← Product Design / Project
AGENT STUDIO
An editorial workflow that gives AI agents distinct responsibilities without handing them creative control.
Problem
Producing a coherent article involves developing a strong idea, researching its claims, making editorial decisions, writing and revising. A single undifferentiated AI chat can blur these responsibilities, making outputs generic or inconsistent and leaving it unclear when a person should intervene.
Approach
I designed a multi agent editorial workflow with five specialized roles: Idea Partner, Researcher, Creative Director, Writer and Editor. Each has a different job. A promising idea is clarified before handoff; research investigates claims and counterarguments; the Creative Director establishes the article’s direction; the Writer drafts it; and the Editor evaluates and revises it.
What happens in practice
- Idea development: The Idea Partner explores the premise and prepares an idea brief. I decide which ideas move forward.
- Editorial direction: The Creative Director assesses the brief and coordinates any research needed before drafting.
- Writing and editing: The Writer develops the article and the Editor checks its coherence and quality. Human review remains part of the process.
Key design decision
The system does not maximize automation. Idea approval, editorial direction and final judgment stay with a human. That makes the workflow slower than fully automatic content generation, but keeps responsibility and authorship visible.
Evidence and current status
A working editorial workflow has been used to develop and produce articles. The system is still being extended. This establishes a functioning process, not yet a measured improvement in article quality or production speed.
My role
Workflow design, AI agent orchestration, creative direction and implementation.