Every Wednesday, one essay on what's actually working in Marketing with AI - tools, workflows, and mindset shifts you can apply immediately.
Claude for Marketers Masterclass is now live.
If you’ve been using Claude as a chat window and wondering why it doesn’t quite deliver - this is why.
The course teaches you how to run Claude as a complete marketing system, built from 30+ live workshops attended by 300+ marketers.
10+ hours of pause-and-do content. 9 bonus assets. One new module every month through 2026 — so the course grows as Claude does.
Trusted by marketers at Adobe, JPMorgan Chase and the Chartered Institute of Marketing. Now an Inc Magazine columnist.
Early-bird pricing available until Monday 15th June at 8am.
Here is an actual campaign brief that went into my MarketingOS the other week.
“Build out the content for my Claude for Marketers coaching programme launch.”
Here’s some of the outputs from that simple prompt:
A sales landing page
A five-email welcome sequence
LinkedIn and Substack Notes for the full launch window
Meta ad variations
And a brand voice consistency check across all of it
The marketing system decided which specialist skill handled which deliverable.
I did not route anything manually.
I did not copy and paste between tools.
I described what I needed in plain English, and the system figured out the rest.
Now to be clear, my system already had context to draw from, such as brand guidelines, product details and campaign goals.
But once that foundation is in place, a plain English brief is genuinely enough to get a project moving.
That is not a tool. That is a team.
And this week, I want to show you exactly how that team is structured, what each department actually does, and how to start building your own from scratch.
The Step Most Marketers Are Missing
Most marketing teams are adding AI to their existing workflow.
A better headline here. A faster brief there.
All very useful, but it’s still the same workflow, just running slightly quicker.
A smaller group of marketers are doing something different. They are not adding tools to their process. They are building systems that replace the process. And the gap in output between the two groups is not small I’m afraid to say.
The system I built is called MarketingOS. I put it together over about 10 days inside Claude Code.
In plain terms: a central orchestration layer that receives a marketing brief, breaks it into phases, and routes each deliverable to the right specialist skill automatically.
Thirty-eight skills are in this system, along with two expert agent advisors, and one command to start any multi-step project.
It’s not a prompt library, and it’s not a shortcut folder.
I’d say it’s much closer to the inner workings of a full marketing department.
How MarketingOS Actually Works
The centrepiece of the system is a skill called /marketing-pipeline.
Think of it as the marketing director of the team.
Any brief goes in as plain English. Then, the pipeline skill takes it through four phases before anything reaches you:
1. Brainstorm. The pipeline starts with clarifying questions and approach options. Before a single deliverable is scoped, the system thinks. What is this really asking for? Who is it for? What does success look like?
2. Plan. A structured execution plan is produced, naming every deliverable needed and which specialist skill will handle each one. This is where the coaching programme brief became nine individual deliverables across four skill categories.
3. Execute. The pipeline calls each specialist skill in sequence, passing the right context to each one. The copywriting skill gets the brief plus the ICP research. The email sequence skill gets the launch strategy plus the audience insight. Every skill builds on what the previous one produced.
4. Review. A quality and consistency pass across all outputs before anything reaches you. Brand voice, messaging alignment, tone consistency. The system checks its own work.
Every task enters as plain English and exits as a set of polished deliverables.
There’s no manual routing or context switching. And thankfully, no dropped balls along the way or missing context from the previous steps.
From one sentence about a Claude for Marketers coaching programme, the pipeline planned nine deliverables and executed them in sequence.
Here is how each department of the system handled it.
The Seven Marketing Departments
Think of the 38 skills as being allocated to seven specialist marketing departments, each with its own remit.
Using the Claude for Marketers coaching programme example I mentioned earlier - I’ll run that example through the MarketingOS below to show you the outputs I got back, stage by stage.
Hopefully that will bring it to life a little for you.
Department 01: Strategy and Planning
Before any campaign runs, this department handles the thinking.
Positioning, launch planning, customer research, pricing decisions, growth approaches.
Most marketers unfortunately skip this layer and go straight to content. The pipeline doesn’t.
Here’s the skills baked into Strategy & Planning:
• /content-strategy Topic planning and editorial calendars
• /launch-strategy Product launches and go-to-market planning
• /pricing-strategy Pricing, packaging and monetisation decisions
• /marketing-ideas 139 proven growth approaches by category
• /competitor-alternatives Competitor comparison and alternative pages
• /free-tool-strategy Engineering-as-marketing planning
• /referral-program Referral and affiliate programme design
• /customer-research ICP, VOC analysis and persona building
With the coaching programme, the pipeline started here.
