Most marketing orgs split the team by channel — Paid Media, SEO, CRM, Analytics, RevOps — and each function reports its own number without anyone owning the customer end to end. This is the folder system I built to run demand generation through Claude Code instead: one brief per audience, with content, ads, and outbound derived from it and living together in the same directory.
.zip — full folder structure, pre-filled with the setup prompts for Claude Code
00 Brand, 01 Content, 02 Always-On, 03 Campaigns, 04 Reports, plus .claude/agents and .claude/skills.voice.md, brand-strategy.md, visual-guidelines.md, and CLAUDE.md each contain the actual prompt to run in Claude Code, in the right order, with instructions for what to do once the real file is generated.brief.md for always-on B2B/B2C content and a duplicable example-segment/ folder for your first niche (ABM-style) campaign.Every company splits the marketing team by channel: Paid Media, SEO, CRM, Analytics, RevOps. Each one measures its own number, without talking to the others. The problem isn't the channel. Nobody specializes in the customer.
AI changed that math. Today you can build a team of generalist specialists: focused not on a channel, but on orchestrating the end-to-end purchase, across multiple channels, for the same audience. Instead of a CRM analyst, a RevOps analyst, and an Analytics analyst, you have ONE specialist in the ICP "SaaS CEOs up to 50 people" who breathes that audience.
That concept becomes 3 connected pillars:
Campaigns are organized by target audience, not by channel. If the target is real estate agencies in Portugal, content, ads, and outbound for that audience live together in the same folder.
Rich material is multichannel by definition. A quiz or ebook is distributed via organic + ads + outbound at the same time, and lives inside the target audience's campaign.
Content is the magnet. It generates the signal (comment, site visit, download) that automations route to outbound. Inbound and outbound aren't separate stages: they're the same machine seen from two sides.
The folder below is the materialization of this. It's the folder my Claude Code uses to operate the 3 pillars with me, every day.
ZM_Demand_Gen/ ├── CLAUDE.md ← operating instructions (the "brain") ├── .claude/ │ ├── agents/ ← subagents (campaign copy, audit...) │ └── skills/ ← custom skills │ ├── 00 Brand/ ← identity shared by the 3 pillars │ ├── voice.md │ ├── brand-strategy.md │ ├── brandbook.html │ └── Assets/ (photos, logos) │ ├── 01 Content/ ← recurring output, no defined ICP │ ├── YYYY-MM/<post>/ (posts from the daily cadence) │ ├── Classes/ │ ├── Commercial Proposals/ │ ├── Research/ (third-party market reports) │ └── Webinars/ │ ├── 02 Always-On/ ← recurring operating engine │ ├── Docs/ (strategy, visual specs, dictionaries) │ ├── Scripts/ │ │ ├── Publishing/ (publishing: LinkedIn, Instagram, X) │ │ └── Automations/ (comment-to-dm, site-visitor-to-outbound) │ └── Campaigns/ │ ├── B2B/ {brief, Content, Ads/Creative, Outbound, Signals} │ └── B2C/ {brief, Content, Ads/Creative, Outbound, Signals} │ ├── 03 Campaigns/ ← tactical niche campaigns (ABM) │ └── <segment>/ e.g. RealEstate-PT/ │ ├── brief.md │ ├── Content/ │ ├── Ads/Creative/ │ ├── Outbound/ │ ├── Signals/ │ └── Scripts/ │ └── 04 Reports/ ← consolidated cross-campaign performance
CLAUDE.md — the operating brain. The file Claude Code automatically loads in every session: the agent's role (demand gen strategist + copywriter), voice rules, publishing cadence, the creation workflow (hook → draft → review → save → update the board), where to save each type of output, and which skills to call for each task. Treat it like code — every time I correct the agent twice for the same reason, the correction becomes a line in CLAUDE.md. It points to deeper docs instead of duplicating them.
.claude/ — agents and skills. agents/ holds specialized subagents (a campaign copywriter that reads any campaign's brief.md, a copy auditor that reviews "blind" against voice.md). Golden rule: generic agents that read the campaign context, not one agent per campaign. skills/ holds custom skills — post pipelines, hook generators, cross-platform repurposing.
