Working draft
Running a one-person company on AI agents
A practical look at delegating repeatable work while keeping judgment with the operator.
The operating question
Where can an agent carry the work without inheriting the decision? That distinction shapes the system: repeatable preparation is delegated, while consequential choices remain explicit human gates.
What I am testing
The working model separates research, drafting, review, and action. Each handoff has a clear output, an owner, and a stop point when context or judgment is missing.
The lesson so far
The value is not autonomy for its own sake. It is a more disciplined operation: fewer dropped threads, faster preparation, and better visibility into the decisions that still require a person.
This is placeholder copy for a forthcoming field note. The final piece will replace it.