The /launch-strategy skill mapped a three-phase campaign: two weeks of waitlist building before launch, a five-day launch window, and a post-launch nurture sequence for people who did not convert first time.
The /pricing-strategy skill worked through the tier structure and the logic for an early-bird discount.
The /customer-research skill produced the ICP profile and a list of the three most common objections to address on the sales page.
None of that was in the original brief. The pipeline generated it before a single word of copy was written.
This is the department that makes everything downstream more accurate.
When the content department knows who the buyer is and what they object to, the copy it produces is a different quality than copy written in a vacuum.
Department 02: Content Creation
Eight skills covering the full content stack. From the sales page to the ad creative, from the email sequence to the social calendar. If it is written marketing output, there is a specialist skill for it.
Here’s the skills baked into Content Creation:
• /copywriting Website copy, landing pages and feature pages
• /email-sequence Drip campaigns and lifecycle email sequences
• /social-content Posts for LinkedIn, X, Instagram and TikTok
• /content-os Weekly content rhythm and Content Matrix
• /cold-email Cold outreach and prospecting sequences
• /ad-creative Ad copy and creative variations at scale
• /sales-enablement Pitch decks, one-pagers and demo scripts
• /lead-magnets Lead magnet strategy and creation
With the coaching programme, this is where the volume of output was produced.
The /copywriting skill wrote the sales landing page, built from the ICP profile and objection list the strategy department had already produced.
The /email-sequence skill created a five-email welcome sequence for waitlist subscribers and a separate three-email launch sequence for the open window.
The /social-content skill produced twelve posts across the launch period: pre-launch teasers, the launch day announcement, and social proof posts for days three, four and five.
The /ad-creative skill produced four Meta ad variations and three Google ad variants, each testing a different angle from the positioning work.
That is 26 pieces of content from one brief.
Drafted, structured, and consistent in voice across all of them, because every skill was working from the same strategic foundation.
Department 03: SEO and Discoverability
Five skills covering how people find you. Traditional search through to the AI-powered discovery surfaces that are now reshaping where buyers do their research before they ever reach your landing page.
Here’s the skills baked into SEO & Discoverability:
• /seo-audit Technical SEO audit and on-page review
• /ai-seo Optimisation for ChatGPT, Perplexity and AI Overviews
• /programmatic-seo Template pages at scale
• /schema-markup Structured data and JSON-LD implementation
• /site-architecture Site structure, navigation and URL design
The /ai-seo skill is worth flagging specifically.
ChatGPT, Perplexity, and Google’s AI Overviews are now primary discovery surfaces for buyers doing pre-purchase research. Being optimised for traditional search results is no longer enough. The /ai-seo skill restructures content to appear in AI-generated answers, not just blue links.
With the coaching programme, the /seo-audit skill reviewed the landing page for technical issues before anything went live.
The /ai-seo skill restructured the FAQ section to match how AI search engines answer questions about coaching programmes. Schema markup was added so the programme appears as a rich result in Google alongside the standard listing.
People have to find the page before they can read it. This department makes sure the discovery layer is working before the campaign spends money on traffic.
Department 04: Conversion Optimisation
Seven skills optimising every step of the funnel, from the first click on the landing page through to what happens when someone decides to cancel.
Most marketing teams spend heavily on traffic and lightly on conversion. This department covers everything after the click.
Here’s the skills baked into Conversion Optimisation:
• /page-cro Landing page, pricing and homepage optimisation
• /signup-flow-cro Signup and registration flow optimisation
• /onboarding-cro Post-signup activation and first value
• /paywall-upgrade-cro In-app paywalls and upsell modals
• /form-cro Lead capture and contact form optimisation
• /popup-cro Popups, modals and exit intent
• /churn-prevention Cancel flows, save offers and dunning
With the coaching programme, the /page-cro skill reviewed the landing page structure and suggested three changes: moving the pricing section higher on the page, restructuring the testimonials to directly address the three objections the strategy department had identified, and tightening the primary CTA.
The /form-cro skill flagged that three fields on the waitlist signup form were creating unnecessary friction.
The /popup-cro skill built an exit-intent offer for visitors leaving the landing page without signing up.
