The Starbucks Problem: Why AI Might Not Kill Jobs

Here's something wild: Starbucks—a company with a $112 billion market cap that sells one of the most automatable products on Earth—recently did something unexpected. After years of pushing automation to squeeze margins, they reversed course. They're hiring more baristas, not fewer. Ceramic mugs are back. Handwritten names on cups matter again.

CEO Brian Niccol said the quiet part out loud: "Small details in hospitality drive satisfaction." Customers wanted to sit and stay when humans were visibly involved in making their $7 latte.

This is the opening salvo in economist Alex Imas's viral essay on what he calls the "post-commodity economy"—and it's one of the more interesting frameworks I've seen for thinking about AI's economic impact that doesn't just gesture vaguely at "new jobs we can't imagine yet."

Imas, who told Fortune his first reaction to AI was "to be very scared," isn't coming from a pro-AI ideology. He's trying to work through the logic of what actually becomes scarce when machines can make almost anything.

The Doomer Framing Is Missing Something

The standard AI jobs narrative goes like this: AI automates tasks → humans become redundant → mass unemployment → societal collapse (or UBI, depending on who's talking). It's the same mental model we apply to every automation wave, just scaled up to nightmare proportions.

But Imas argues this framing assumes the economy is static—that there's a fixed set of goods and services, and if machines can produce them cheaper, humans are simply out of luck. History suggests otherwise.

When agriculture got automated, 40% of American workers didn't just become permanently unemployed. They moved to manufacturing. When manufacturing automated, they moved to services. Not because anyone planned it, but because as productivity rose and prices fell in automated sectors, people got richer and started wanting different things.

The economic term for this is "structural change," and it's driven by something called non-homothetic preferences—a deeply unsexy phrase that means: when you get richer, you don't just buy more of the same stuff. You buy fundamentally different stuff.

You can only eat so much food (agriculture's problem). You can only use so many identical phones (manufacturing's eventual problem). But there's always a better restaurant experience, a more attentive doctor, a more meaningful educational relationship. These sectors have high "income elasticity"—demand grows faster than income.

What Actually Becomes Scarce

Imas's core insight is that AI doesn't eliminate scarcity. It relocates it.

Before industrialization, products were inseparable from the people who made them. You knew your baker, your weaver. Their reputation was tied to the bread, the cloth. Capitalism's big trick was the "commodity form"—making products valuable independent of who made them. A table is a table. An iPhone is an iPhone. Doesn't matter if it was assembled in Shenzhen or Chennai.

This was Marx's whole alienation critique (exploitation through interchangeability), but it was also an engine of extraordinary wealth. Breaking craft into standardized steps meant you could scale, optimize, ship across oceans.

AI takes this logic to its conclusion: if any human can be replaced in commodity production, then eventually all humans can be. Labor becomes entirely substitutable with capital.

But here's where it gets interesting. Imas points to research showing that rich households already spend disproportionately on goods and services where the human element matters. In 2022 data, the highest income quintile spent 4.3x more than the lowest quintile overall—but way more than that on in-person dining, entertainment, education. Not just "more restaurant meals" but specifically meals where the experience, the hospitality, the relationship is part of what you're buying.

He calls this the "relational sector"—the parts of the economy where human involvement isn't a bug to be automated away, but the entire point.

The Pattern Is Already Visible

The wealthy aren't just buying more stuff. They're buying providence, craft, exclusivity, taste, care. Things that are explicitly not commodities. A handmade ceramic mug from a specific artisan. A meal prepared by a chef whose judgement you trust. Medical care from a doctor who knows your history and actually listens.

Imas writes: "As people get richer, they don't just want more commodities. They want things that aren't commodities in the standard sense of the word. The social aspects of products such as the relationships, the status, and exclusivity."

This isn't even a new observation. Economist Joachim Hubmer documented that higher-income households spend relatively more on labor-intensive goods and services as a share of total consumption. The pattern exists now, pre-AGI.

If AI radically drops the cost of commodity production—making everyone effectively richer in real terms—this logic predicts a massive sectoral shift. Automated sectors shrink as a share of GDP (even as they produce more). Relational sectors grow.

The Catches (Because There Are Always Catches)

Imas is careful about what he's not claiming. He's not saying labor's share of the economy must stay constant, or even that it won't decline. Automation might still reduce labor's aggregate share of GDP. His narrower claim is about sectoral reallocation: even if labor's share falls, the relational sector could remain substantial because of how human preferences work.

There's also a huge caveat about geography. This framework works best for developed economies where rising incomes can fund the transition to relational consumption. For developing countries whose economies were built on producing commodities for rich nations, the picture is "more complicated and potentially more worrying."

And there's the question of whether this is actually good. A post-commodity economy where human labor shifts to "relational" work could mean meaningful jobs centered on care, craft, and connection. Or it could mean a massive servant economy where the wealthy pay humans to perform inefficiency as a luxury good. The framework doesn't determine the outcome—policy does.

What This Changes

If Imas is right, the AI jobs debate is asking the wrong question. It's not "how do we save jobs from automation?" It's "what do people want to consume more of when basic production becomes cheap?"

That's a wildly different problem with wildly different solutions. It suggests we should be thinking less about preserving existing jobs and more about building infrastructure for the sectors where demand will grow—healthcare, education, hospitality, craft, care work, cultural production.

It also suggests the transition might be less about finding "new jobs we can't imagine yet" and more about expanding capacity in jobs we already know matter but currently underprovide because they're too expensive. More teachers with smaller classes. More nurses with more time per patient. More artisans who can actually make a living.

Whether we get the good version of this transition or the dystopian one depends entirely on choices we make now about how income gets distributed, what work gets valued, and who gets to participate in the post-commodity economy.

The Starbucks baristas coming back to write names on cups? That might not be a weird corporate quirk. It might be the leading edge of something much larger.

— Yuki Okonkwo