00 Brand — shared identity. Voice and visual identity live outside the pillars because they feed all three: the same tone in the organic post, the ad copy, and the outbound email. voice.md carries the archetype, tone rules, banned vocabulary, and pre-publish review filter — the agent loads it before writing anything. brand-strategy.md holds positioning, content pillars, ICPs, and how the 3 pillars feed each other. brandbook.html is the visual reference: colors, typography, components, grid. It's the first folder to set up — without documented voice, every piece of copy comes out with a different tone and authority doesn't compound.
01 Content — recurring content with no ICP. The output of the daily cadence (1 post/day, format varying by day of the week) when the content is for the general feed, with no specific target audience. Each post gets a subfolder YYYY-MM/<post-name>/ with final copy and media. It also holds long-form materials (classes, webinars, proposals) and Research/ with third-party market reports that become topic fodder. This is the Authority pillar in continuous mode — the goal isn't to go viral, it's to generate a constant intent signal the other pillars consume.
02 Always-On — the engine. Everything that runs continuously, with no end date. Docs/ is the canonical reference (content strategy, carousel/infographic specs, management board field dictionary). Scripts/Publishing/ holds the scripts that publish or schedule via API. Scripts/Automations/ is the glue between inbound and outbound — each automation is a trigger → action → destination rule, documented in its own folder before being coded (comment-to-dm; site-visitor-to-outbound). Campaigns/B2B/ and B2C/ are always-on campaigns for a broad ICP, with the same internal structure as niche campaigns. The distinction between a script and an automation: a script runs once and finishes; an automation keeps listening for an event and reacts on its own — and can call a script under the hood.
03 Campaigns — tactical niche campaigns (ABM). One folder per segment, with a defined start and end. Real example: RealEstate-PT/ (prospecting real estate agencies in Portugal). brief.md — ICP, pain point, angle, offer — is the file subagents read before writing any campaign copy. Content/ holds targeted posts plus the rich material that anchors the campaign. Ads/Creative/ holds ad creatives. Outbound/ holds lists, sequences, and outreach scripts. Signals/ holds raw intent data. A new campaign is just this structure duplicated with a different segment — the 3 pillars live together because the same rich material is distributed organically, amplified in the ad, and used as the reason for cold outreach, all for the same target.
04 Reports — consolidated performance. The cross-campaign view: what each pillar generated, which campaigns perform, where to reallocate effort. Per-post metrics live in the management board; consolidated numbers live here.
| Content type | Folder |
|---|---|
| Recurring post, no defined ICP | 01 Content/YYYY-MM/<post>/ |
| Recurring post targeted at an ICP (B2B or B2C) | 02 Always-On/Campaigns/<ICP>/Content/YYYY-MM/<post>/ |
| Niche campaign content | 03 Campaigns/<segment>/Content/ |
| Rich material (quiz, ebook, tool) | Inside the Content/ of the campaign that will distribute it. Never loose. |
All connected via MCP or API:
| Tool | Role in the system |
|---|---|
| Figma | Creation of visuals (carousels, infographics, posts) from ready-made templates in a reference file. Claude Code draws directly on the canvas via MCP and exports the final piece. |
| ClickUp | Content management board. Each post is a card with a status (ideas → research → approve → booked → published → analyzed), publish date, and post-publish metrics that become input for the next topic. |
| Notion | Home for rich materials distributed in comment-gates and the system's living documentation. |
| Nuvia | Outbound engine — the email/DM sequences that reach out to leads generated by content and ads signals. |
| Apify | Signal collection — scraping who comments, interacts, and visits, feeding the Signals/ folders and the routing automations to outbound. |
Voice → strategy → brain → visual identity → tools → board → cadence → campaign → automation. Each layer only works on top of the previous one.
You're going to create the file "00 Brand/voice.md": the voice guide you'll load yourself before writing any of my copy. Before writing, interview me. One question at a time: 1. Ask for 3 to 5 pieces of my writing that I feel sound like me (posts, emails, messages). 2. Ask who I write for (role, technical level) and how I want to sound to that person, in 3 adjectives. 3. Ask what I NEVER want to sound like (e.g. LinkedIn guru, course seller). With the answers, generate voice.md with these sections: - Archetype (1 paragraph) - 5 to 10 actionable tone rules (no "be authentic" — rules you can check against a finished text) - Banned vocabulary (banned words and clichés) - Standard text structure (how it opens, develops, closes, CTA style) - 3 on-brand and 2 off-brand examples, each with a 1-line explanation - Review filter: 5 yes/no questions to run before publishing any text Create the "00 Brand" folder if it doesn't exist and save the file there.
