Carl Benedikt Frey, a professor at the Oxford Internet Institute, argued in a June 14, 2026 Financial Times opinion piece titled "'Can a machine do this job?' is the wrong question" that AI will bring about a "self-service economy" by shifting tasks once handled by professionals onto consumers themselves, rather than simply destroying jobs.
June 14, 2026 · The AI & Work Debate
"Can a machine do this job?" is the wrong question
The labor debate is shifting away from a yes-or-no verdict on whole occupations toward a granular view: which tasks get automated, what stays human, and whether a machine should — not just can — take over.
80%
of a job — the tedious, repetitive part people dislike — is where AI is most useful
20%
judgment, relationships & creativity — the part that stubbornly still needs a human
CAN→ SHOULD
the new test weighs cost of failure, not just technical feasibility
Splitting the job: what AI takes vs. what stays human
80% AI
automate the repetitive work
20% Human
judgment · relationships · creativity
Each block ≈ one fifth of the job. The better question: how do you make AI handle the 80% you hate?
Reframing the question
Old question · 2010s–2024
Unit: the whole occupation
Criterion: can a machine do it
Focus: replacement
Test: technical feasibility
New question · 2025–2026
Unit: the individual task
Criterion: should it be delegated
Focus: augmentation & remaining role
Test: cost of failure, quality of outcome
Why "should" can outweigh "can"
Generative AI may handle a substantial share of tasks — but where a mistake is catastrophic, capability does not equal permission.
Cost of failure
Emergency-call operators: extreme stakes keep humans in the loop
Quality of outcome
Education: AI drafts and grades; emotional support stays human
Scarcity rises
A Jevons-style effect: efficiency lifts demand, making human skills more valuable
Pragmatic view
Let AI absorb the tedious, repetitive work and free people for judgment, relationships and creativity. Freed time can become a path to promotion.
Cautious view
AI can perform tasks, but human experience guarantees the quality of outcomes. Full automation is unsuitable where the price of error is high — and freed time can also become a path to irrelevance.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…