two minutes. Is your team actually ready for AI? Get clarity on where your team
for AI? Get clarity on where your team stands. AI readiness score. Personalized
stands. AI readiness score. Personalized
action plan. Free crash course access.
This is also pretty spot-on. And we have that first question of the quiz ready for preview. We also get all of the
for preview. We also get all of the images that it created for our ads here for review and even went ahead and created a five slide carousel. Now,
finally, with those assets creative, the creative director can hand everything over to our performance marketer. And we
can see here the performance marketer has straight away confirmed the assets received from the creative director.
We've got two landing pages, five single image ads, two carousels with 10 cards total, and eight video scripts. Before
the performance marketer sets everything up in meta, it's going to validate the targeting. So, here it originally
targeting. So, here it originally suggested the Duck region with German and English ads. And I said actually we want to target US, UK and DH but only English ads. Go ahead and create the
English ads. Go ahead and create the campaigns. And here we get the
campaigns. And here we get the confirmation all created successfully.
An overview of the campaigns and a link where we can review all of this in our ads manager. Let's take a look. And here
ads manager. Let's take a look. And here
we go. All four campaigns created in draft mode for us to review. We can see two campaigns for the top of funnel, a middle of the funnel campaign, and a bottom of the funnel focused on team training. Let's take a look at one of
training. Let's take a look at one of these campaigns. If I check out the live
these campaigns. If I check out the live workshop. So here we've got the
workshop. So here we've got the targeting set up on the adset and we have the four different ads for those videos where we've just created the scripts and we need to record them. We
have the video placeholders. So as soon as I record that, I still want my videos to be authentic. We could think about using AI models for that. But in my case, we know that real videos perform much better on meta. So we have the
placeholders here. But let's check out
placeholders here. But let's check out for instance this carousel ad. So here
we even can see that carousel ad which has been completely set up including the headline, the description, the text, everything. Here we have what we build
everything. Here we have what we build at 9x live and it's basically a highlight of some of our recent workshops. So AI agents for beginners,
workshops. So AI agents for beginners, build a customer support agent live, make automations masterclass, lead qualification workflow, prompt engineering deep dive, build a prompt
library, join the next one free every week. This is absolutely amazing and I
week. This is absolutely amazing and I have to admit the quality of this is pretty good. If I compare this to some
pretty good. If I compare this to some of the actual ads that I see on meta quite regularly. So that's the full
quite regularly. So that's the full output. Now let me break down exactly
output. Now let me break down exactly how this works so that you understand it and can run it for yourself. This whole
thing is powered by Claudebot skills.
Skills are basically a set of instructions that tells your AI agent exactly how it should work on a certain task. So here we have four orchestrator
task. So here we have four orchestrator skills. We have those four employees.
skills. We have those four employees.
the ads analyst, the head of marketing, the creative director, and the performance marketer. And in each case,
performance marketer. And in each case, all of these skills hand off or like the first three skills hand off to that next skill. And then each of these major
skill. And then each of these major orchestrator skills, the ads analyst, head of marketing, creative director, and performance marketer have specific subsklls. So, let's start with the meta
subsklls. So, let's start with the meta ads extractor. I can show you exactly
ads extractor. I can show you exactly how that skill is set up. This skill
instructs our agent to use a couple of tools. First of all, a browser tool.
tools. First of all, a browser tool.