You're going to create the file "00 Brand/brand-strategy.md": the strategy that connects my 3 demand generation pillars (Content and Authority, Ads and Performance, GTM and Outreach). Before writing, interview me. One question at a time: 1. What I sell and to whom (my ICPs, separating B2B and B2C if both apply). 2. What transformation I deliver and what proof I have of it (cases, numbers, results). 3. What topics I know well enough to sustain daily content without making things up. 4. Where my audience actually is (LinkedIn, Instagram, X, email, communities). With the answers, generate brand-strategy.md with: - Positioning in 1 paragraph (who I help, to do what, how) - 3 to 5 content pillars (topics), each with an angle and a sample topic idea - ICP definitions: main pain point, buying trigger, preferred channel - Weekly publishing cadence (day-by-day table, with each day's format) - How each pillar feeds the other two (content → signal → outbound; content → ads for the same audience) Save it in the "00 Brand" folder.
Read "00 Brand/voice.md" and "00 Brand/brand-strategy.md" and create CLAUDE.md at the root of the folder: the operating instructions you'll load in every session from now on. Structure it with these blocks: 1. Role: Senior Demand Generation Strategist + Copywriter, covering 3 pillars (Content and Authority, Ads and Performance, GTM and Outreach). 2. Strategy: the 3 pillars and the 3 non-negotiable premises — campaigns organized by target audience, not channel; rich material is multichannel by definition; content is the magnet that generates the signal automations route to outbound. Summarize the essentials of brand-strategy.md. 3. Voice: the 5 most important rules from voice.md, with an instruction to ALWAYS load the full file before writing. 4. Publishing cadence: the day-by-day table from brand-strategy.md. 5. Creation workflow with approval gates: HOOK (5 options, wait for my choice) → FULL DRAFT → REVIEW (voice.md filter) → SAVE in the right folder → UPDATE the management board. 6. Where to save: decision rule between "01 Content" (no ICP), "02 Always-On/ Campaigns" (broad ICP B2B/B2C), and "03 Campaigns" (niche). 7. Folder structure: the full 00-04 tree. 8. Execution rules: deliver directly without confirmation loops beyond the gates; when I correct something, adopt it immediately without justifying the previous output. Be concise: CLAUDE.md points to the other docs, it doesn't duplicate them. After creating the file, also create the complete 00-04 folder structure.
You're going to create the file "00 Brand/visual-guidelines.md": the minimum visual spec for any piece (carousel, infographic, static post). Ask me, one question at a time: 1. My colors. If I don't have a defined palette, ask for 2-3 visual references I like and propose: background color, text color, 1 accent color, and neutrals (4-6 colors total). 2. My fonts (heading and body). If I don't have any, propose Google Fonts pairings that fit the voice.md archetype. Generate visual-guidelines.md with: - Palette with hex codes and each color's function - Typography: family, weights, and sizes for heading, subheading, and body - Standard piece dimensions: 1080×1350px (4:5), 9% safety margin - 3 to 5 usage rules (e.g. accent color on 1 element per piece, max 10% of the area; no gradients) Save it in the "00 Brand" folder.
Create a "Content" list in my management tool (ClickUp, in my case) with: - Status: ideas → research → approve → booked → published → analyzed - Custom fields: Content type (dropdown), Language (dropdown), Target audience (text), CTA (text), Final post link (url), Final media link (url), Impressions (number), Reactions (number), Comments (number), Reposts (number) Then document in "02 Always-On/Docs/board-fields.md" the field dictionary (name, type, meaning of each one), so you can reference it yourself in future sessions without having to rediscover the structure.
This board is what closes the learning loop: post-publish metrics go into the card and become input for the next topic.
Want more on AI-assisted marketing operations?
Read Marketing Operations articles →Need prompts for the rest of your stack?
Open Prompt Library →