Most teams treat content as one job: someone writes a thing. Real content teams don’t work that way. They split the work across seven roles, and each role has a different skill, different tooling, and different output.
Rob Hoffman wrote up the seven-role breakdown in his “Future of Content” pieces. The roles are:
- Strategist - decides what content to make, why, and for which audience
- Researcher - collects sources, finds the angle, validates the claim
- Outliner - structures the argument into a draft scaffold
- Drafter - writes the prose
- Editor - tightens, kills clichés, sharpens claims
- Distributor - adapts the piece for channels (LinkedIn, X, Substack, podcast)
- Analyst - tracks performance and feeds learnings back to the strategist
A solo content person does all seven, badly. A real content team has different people for each, and the team is faster and produces better work than the solo writer.
The AI-content-tooling industry is mostly built around role 4 (drafting). Chat tools, AI ghostwriters, “write me a LinkedIn post.” Some have added editing. Almost nothing addresses strategist, researcher, or distributor systematically.
That’s the opportunity.
The Architecture I’d Build
Seven specialized agents, one per role. Each has a tight, opinionated system prompt and a small toolset matched to its job. Output flows through a shared dashboard where the human (the content director) approves or kicks back at each stage.
Strategist agent. Reads your goals (YAML OKRs), audience definitions, and recent performance data. Proposes 3-5 content opportunities per cycle with target audience, angle, and channel mix. Output: a content brief.
Researcher agent. Takes the brief, pulls sources (web, your own knowledge base, your past pieces, your CRM). Validates the angle isn’t already saturated. Output: a research dossier with quotes, stats, and counterarguments.
Outliner agent. Takes the brief + dossier. Proposes 2-3 different structural approaches (problem-solution, war story, contrarian take). Operator picks one. Output: a structured outline.
Drafter agent. Takes the outline + voice guide. Writes the piece in your voice. Output: a draft.
Editor agent. Takes the draft and a strict voice rules file. Cuts clichés. Tightens claims. Flags hedging. Output: an edited piece + a punch list of changes.
Distributor agent. Takes the published piece. Generates channel-specific adaptations: LinkedIn post pulling the strongest insight, X thread breaking the argument into beats, Substack email with the full piece + intro, Reddit posts for relevant subs. Output: adaptations ready to schedule.
Analyst agent. After 30 days, pulls engagement data per channel. Identifies what worked. Feeds insights back to the strategist for the next cycle. Output: performance memo.
The Shared Dashboard
The dashboard is the orchestration layer. Each piece moves through stages. The operator (the content director) sees:
- Pieces currently in research (strategist done, researcher running)
- Pieces awaiting outline approval
- Pieces in editing
- Pieces ready to publish
- Pieces post-publication awaiting analyst review
At each stage, the operator can approve, send back with notes, or kill. The agents act on the approvals. The dashboard tracks state.
This is the part that’s missing from every “AI content tool” I’ve used. They’re stuck at role 4 with a chat box. They don’t have a workflow because they don’t have a model of the team.
Why This Works
Three reasons:
Specialization beats generalization. A strategist agent that does ONLY strategy will be sharper at it than a chat-mode agent that does everything. Same logic that makes specialized human teams beat solo writers.
The dashboard is the manager. Without a workflow layer, AI content tools become “I asked the AI for a draft and got a draft.” With a workflow, you get a process the operator manages, not a transaction the operator completes.
The human stays in the director role. Per the take in Autonomous Agents Need a Director, AI in creative work belongs on the paintbrush. The human stays the painter. The seven-agent model keeps the human deciding what to make, while the agents handle the labor of making it. No autonomous content slop. Just a faster team.
What Already Exists
Partial pieces:
- Strategist: Some marketing tools have AI brief generation. None integrate with your goals/CRM the way this would.
- Researcher: Perplexity is the closest thing. Could be the engine for this role.
- Drafter: Saturated category. Any LLM does this.
- Editor: Some tools (Hemingway, Grammarly’s AI mode) exist but aren’t workflow-integrated.
- Distributor: Buffer, Hootsuite, Typefully all do scheduling but don’t generate channel-specific variants from a single canonical piece.
- Analyst: Channel analytics exist; nothing feeds back into strategy automatically.
The opportunity is the integration. Not building each agent. Building the dashboard that makes them work as a team.
What I’d Charge For
This is a service offering, not just a tool. The AI content team works in your voice, with your knowledge base, against your goals. That requires onboarding (define your voice, load your KB, set your KRs). That’s where the human consulting happens.
The tooling is the leverage. The consulting is the customization. The combination is what makes it work for any specific company.
Notes for a product brief, when I get to it.