Keep directional decisions human…
When using AI to reduce production costs, maintain human control over directional decisions (what to produce, what argument to make, what judgment to render) while delegating mechanical execution (format conversion, draft expansion, repurposing) to preserve sovereignty through augmentation not replacement.
When This Fires
- When integrating AI tools into your production workflow
- When AI-generated content feels "off" despite being technically competent — likely the AI made directional choices
- When deciding which production steps to automate and which to keep manual (Four-category automation triage)
- Complements Sovereignty check before automation — if the automation… (sovereignty check) and Insert AI as a transformation step between workflow stages… (AI as transformation step) with the production-specific boundary
Common Failure Mode
Directional delegation: "AI, write an article about productivity." The AI produces a competent article with generic arguments that don't reflect your specific perspective, experience, or judgment. You publish it because it's "good enough," but it could have been produced by anyone — your unique voice and thinking are absent.
The Protocol
(1) For each production step, classify it as directional (requires your judgment) or mechanical (requires execution without judgment). (2) Keep all directional steps human: choose the topic yourself, decide the argument yourself, select the evidence yourself, form the conclusions yourself. (3) Delegate mechanical steps to AI: "Here's my outline and key arguments — expand this into a draft." "Here's my article — reformat it for LinkedIn." "Here's my data — produce a visualization." (4) Always review AI outputs before delivery (Sovereignty check before automation — if the automation…): the AI may have made implicit directional decisions during mechanical execution. Correct any divergence from your intended direction. (5) The test: "Could someone reading this output identify it as distinctly mine, reflecting my specific thinking?" If yes, sovereignty is preserved. If it could have been produced by any AI user, the directional decisions were delegated.