What will be scarce? - By Alex Imas
TL;DR
Starbucks is a warning shot against the pure-automation story — Despite having the technology to automate coffee service, Starbucks reversed course, with CEO Brian Niccol citing handwritten cup notes, ceramic cups, seating, and more baristas as the things that actually improved customer satisfaction.
Alex Imas’s core claim is that AI changes scarcity rather than eliminating it — If AI makes commodity production cheap, the scarce thing won’t just be labor in general but specifically human involvement, provenance, attention, and exclusivity in what he calls the “relational sector.”
History suggests automation shrinks a sector’s share of the economy even as output rises — Imas points to the US moving from about 40% of workers in farming in 1900 to under 2% today, and to Kongsamut-Lashkari-Mestieri-style structural change logic where richer people shift spending toward higher-income-elasticity sectors like services.
The demand shift is driven more by income effects than price effects — Citing Diego Comin, Daniel Lashkari, and Marti Mestieri, he says over 75% of structural change comes from rising incomes changing what people want, not just from cheap automated goods freeing up spending.
Mimetic desire makes human-made goods unusually sticky in an AI world — Drawing on René Girard and experiments with Kristoff Matarash and Gyanan Mandal, Imas says willingness to pay roughly doubled under random exclusion, while AI involvement weakened exclusivity premiums: human-made art got a 44% boost from being one-of-one versus 21% for AI-generated work.
The durable jobs may be less ‘prompt engineer’ and more nurse, teacher, therapist, trainer, and guide — His bet is that long-run employment shifts toward care, education, hospitality, therapy, craftsmanship, and community-facing roles where the human is not just an input but part of the product itself.
The Breakdown
Starbucks, of all companies, brings the human back
Imas opens with a deceptively simple case: Starbucks, a $112 billion company selling one of the most standardized products imaginable, should have been an automation slam dunk. Instead, after tightening processes and reducing labor, it reversed direction — Brian Niccol highlighted handwritten notes, ceramic cups, better seating, and more hospitality as the things customers actually valued.
The big economic question: what stays scarce after AI?
His framing is clean: economics doesn’t disappear in abundance, it just revolves around a different bottleneck. If AI can produce commodities at near-zero marginal cost, then the whole game becomes identifying what people still can’t get enough of — and his answer is human-linked value, not just output.
From Marx’s commodity form to a post-commodity economy
He walks through the classic industrial move: detach the product from the person who made it, standardize production, and scale it globally. AI looks like the endpoint of that logic, but Imas argues it may also trigger its reversal in spending patterns, pushing demand toward goods and services whose value is inseparable from the human behind them.
Structural change is the real template, not job apocalypse
This is where the essay gets more concrete. Using the old agriculture-to-manufacturing-to-services story — 40% of Americans on farms in 1900, under 2% today — he argues that automation usually makes a sector smaller as a share of GDP, because rising real incomes push spending elsewhere; he leans on Comin, Lashkari, and Mestieri’s result that income effects explain more than 75% of this reallocation.
Rich people don’t just buy more — they buy more human
Imas says the key mistake is assuming preferences stay fixed. He points to 2022 BLS spending data and Yakim Hubmer’s work to argue that higher-income households disproportionately spend on labor-intensive, experience-heavy categories like dining, entertainment, education, and care — not just more stuff, but more goods where the human element matters.
René Girard, Armani suits, and why exclusivity keeps mattering
The philosophical center of the piece is mimetic desire: we want things partly because other people want them and can’t have them. Imas connects Girard to Augustine, Hobbes, Rousseau, and Dave Hickey’s Armani example, making the point that once basic needs are met, provenance, status, and social meaning become a bigger share of what people are actually paying for.
The experiments: exclusion raises value, AI lowers it
He backs the theory with his own research. In experiments with Kristoff Matarash, willingness to pay nearly doubled when people learned others would be randomly excluded from getting the same good; in new work with Gyanan Mandal, AI involvement made objects feel less exclusive and more reproducible, with human-made art receiving a 44% exclusivity premium versus just 21% for AI-made art.
What the future of work could look like
By the end, the picture is surprisingly specific and not especially sci-fi: teachers, nurses, therapists, childcare workers, hospitality staff, trainers, clergy, guides, bespoke makers, and “provenance certifiers” all become more central. His line is that the durable jobs won’t be about supervising AI forever, but about being the person whose judgment, warmth, memory, or presence is part of what the buyer is buying.