Getting traffic to a poorly converting page is an expensive way to learn what is not working. This department makes the traffic work harder before you spend money sending more of it.
Department 05: Measurement and Ops
Four skills that keep the engine running. Without measurement, every decision is a guess. With it, every campaign produces data that makes the next one sharper.
Here’s the skills baked into Measurement and Ops:
• /analytics-tracking GA4, event tracking and GTM setup
• /ab-test-setup A/B test design and experiment planning
• /paid-ads Paid ad strategy, targeting and ROAS optimisation
• /revops Revenue ops, lead scoring and CRM automation
With the coaching programme, the /analytics-tracking skill has set up event tracking for three key moments: waitlist signup, checkout initiation, and purchase completion.
The /ab-test-setup skill designed a headline test for the landing page with two variants, a minimum sample size defined before starting, and a clear success metric agreed in advance.
The /paid-ads skill reviewed the Meta campaign structure and recommended segmenting warm audiences separately from cold for the first two weeks.
The measurement layer is what turns a first campaign into a second one that performs better. Without it, you are optimising by instinct. With it, you are optimising by evidence.
Department 06: Copy Quality
Two skills for when the writing needs to be sharper. The first is a systematic editing pass. The second applies psychology to the copy itself.
Here’s the skills baked into Copy Quality:
• /copy-editing Edit, review and polish existing copy
• /marketing-psychology 70+ psychological principles applied to copy
The /copy-editing skill runs a structured pass: weak verbs, passive voice, buried leads, inconsistent tone, CTA strength, headline scoring.
The /marketing-psychology skill goes a layer deeper. It maps 70+ psychological principles to specific copy patterns: loss aversion, urgency triggers, social proof frameworks, reciprocity patterns, cognitive bias application. Not generic notes. Specific rewrites.
With the coaching programme, every piece of content that came out of the content creation department went through /copy-editing before it was presented for review.
The /marketing-psychology skill then reviewed the sales page specifically and identified three places to add more specific social proof, two places where loss aversion would strengthen the CTA, and one section where the urgency framing was too vague to convert.
This is the department that separates copy that reads well from copy that converts.
Department 07: Video Production
Two skills for producing video inside the system, without a production team or external tools.
Here’s the skills baked into Video Production:
• /remotion-video Locally rendered MP4 videos via Remotion
• /outscal-video AI-generated animated videos with voiceover
With the coaching programme, the /remotion-video skill produced a ninety-second programme explainer for the landing page and a thirty-second social cut for launch week. Described in plain English.
Produced without a shoot, a production team, or a brief to an agency.
The Agents Layer
Now, ‘Skills’ handle execution. Agents handle a different kind of thinking.
I built two agents into MarketingOS, and I want to be upfront about why I picked these two specifically.
The first is built around Alex Hormozi’s frameworks.
I have spent a lot of time with his approach to offer design, specifically the Value Equation and Grand Slam Offer methodology.
So, when a project touches pricing, packaging, or offer structure, I wanted that thinking available inside the system, not just as something I occasionally remember to apply.
I fed in his core frameworks, his pricing logic, and his lead generation approach. The agent now applies those lenses whenever the pipeline routes an offer-related brief to it.
The second is built around Matt Gray’s Founder OS.
I have followed Matt’s content for a while and kept coming back to specific things he teaches: his hook templates, his Pitch OS sales system, his content challenge framework.
So rather than bookmarking them and forgetting them, I packaged them into a reference library and built an agent trained to use them. When the pipeline needs content that hooks from the first line, or copy for a sales call, this is the advisor it calls.
Think of each agent as a second brain, built on a framework you already trust.
Not a generic AI assistant. An advisor who approaches every brief the way a specific expert would, drawing on documented methodology rather than general knowledge.
These two agents were built by me, for me.
But the structure is completely open. You can build your own agent around any framework, methodology, or expert whose thinking you want permanently available in your system.
A favourite book on positioning. A sales methodology your team swears by. A copywriting approach you keep coming back to.
Capture it in a reference document, describe how you want the agent to use it, and it is available for every future brief.
You may not need agents at all when you are starting out. The thirty-eight specialist skills are powerful on their own. Agents are the layer you add when you want strategic frameworks baked into the system, not just execution skills.
How This MarketingOS Was Built in Claude Code
I built the entire system in Claude Code. And I’m not a developer.
If you want the full picture of what Claude Code actually is, I covered it in The Marketer’s Guide to Claude Code (The Non-Techie Version) earlier this year.
