Building in Public: How AI Agents Run Our Content Factory

Description and use

A company employs 19 AI agents to draft content, prepare sales follow‑ups, generate SEO pages, and analyze results, with human review for public output. The workflow is split into roles—writing, sales, SEO, analytics—improving speed and quality.

Problem

Automating content creation and sales follow‑ups while maintaining quality and preventing fabricated data.

Outcome

The workflow increases content output (67 pieces in a week) while keeping quality through human review and iterative guardrail improvements.

Workflow

• 1. Analytics agent reads weekly performance data and identifies successful and stalled topics.
• 2. Topics and constraints are set for the week.
• 3. Writing agent drafts articles from approved briefs.
• 4. Sales agent prepares replies and follow‑ups for a person to send.
• 5. SEO agent generates pages from vetted templates and checks them against rules.
• 6. Human reviewers approve any content that goes public.
• 7. Analytics agent reads results, records failures, and suggests instruction adjustments.
• 8. Team reviews adjustments and updates agent prompts.
• 9. Repeat cycle each week.

Evidence

Building in public means we show how our AI agents actually run a content factory. They draft articles, prepare sales follow‑ups, generate SEO pages and read analytics, while people set the goals and approve the result.

Source

We split the work into narrow roles: Writing agent, Sales agent, SEO agent, Analytics agent.

Source

Early in the week, analytics shows what worked and what stalled. We pick topics and set constraints. The agents draft, and a person reviews anything that goes public.

Source

They repeat themselves, drift from our tone, and sometimes claim things they can't support. That is why we forbid invented data.

Source