What it's going to do, it's actually going to open the browser on the computer your Claudebot is running on and literally visit that ad library page and then download all of the assets. Get
the correct ad library page ID, open the ad library with filters, extract all media URLs from DOMs. This basically just means find out where the videos and images are hosted. Actually, download
the videos and images and capture the right dimensions. Extract the aspect
right dimensions. Extract the aspect ratios and very important here, scroll for more results. The ad library lazy loads ads to get more than the initial batch. Scroll down and wait then
batch. Scroll down and wait then re-extract. Inside each of these subs
re-extract. Inside each of these subs skills, we're also doing handoffs to other subsklls. So after you've
other subsklls. So after you've extracted all the ads, use ad creative analysis. Analyze each creative video
analysis. Analyze each creative video via Gemini and images via vision. And as
well for landing pages, visit and analyze every single landing page. And I
should probably also show you what the skill looks like for the orchestrator skill that ads analyst here. This just
gives the highlevel workflow and lays out how it can use each of its sub tools. Meta ads extractor analyzer as
tools. Meta ads extractor analyzer as well as the ad creative analysis and the landing page analysis. And in those orchestrator skills, this is where we have those handoffs to the next
orchestrator. So here in phase six of
orchestrator. So here in phase six of the ads analyst, once all of its subs skills has run, it's going to hand off to the head of marketing. The head of marketing has two subs skills. First of
all, a website brand analysis. So, this
is that moment where you saw it actually went and visited every single page on our own website. This is the best way for it to understand what products we offer, how we communicate, and our complete design system, which is very
important for when it actually creates the landing pages and the image ads. And
what's pretty cool here in the campaign planner subskll for our head of marketing because this is already one step down in our chain, we start off by laying out the inputs. what should you already have access to before you get to
work? And if it doesn't have any of
work? And if it doesn't have any of these, it's going to proactively ask us like, hey, I haven't run my brand analysis yet. I don't know exactly what
analysis yet. I don't know exactly what products we offer. And also in this one, we can see in all of these skills, we really reiterate how it should work with those other skills in our marketing team. So for the campaign planner, this
team. So for the campaign planner, this skill works best when combined with the website brand analysis. Generate the
brand bible first and the meta ads analysis. Analyze competitor ads after
analysis. Analyze competitor ads after approval, create the assets. Now we get to the actual asset creation and that is of course done by our creative director.
This creative director has four subs skills. First of all, an ad designer.
skills. First of all, an ad designer.
This one's going to use Nano Banana Pro to actually create the image ads. We
have a script writer who's trained on creating very short punchy video ads, writing the complete scripts, VSSLs, that sort of things. They all be 20 to 40 second videos. Then of course for
creating the landing pages, we need two skills. We've got the page designer as
skills. We've got the page designer as well as a skill for front-end design.
So, probably the most interesting one here is that ad designer subs skik skill where we are using Nano Banana Pro to actually come up with the images. You
will need to connect your Gemini API key to have this one working. We can see here we lay out all the different aspect ratios it can use. Obviously, this will be heavily influenced by any analysis
that was previously done. And then all we're doing here is basically telling it how to create very very good prompts that we can send to that nano banana model. And from testing this, what I
model. And from testing this, what I found is less is definitely more. So
what I've told it very clearly is it should not include logos, should not include brand names, company names, watermarks, or any additional text. Only
what it's going to give it in the prompt. And the best way for it to
prompt. And the best way for it to actually learn this is through example prompts. So we can see here a few
prompts. So we can see here a few different example prompts that we would send off to Nano Banana Pro to get actually good results and then also the exact structure of how to make those API
calls. And then the workflow is actually
calls. And then the workflow is actually pretty simple. Whenever you're creating
pretty simple. Whenever you're creating an image, receive the brief. This is
what it's going to get from the creative director. Construct the prompt, generate
director. Construct the prompt, generate the image, and this next part is critical. Review before sending. This is
critical. Review before sending. This is
how we sort of avoid that AI slop. What
our AI agent is going to do whenever it actually generates an image ad, it's going to check its work. Check that the text is correct, there's no hallucinated logos or brand names, no unexpected elements, the aspect ratio looks
correct, and the overall quality is acceptable. And then iterate if needed.
acceptable. And then iterate if needed.
If the image has issues, simplify the prompt, be more explicit, try regenerating. Then finally, with all of
regenerating. Then finally, with all of that created, we hand over to our final orchestrator, which is the performance marketer. Right now, it's very, very
marketer. Right now, it's very, very simple. just has one subs skill to
simple. just has one subs skill to publish these ads to Meta, but I'm already thinking about new subs skills that I want to potentially give our performance marketer. For instance,
performance marketer. For instance, actually checking on running campaigns and seeing how they're performing. So,
actually looking into our meta ads publisher skill. Obviously, this needs
publisher skill. Obviously, this needs to connect with your Meta Ads account.
