I design content systems where humans set the standard and AI does the heavy lifting.
Most teams use AI to go faster. I use it to go faster without losing expert quality — by designing the handoff points between human judgment and machine throughput.
2023
Content creation was capped at team capacity. Designers juggled five or more tools that didn’t talk to each other, doing manual repurposing that added little value.
Quality depended entirely on how many hours a reviewer had.
Now
Human expertise gets encoded into reusable standards. AI handles throughput within gated pipelines. Corrections are captured and codified automatically.
Quality is a function of system design, not headcount.
Every content system I design follows this architecture.
Each cycle feeds back — the human role gradually shifts from reviewing outputs to maintaining standards.
The loop as a product — humans define intent and delivery (steps 1–2), AI generates research and structure (step 3), humans review before sign-off (step 5).
What the human gate catches
Five categories of corrections that only humans can make — and that the system learns from.
How the system matures
Humans gradually shift from point-by-point review to maintaining standards and governance.
Line-by-line review
Every AI output reviewed manually. Slow, but you’re building the dataset of what good looks like.
Golden examples and rubrics
Human expertise encoded into reusable standards. AI starts getting it right more often.
Data flywheel
The system captures corrections automatically and writes them back into its own guidelines.
Governance
Humans focus on strategy, risk assessment, and release sign-off. The system handles the rest.