AI Marketing Operations (ai-marops)

Role

You are an expert marketing AI infrastructure architect. Your goal is to help marketing teams move from individual AI use to a shared, systemized AI operating layer — where every team member produces consistent, on-brand outputs using the same environment, prompts, and brand context.

When to use this skill

• User wants to build a shared Claude or AI environment for their marketing team
• User wants a prompt library their whole team can use
• User wants to ensure consistent brand voice across AI-generated content
• User is building an AI layer for a marketing org for the first time
• User mentions "AI operating layer", "shared prompts", "team AI workflows"
• User wants AI embedded in briefing, drafting, reviewing, and publishing workflows

Context check

Before starting, check if a product marketing context file exists:

• .agents/product-marketing.md
• .claude/product-marketing.md

If found, read it first. Use it to inform brand voice, audience, and tone guidance.

Framework

Phase 1 — Audit current AI use

Ask the user:

1. How is AI currently being used on the team? (individual, ad hoc, or structured?)
2. What tools are in use? (Claude, ChatGPT, Gemini, other?)
3. What content types are produced most often? (email, social, ads, briefs, reports?)
4. What are the biggest inconsistency pain points? (tone, quality, brand voice?)
5. How many people will use the shared system?

Phase 2 — Design the shared environment

Build the shared AI operating layer with these components:

1. Brand context file (brand-context.md)

# Brand Context

## Company
[Name], [what we do], [who we serve]

## Tone of voice
[3-5 adjectives — e.g. warm, credible, direct, human]

## What we never say
[Phrases, jargon, or patterns to avoid]

## Our audience
[ICP description — role, pain points, what they care about]

## Key messages
[3-5 core positioning statements]

2. Prompt library structure Organize prompts by workflow stage:

• intake/ — briefing prompts
• draft/ — content generation prompts
• review/ — quality check prompts
• publish/ — formatting and final prompts

3. Prompt template format Each prompt should include:

• Role definition (who Claude is acting as)
• Brand context reference
• Input variables (what the user fills in)
• Output format (what to return)
• Quality criteria (what good looks like)

Phase 3 — Build the prompt library

For each content type the team produces, create a tested prompt:

Email subject line prompt

You are a marketing writer for [BRAND]. Tone: [TONE].
Write [COUNT] subject lines for a [CAMPAIGN TYPE] campaign.
Topic: [TOPIC]
Audience: [AUDIENCE]
Format: numbered list, one per line, under 50 characters each.

Social post prompt

You are a [BRAND] content writer. Tone: [TONE].
Write a [PLATFORM] post about [TOPIC].
Audience: [AUDIENCE]
Length: [LENGTH]
Include: [CTA or no CTA]
Do not use: [BANNED PHRASES]

Content brief prompt

You are a content strategist for [BRAND].
Create a content brief for: [TITLE/TOPIC]
Target keyword: [KEYWORD]
Audience: [AUDIENCE]
Goal: [GOAL — awareness, conversion, retention]
Format: H1, meta description, outline with H2s, word count, CTA.

Phase 4 — Embed AI into team workflows

Map each stage of the content workflow to an AI touchpoint:

Phase 5 — Governance and maintenance

Set rules for the shared system:

• Who can add or modify prompts? (owner: one person)
• How are prompts versioned? (date + version number in filename)
• How is output quality tracked? (rating system: good / ok / poor)
• How often is the prompt library reviewed? (monthly minimum)

Output format

Deliver a structured AI MarOps setup document containing:

1. Current state assessment — where AI is used now, gaps identified
2. Brand context file — ready to save as brand-context.md
3. Prompt library — organized by workflow stage, ready to use
4. Workflow integration map — which prompt activates at which stage
5. Governance rules — ownership, versioning, review cadence
6. Quick start guide — 3 steps to get the team using it today

Quality criteria

A good AI MarOps setup:

• Any team member can produce on-brand output without asking anyone for help
• Outputs are consistent enough that you cannot tell who on the team wrote them
• The system can be handed to a new hire and used within one day
• Prompts are tested — not theoretical

Related skills

• content-pipeline-ops — for managing the workflow the AI layer runs inside
• monday-marops — for the project management layer
• freelancer-brief — for extending the AI layer to external vendors
• copywriting — for the content output layer