Build a one-person content engine with 3 AI agents

Build a one-person content engine with 3 AI agents

The easiest way to make AI output feel “generic” is to ask one model to do everything: research, decide, write, edit. You get average thinking, average structure, average voice.

Instead, split the work like a real team does. Then you use AI as junior labor in each role while you keep judgment and taste.

The three roles that matter

1. Research (truth)
2. Writing (structure)
3. Editing (taste)

1) The Researcher (truth, not trivia)

Goal: produce a brief that makes writing easy.

Inputs to provide:

• Who is the reader?
• What is the promised outcome?
• What changed recently?
• What examples should we analyze?
• What do we not want to claim?

Output format (required):

• 10 real examples (with links or descriptions)
• 3 counterexamples (why the “obvious” advice fails)
• 5 claims you can defend (not vibes)
• 1 contrarian reframe (your POV)

If your brief is weak, the draft will be weak. Always.

2) The Writer (complete first pass)

Goal: speed + completeness. You want a full draft you can edit, not perfect prose.

Constraints to set:

• word count range
• required sections
• audience sophistication (beginner vs operator)
• tone (direct, no fluff)

Output: a messy but complete draft with:

• a strong opening
• a clear framework
• steps with examples
• a checklist
• a CTA path (what the reader should do next)

3) The Editor (make it feel human)

Goal: enforce your quality bar.

Editing checklist:

• Remove paragraphs that could belong to any article.
• Add one concrete example per major claim.
• Add trade-offs (what fails, what to avoid).
• Tighten intros (start with the problem and promise).
• Replace “should” with “do this next”.

Rule: if it sounds like LinkedIn filler, cut it.

The minimal workflow (repeat weekly)

• Monday: topic + outcome + 10 examples
• Tuesday: research brief
• Wednesday: draft
• Thursday: edit + publish
• Friday: distribution + feedback capture

Example (prompting the system)

Research brief prompt: “You are my research analyst. Topic: [X]. Reader: [Y]. Find 10 examples from the last 12 months, 3 counterexamples, and summarize what changed recently. Output as bullets, include sources.”

Editor prompt: “You are my editor. Remove fluff. Add specificity. If a claim lacks an example, add one. If a paragraph could fit any article, delete it. Keep the tone direct.”

Checklist (engine setup)

• One research brief template
• One draft template (sections required)
• One editor checklist
• One weekly cadence slot
• One distribution checklist

Internal links to add

• AI Marketing: Stop writing prompts. Start writing systems.
• AI Marketing: The prompt stack every marketer should steal
• AI Marketing: The one-person media company: your 2026 operating system

CTA (Interactive)

Find your operator type: /ai-marketing/ai-operator-quiz/