So, you're going to need a Meta Marketing API. What's great here is you
Marketing API. What's great here is you can actually ask your Open Claw, your Claudebot to walk you through the process. That's exactly what I did.
process. That's exactly what I did.
After you import these skills, you can say, "Hey, I actually need to go ahead and create this API and it will walk you through everything step by step. Once
you have all of those credentials, you can add it here to your environment file." And that way it can actually
file." And that way it can actually connect to your ad account. And again,
here we have this little review step added to the workflow before creating any ads, review the full picture that make sure everything is cohesive and that campaigns aren't contradicting each
other. And then we simply equip our AI
other. And then we simply equip our AI agent with all the API calls it needs to make. How to create campaigns, how to
make. How to create campaigns, how to create adets, how to upload images, and then how to create the different types of ads. If it's a single image ad, if
of ads. If it's a single image ad, if it's a carousel, and if it's a video ad.
Now, if you want to set this up for yourself, I've put the full marketing team template in the link in the description below. You can download the
description below. You can download the whole thing for free. And the easiest way to get it running is honestly just sharing that folder with your Claudebot and ask it to help you import those skills. You literally just tell it
skills. You literally just tell it something like, "Hey, I'm sharing with you a set of skills for a marketing team. Can you use your skill creator
team. Can you use your skill creator tool to import all of these skills for yourself?" And Claudebot will handle the
yourself?" And Claudebot will handle the rest. It'll read through each file, set
rest. It'll read through each file, set everything up, and it will probably walk you through setting up any of those API keys that it needs along the way. All
right, let's talk about pricing. And the
important thing to understand is how much this costs to run really depends on which AI model you are using. With
Claudebot or Open Claw, you can actually choose exactly which AI provider and model you want to run it on. You're not
locked in to only using Anthropic or only using OpenAI, but your costs will be different depending on which model you use. Now, I strongly recommend
you use. Now, I strongly recommend running this on Claude from Anthropic.
For marketing work specifically, it is the best. It has the best personality,
the best. It has the best personality, the best tone of voice, and Opus 4.6 right now is the strongest model. That's
what I used for everything you just saw today, but it is definitely one of the more expensive models. When you set up Claudebot, you'll see this screen where you can select your AI provider. And to
connect with Anthropic, you have two options. one using an API key or two
options. one using an API key or two through something that's called a setup token which actually uses your monthly clawed subscription. Let me show you
clawed subscription. Let me show you what it actually cost me to run this whole campaign pipeline from start to finish on the API. So this whole flow from researching our competitors all the
way to actually creating the ad assets and uploading everything to Meta cost me just under $24 on Opus 4.6. That might sound like a lot, but think about what you'd pay even
one marketing freelancer for a day's work. It's not even close. Now, as I
work. It's not even close. Now, as I said, this was running via the API, but the other option is your Clawude subscription, which starts at just $20 a month. It's way cheaper, obviously, but
month. It's way cheaper, obviously, but I do want to flag something. There's
been reports that Anthropic has been shutting down accounts that are using their Claude subscription to run Claudebot or Open Claw because technically you're meant to be connecting through the API. So, if you go the subscription route, while it is
cheaper, just know that there is a risk and you need to make that call for yourself. And then we also have the cost
yourself. And then we also have the cost of generating the images for our image ads. I'm using the Gemini Nano Banana
ads. I'm using the Gemini Nano Banana Pro model. For 2K resolution, you're
Pro model. For 2K resolution, you're looking at about 13 cents per image. And
if you instead opt for 4K resolution, it's about 24 cents per image. So even
if you're generating a dozen ad creatives, you're talking just a few dollars, which is totally reasonable.
Now, if this is the first time you're seeing Claudebot in action and you want to understand exactly what it is, how to set it up, and what else it can do beyond marketing, click on this video right here because in it, I break down
the four capabilities that make it different to any other AI tool out there and walk you through the full setup step by step. So, if that sounds interesting,
by step. So, if that sounds interesting, I'll see you in the next
Loading video analysis...