Growing up, when I was first getting into business, I naively believe that if you build it, they will come.
I thought a great product would market itself. Customers would show up out of nowhere. The person with the best product would always win.
Then I started cutting my teeth in the real world of entrepreneurship, building businesses and creating products, and I realized nothing could be further from the truth.
The best product rarely wins. In fact, the best product usually loses, because the people who are really good at making products are usually really bad at marketing.
What I quickly realized is the best marketing wins.
And the way to be the best marketer is to have a solid system underneath it. For a long time, my "marketing strategy" (and most people's) looked like this:
- Launch a new feature
- Post about it on social media a few times
- Pray that customers show up
What that leads to is inconsistent sales. It certainly wouldn't be called a strategy.
That's why Justin Brooke's system blew me away. He's been in the marketing world for 20 years, and when he talks, I listen.
Justin calls it CMO HQ, and what I love about this system is how simple it is. The entire thing is just folders and markdown files.
Justin is a copywriter, not a developer, and he built a full marketing and sales operating system that works with any AI. Claude, Codex, Hermes, OpenClaw. Doesn't matter. Because folders and text files are the one format every AI agent already understands.
His mental model is dead simple. Every AI agent needs three things:
- Memory
- Instructions
- Tools
Folders full of historic files are the memory. Markdown files are the instructions. MCPs and CLIs are the tools. If you have all three, the agent can do any job in your business.
And "jobs" is the right word. Justin uses the Jobs To Be Done framework: list every job that has to happen in a business, and every job becomes a folder.
Every folder gets an instruction file. A dentist has phone calls, appointments, billing.
A marketer has paid, earned, and owned channels.
Same system either way.
Here's what it looks like in action. When the Pope mentioned "digital sobriety" in the news, Justin told Claude to write an article about it for one of his niche sites.
Claude wrote the draft, generated the image prompts, waited for approval, then pushed the finished post straight to his blog via MCP. Justin never opened a doc. He just vibed, reviewed, and approved.
The system even watches its own scoreboard. It's connected to kit.com for leads, Google Analytics for clicks, and a monthly P&L drop for the money.
Clicks to leads to sales to profit, all visible to the agent.
But the real genius is how he sells it.
The whole system delivers as a zip file. Inside are two files: an operations manual and an installer.md.
The client hands the installer file to whatever AI they use, and the AI interviews them (name, company, products, channels), then builds the entire folder structure customized to their business.
You can build the system once then let the machine do the work forever.
Here's how to build your own this week:
Step 1: List the jobs to be done.
Write down every recurring job in your business. Content, reporting, lead follow-up, finances. Each job becomes a folder.
Step 2: Have your AI build the structure.
Tell Claude (or your AI of choice) to create a folder per job, plus an HQ folder with files about you and your business. Justin adds a brands folder (with intelligence, departments, and metrics inside), a finance folder, and a shared folder for his team.
Departments split into paid, earned, and owned, the three channels of marketing.
Step 3: Build it in a cloud folder.
Justin builds his in iCloud. Google Drive or Dropbox work the same. Since it's all text files, it syncs instantly, and you can pick up from your phone, laptop, or office machine exactly where you left off.
Step 4: Give it memory.
Create a session-logs folder. When you finish working, say "I'm done" and have the AI write a summary of the session into that folder.
Add a changelog file for big decisions like price changes. This is how the system remembers what you did on Tuesday at 3 p.m.
Step 5: Connect your data.
Export your leads (Justin uses kit.com), sales (ThriveCart), ad data (Google Ads), and monthly P&L, and drop the files into the right folders.
Each folder has instructions, so when a new file lands, the AI knows to build the expense summaries, cost-per-acquisition dashboards, and monthly reports automatically.
Step 6: Package it to sell.
Once your system works, tell your AI: "Write an operations manual for this, section by section. Then write an installer.md that walks any AI through interviewing a new user and building this entire structure for them." Zip those two files together. That's your product.
Step 7: Sell it with a slide deck.
Justin's best sales advice: stop winging your calls. Use a simple deck, even black and white, and run the same one every time. Tweak one slide after every call.
By call seven it's dialed in, and now you can hand it to someone else to sell for you.
That's the entire system.
Folders, markdown files, and instructions. A marketing department that remembers everything, works while you vibe, and ships as a zip file you can sell for thousands.
Build the first folder this week. Reply and tell me which job in your business you'd turn into a folder first. I read every response.
- Corey
P.S. Justin demoed the entire system soup to nuts in this week's episode of the Build With AI podcast.
A few notes/tweets/cool things sourced from the AI community.
1)
Updated Codex masterclass from Riley Brown.
2)
10 free Github repos from my friend Andrew Warner.
3)
My plan to host in person AI meetups.
Thanks for reading the Corey's Notes newsletter. I'd appreciate it if you sent it to a friend that might find it interesting.
Be sure to check out the Build With AI podcast on Apple, Spotify, and YouTube. If you're a non-technical entrepreneur who wants to learn how to make money with AI, you'll love it.
Be back next week.
-Corey
