Human Depth, AI Scale Task Filter
Assign deep craft to humans and high-volume synthesis to AI
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 6
- Confidence
- 97%
This task-allocation filter compares two dimensions: depth of expertise and required scale. If success depends on exceptional creative craft, subtle interpretation, or a deep understanding that a skilled human already performs well, keep the decisive work human-led. If the task requires reading, transforming, or personalizing more material than a person can process economically, use AI to supply scale. Formulaic informational writing may therefore be highly automatable, while diagnosing why top-tier writing works remains human territory. Mixed tasks combine both advantages: AI performs broad collection or first-pass transformation, and a human supplies judgment, standards, and final approval. The model prevents teams from treating every task as equally suitable for automation.
Origin
Extracted from Marketing Against The Grain
Core principles
- 01Use comparative advantage rather than hype to allocate work
- 02Keep deep creative and expert judgment with skilled humans
- 03Use AI where volume exceeds practical human time and effort
- 04Distinguish formulaic output from exceptional craft
- 05Require human oversight when quality depends on subtle judgment
How to run it
- 1
Define the success criterion
State what makes the output valuable: exceptional insight, creative distinction, speed, coverage, personalization, or consistency.
Pro tip Separate must-have quality from merely desirable polish.
Watch out A vague definition of success makes any automation decision arbitrary.
- 2
Assess human depth
Determine whether a skilled person can perform the task substantially better through expertise, taste, or reasoning.
Pro tip Use evidence from top performers rather than average output.
Watch out Do not classify a task as automatable merely because average human work is mediocre.
- 3
Assess the scale burden
Measure how much reading, repetition, personalization, or throughput the task demands. Identify where time and effort make all-human execution impractical.
Pro tip Estimate units per day and the research required per unit.
Watch out Large volume does not remove the need for accuracy controls.
- 4
Choose the operating mode
Keep depth-dominant tasks human-led, make scale-dominant tasks AI-led, and design a hybrid workflow for tasks requiring both.
Pro tip Let AI handle collection and variation while humans own standards and judgment.
Watch out Do not give AI final authority merely because it performed the bulk of the work.
- 5
Test representative outputs
Compare AI, human, and hybrid results on real examples. Evaluate both quality and throughput before expanding the workflow.
Pro tip Include difficult edge cases in the sample.
Watch out A successful easy example can conceal failures on nuanced cases.
- 6
Set review boundaries
Specify what humans must verify, revise, or approve before the output is used. Tighten oversight as consequences increase.
Pro tip Turn recurring review findings into explicit AI guardrails.
Watch out Hallucinations and averaged responses can survive a superficial review.
In the wild
AI reads one hundred company websites and drafts a relevant opening for each prospect, a workload a human cannot research in a day. A salesperson reviews the drafts, corrects weak claims, and handles the actual relationship judgment.
→ The team gains research scale without surrendering final communication judgment.
A company automates formulaic informational articles but retains a senior editor for pieces where voice, original insight, and cultural judgment create the value. AI supports research and formatting rather than deciding what makes the piece exceptional.
→ Automation reduces routine effort while distinctive work remains expert-led.
Common mistakes
Automating because a task uses words
Language tasks vary widely; formulaic production and exceptional creative judgment do not have the same automation profile.
Ignoring AI's scale advantage
Keeping repetitive research entirely manual wastes AI's strongest comparative advantage.
Removing expert oversight
Scale does not compensate for hallucinations, shallow reasoning, or failure to recognize exceptional quality.
Is it for you?
Best for
Managers designing practical AI workflows across research, personalization, writing, and analysis.
Not ideal for
High-stakes decisions where neither AI scale nor unaudited human intuition is sufficient.
From the transcript
“anything that a human can do really, really well, AI really can't do. The things that a human can't do because it doesn't have the…”
“But if you're like, hey, I need a deep, deep, deep understanding of something, a human's better at that still today.”
“And I suspect that write in is formulaic, right? It is basically formulaic, informational, educational. There is one answer. It's pretty easy to have someone…”
From the episode
Generate 10x Views On A LinkedIn Post With These GPT-4o Prompts