The capability thesis
AI Is a Human Skill
The quality of the answer is constrained by the quality of the thinking around it.
Definition
AI capability is not the ability to operate a tool. It is the human ability to frame, question, judge, imagine and learn with machine intelligence.
The argument
Learn it like painting
You do not learn to paint by reading the manual for a brush. Craft emerges through practice, feedback, taste and reflection.
Thinking partner
Use AI to expose blind spots and generate better questions—not simply as an output machine.
The model
Mindset → Skillset → Toolset.
Metaphor / example
A precise prompt can still solve the wrong problem. Metacognition helps people notice the assumptions shaping the exchange.
Leadership implication
Build practice into real work and reward better judgement, not prompt theatre.
Common misunderstanding
Calling AI a human skill does not minimise technical expertise. It explains why technical access alone is insufficient.
External evidence and interpretation
What the research shows—and what Mark reads into it.
Published finding
In a 2023 field experiment with 758 BCG consultants, GPT-4 improved speed and quality on tasks inside the AI’s capability frontier, but the authors describe a ‘jagged’ frontier where AI falls short on other tasks—so human judgement about when to rely on it matters.
Harvard Business School / D^3 Institute — Navigating the Jagged Technological Frontier (HBS Working Paper 24-013, September 2023) ↗Mark’s interpretation
Judging where AI helps and where it misleads is a human skill—Mark’s interpretation of the ‘jagged frontier’ finding.
Published finding
In one company’s customer-support operation (5,179 agents), access to a generative-AI conversational assistant raised issues resolved per hour by about 14% on average in the November 2023 working-paper version, with much larger gains for novice and less-experienced agents. Later published versions may report revised figures.
National Bureau of Economic Research — Generative AI at Work (NBER Working Paper 31161, April 2023, revised November 2023) ↗Mark’s interpretation
The same tool produced very different gains for different people, suggesting capability depends on the person and context, not access alone.
Questions people ask
Why call AI a human skill?+
Because useful outcomes depend on framing, questioning, judgement, imagination and reflection—not merely access to a model or prompt pattern.
Does technical AI expertise still matter?+
Yes. Technical expertise matters greatly; the point is that technical access alone cannot create organisational judgement or learning.
Sources
Harvard Business School / D^3 Institute — Navigating the Jagged Technological Frontier (HBS Working Paper 24-013, September 2023) ↗National Bureau of Economic Research — Generative AI at Work (NBER Working Paper 31161, April 2023, revised November 2023) ↗Mark Cameron — AI profile ↗Research library →Bring it into the room
Start a speaking enquiry →