The very short version:
Claude Code is Anthropic’s environment for building things with AI. You describe what you want. Claude builds it. All in plain English.
The process for building each skill followed the same four phases the pipeline uses for marketing projects.
1/ I described what good output looks like for a given deliverable.
2/ Claude then asked clarifying questions. We planned and agreed the structure together.
3/ Claude wrote the skill files.
4/ I finally tested it against real briefs, refined the instruction set where the output was not right, and kept what worked.
That loop, repeated across thirty-eight skills over a couple of weeks, is how the system was built.
I did not write the skills. I described them. And some I borrowed from others and then refined them for my use cases.
I gave Claude the logic (Brainstorm, Plan, Execute, Review), the output format I wanted, the quality standards it had to meet, and the constraints. Claude Code then turned those descriptions into working files.
The system lives in a GitHub repository. If you’ve never used GitHub before, do not let that word put you off. At its simplest, GitHub is a folder that lives in the cloud, tracks every version of your files, and can be shared with your team.
And you can ask Claude Code to set up the connection for you, again in plain English. It handles the technical steps.
You only need to confirm them and once it is connected, updating a skill or adding a new skill or agent to your Marketing OS is simple: make the change, push it to the repository, and everyone on your team pulls the latest version.
One update activates instantly. No manual sharing. No version confusion.
The Simple Way to Get Started
You do not need thirty-eight skills from day one. You need one.
1. Pick your most repeated deliverable. What do you write or produce more often than anything else? Cold emails? Social posts? Landing page copy? That is your first specialist skill. The one that will save you the most time the fastest.
2. Write out what good looks like. Before you build the skill, document your standard. What makes your best cold email actually work? What structure, what tone, what length, what CTA? What do you always change in AI drafts before they are good enough to send? This document becomes the instruction set your skill will follow. If you do nothing else this week, write this document.
3. Build the skill in Claude Code. Open Claude Code and describe what you want the skill to do. Share your good-looks-like document. Claude Code will generate the skill file. Test it against a real brief. Adjust the instructions until the output consistently matches what you would have produced yourself.
4. Add a second skill. Then a third. Each one compounds the others. A copywriting skill works better when a copy editing skill follows it. A launch strategy skill works better when an email sequence skill is ready to execute against the plan. Build towards having coverage of your most common project types.
5. Build the orchestrator. Once you have five or six specialist skills, build the routing layer. This is the pipeline that receives a brief and decides which skills handle which parts of it. With enough specialists in place, the orchestrator becomes the multiplier.
6. Connect it to GitHub. Ask Claude Code to help you set this up. Tell it, in plain English, that you want to commit your skills folder to a GitHub repository so you can share it with your team and track changes. Claude Code will walk you through it. It is a ten-minute setup the first time. After that, updating your system is a single step: push the change, team pulls it, everyone runs the new version. That is also what makes this different from a prompt you save in a doc somewhere. It is a living system, maintained in one place, deployed everywhere.
The AI Marketing Team of The Future (or now)
This is not a story about replacing your marketing team.
This kind of system handles the middle 80 per cent of any marketing workflow. The repeatable, structured, execution-heavy work that consumes hours without requiring judgment.
For example, writing the same type of email sequence for the fourteenth time. Running the same technical SEO check for a new landing page. Producing ad variations from a brief that is basically identical to last month’s.
You still own the first 10 per cent of the workflow: the strategy, the positioning, the brief, the context that only you have.
And you still own the last 10 per cent of the workflow: the review, the decision on what ships, the judgment call on what is on-brand and what is not.
The system does not make those decisions. You do.
It just means your decisions become deliverables faster, and at a baseline quality that most teams cannot match consistently when doing it manually.
One person with a well-built system can outship a team without one. A team with a shared system can operate at a scale that simply was not possible two years ago.
That is the AI marketing team of 2026. Not a replacement for people. A multiplier for the people who build it.
Want hands-on help with Claude for Marketing?
I work with marketers 1:1 in Claude Marketing Power Hours and help teams embed Claude into their workflows from first steps to full-scale systems. Book a call using the link below.
Thanks for reading Marketing with AI! If you enjoyed this read, the best compliment I could receive would be if you shared it
Clear thinking here. Thank you for sharing
This 👇🏼
“A smaller group of marketers are doing something different. They are not adding tools to their process. They are building systems that replace the process.”