Product-Proximity Human-in-the-Loop Rule
Increase human creative control as content moves farther from the product.
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 5
- Confidence
- 96%
Place each content task on a continuum from tightly grounded product material to broad editorial or highly creative work. Product marketing, tailored case studies, and buyer-use-case pages can often be generated effectively because the model has concrete facts, constraints, and an obvious objective. As the task moves toward original cultural commentary, distinctive editorial products, or ambitious creative ideas, AI quality becomes less dependable and human judgment should increase. The rule determines the human-in-the-loop level before production begins: more factual grounding can permit lighter review, while greater abstraction, novelty, or reputational exposure requires human ideation, editing, and approval. Teams should periodically recalibrate the boundary as models improve, but never confuse fluent writing with verified truth or genuine taste.
Origin
The three speakers reached this rule while debating whether AI could independently create account-specific microsites versus editorial newsletters and highly creative content.
Core principles
- 01Match oversight to the type of content, not one universal policy.
- 02Let AI handle grounded, product-adjacent material more independently.
- 03Increase human involvement as abstraction and creative ambition rise.
- 04Evaluate quality against the intended audience and purpose.
- 05Keep humans accountable for consequential claims.
How to run it
- 1
Classify Proximity
Decide how closely the requested asset is tied to verified product facts, buyer needs, and a defined use case.
Pro tip Use a simple scale from direct product explanation to open-ended editorial creation.
Watch out A product mention alone does not make an asset well grounded.
- 2
Assess Creative Stakes
Estimate how much originality, taste, cultural awareness, and reputational judgment the task requires.
Pro tip Treat flagship editorial work as higher stakes than routine variants.
Watch out Fluent generic copy can appear finished while lacking a valuable idea.
- 3
Set Human Involvement
Assign AI, reviewer, editor, or human creator roles according to proximity and stakes.
Pro tip Reserve senior creative attention for the farthest and most consequential end of the spectrum.
Watch out Do not use review as a rubber stamp.
- 4
Verify the Output
Check factual grounding, audience resonance, tone, and originality before release.
Pro tip Compare the output with strong human examples in the same format.
Watch out Never assume model confidence indicates accuracy.
- 5
Recalibrate
Record quality results and move task types toward more or less automation based on evidence.
Pro tip Reassess by task category rather than declaring one model universally capable.
Watch out Do not lower standards merely to increase automation.
In the wild
A team lets an agent draft an account-specific product page from verified company research and product documentation, then gives it a factual review. For a flagship editorial newsletter intended to establish a distinctive industry point of view, a human develops the central idea and structure before using AI for research support and variants.
→ Human creative effort is concentrated where abstraction and taste matter most, while grounded work gains efficiency.
Common mistakes
Using One Review Policy
Applying identical oversight to routine product variants and flagship creative work either wastes human effort or accepts weak quality.
Mistaking Fluency for Creativity
Polished prose can still lack a resonant idea, original judgment, or accurate grounding.
Automating High-Stakes Claims
Even product-adjacent content needs specialized review when errors could create material harm.
Is it for you?
Best for
It is best for teams deciding where AI can safely automate content and where expert creative judgment remains essential.
Not ideal for
It is not ideal as permission to publish high-stakes product, legal, medical, financial, or reputational claims without specialized review.
From the transcript
“the further AI content gets away from the product buyer use cases and out to like more broader ideas, the worse it is.”
“I think the more creative the content is, the more the human is still involved.”
From the episode
Marketing Tactics You Need to Learn in the AI Era ft. Morning Brew Founder