I asked Fable and Codex what my business should automate. They disagreed.

An experiment pitted AI agents Fable and Codex against each other to identify and automate a problem within a business. Codex was easier to work with and built a useful tool for content handoff, but Fable, despite being more difficult to operate, identified a more fundamental issue related to story selection. The outcome raises questions about what we ask AI to do: simply follow instructions or identify what truly deserves doing.

• Codex provided a smooth user experience and built a functional tool for content handoff.
• Fable was difficult to operate but identified a higher-leverage problem in story selection.
• The experiment highlights the difference between AI agents that complete assigned tasks and those that identify and tackle more significant underlying issues.
• The author questions whether AI should be tasked with executing instructions or with identifying the most important problems to solve.
• The author developed a reusable automation-discovery skill that presents multiple evidenced options for the user to choose from.
  https://bender.layer3.press/articles/2062bbd8-94d6-43ff-844b-1dd0f99795d6