As artificial intelligence absorbs more and more of the economy's cognitive labor, one question sits underneath all the anxiety and excitement: what is left for humans to do? The answer, according to emerging economics research, is more precise than most people expect. What remains scarce after AI automates everything it can is humanness itself — specifically, goods and services where the fact that a human was involved is part of the value. Economists call this the relational sector, and understanding it may be the most important framework for thinking about your career, your skills, and humanity's place in an increasingly automated world.
What Is the Relational Sector?
The relational sector is defined as goods and services where the presence of a human in the process is itself part of the product's value — not just a means to an end.
This is a subtle but important distinction. For most of economic history, humans were involved in production because they were the best available tool. A horse pulled the plow; then a tractor replaced it. A telephone operator connected calls; then automated switching replaced her. In each case, the human was valued for the output, not for being human. When something better came along, the output was preserved or improved, and the human was no longer needed.
The relational sector is different. Here, the human is the output — or at minimum, an irreplaceable part of it. A therapist you've seen for three years isn't valuable merely because they give good advice. An AI might give equally good or better advice. The therapist is valuable because the trust, the history, the felt sense of being understood by another conscious person — these things are part of what you're paying for. Replace the therapist with an AI and you have a fundamentally different product, even if the advice is identical.
Alex Imas, Director of AGI Economics at Google DeepMind and Professor of Economics at the University of Chicago Booth School, introduced this framing in a recent essay and expanded on it in a conversation on the Dwarkesh Podcast. His core claim: because humans are naturally scarce and fixed in number, anything that requires a human as a human will retain scarcity value even as AI makes most other things abundant.
The Economics of Scarcity Have Always Worked This Way
When something becomes abundant, its price falls. When it remains scarce, its price holds or rises. If AI can produce software, legal analysis, financial models, and creative content at near-zero marginal cost, the price of all of those things — as commodities — will trend toward zero. What won't trend toward zero is anything that cannot be replicated or substituted.
Humans cannot be replicated. There are approximately 8 billion of us, and that number isn't growing fast enough to matter relative to the pace of AI capability gains. Every human you will ever interact with is, by definition, rare.
David Ricardo grappled with a version of this problem in 1820, when the Industrial Revolution was automating the jobs of his era. He predicted mass unemployment. He was wrong — not because automation didn't happen (it did), but because he failed to anticipate that abundance in one area generates new demand in others. What he got right was the underlying mechanism: what becomes scarce commands value. He just misjudged what would remain scarce.
Phil Trammell, Head of Economics at Epoch AI and Research Scholar at Stanford, offers a useful analogy. Imagine a Mongolian economist in 1400 trying to predict what would be valuable in a highly automated future. He might reason: "We'll have machines that replace horses for transportation, machines to make food and shelter. What will be left? Singers — humans performing for other humans." He wouldn't be entirely wrong; live performance does remain valuable. But he would dramatically underestimate how much the variety of non-human goods would expand, absorbing demand and keeping the singer's share of the economy small.
The lesson isn't that relational goods don't matter. It's that predicting the size of the relational sector in advance is genuinely hard.
This Changes How You Should Think About Your Career
The question to ask about any role or skill is not "Can AI do this?" — in the limit, AI will be able to do nearly everything. The better question is: Does it matter to the person receiving this service that a human did it?
The key insight Imas adds is that this isn't binary by job — it's binary by task. Most jobs are bundles of tasks. A doctor fills out insurance forms (not relational), manages pharmaceutical relationships (not relational), and holds a patient's hand through a frightening diagnosis (relational). AI will likely absorb the first two. The third is where irreplaceable human value lives. The doctors who thrive won't necessarily be the most technically brilliant — they'll be the ones who excel at the human tasks that remain.
This pattern shows up across industries:
- Education: AI will be a remarkable tutor for factual content. But the teacher who sees a struggling student and knows — from genuine care and years of human pattern recognition — that something is wrong at home, and acts on it, is doing something AI cannot replicate.
- Finance: Algorithmic advice will likely outperform most human advisors on pure returns. But the financial advisor who builds a twenty-year relationship with a family, understands what money means to them emotionally, and calls when they're panicking during a market crash — that's relational work.
- Leadership: Managing, motivating, and developing other humans is deeply relational. The trust that makes an organization function is built between people.
- Healthcare: Diagnosis and treatment planning may be AI-dominated within a decade. The care, the presence, the advocacy — these will remain human.
This doesn't mean the answer is simply "go be a therapist." Relational value exists in almost every field. The work is figuring out where it lives in your field, and deliberately building toward it.
The Honest Uncertainty
It would be dishonest to present this framework as a complete map.
Imas is direct about this: we don't yet have the data to know how large the relational sector actually is, or how durable human preferences for human-involvement will prove to be. His lab ran experiments with art prints — asking people to pay real money for prints made by humans versus AI. The human premium was real, but sensitive to context. When the same print was produced in a run of 500 instead of one, the premium collapsed. The humanness mattered less when it was no longer a form of genuine connection with a specific person.
Will people eventually habituate to AI therapists, AI teachers, AI companions? The honest answer is we don't know. Some researchers argue that the preference for human connection is deeply evolutionary — that people who genuinely prefer AI relationships will, over generations, be selected against. Others argue that preferences are more plastic than we assume. What we can say is that the human premium is real, measurable, and meaningful today. And building skills in the relational dimensions of your work is, at minimum, a durable hedge against an uncertain future.
What This Has to Do with Appreciating AI
AI Appreciation Day exists to foster a more intentional, honest relationship between humanity and artificial intelligence. Part of that intention is seeing AI clearly — understanding what it is, what it can do, and what it cannot.
The relational sector framework is a gift in that sense. It doesn't ask you to fear AI or to dismiss it. It asks you to think carefully about what humans bring to the world that is genuinely valuable — and to recognize that the answer isn't "everything" or "nothing." It's something more interesting: humanness itself, in the right contexts, at the right moments.
This should be clarifying rather than frightening. The path forward for most people isn't to compete with AI on its own terms — faster, more accurate, cheaper — but to deepen what is irreducibly human about the work they do. To build relationships, not just outputs. To bring genuine care, not just competence.
AI will handle the parts of your job that don't require you to be human, which will free you to focus on the parts that do.
That seems like a future worth appreciating.
Frequently Asked Questions
What is the relational sector in economics?
The relational sector refers to goods and services where the fact that a human was involved in producing or delivering them is itself part of the value. Unlike most economic production — where humans are valued for their output and can be replaced by anything that produces the same output — relational goods and services are ones where the humanness is the point. A therapist, a live performance, or a longtime mentor are examples. The term was developed by Alex Imas, Director of AGI Economics at Google DeepMind, to describe what remains economically valuable as AI automates other tasks.
Will AI replace human jobs entirely?
The historical evidence suggests no. Every major wave of automation since the Industrial Revolution displaced specific categories of work while generating new demand elsewhere. In 2026, prime-age employment in the US is at its highest level since 2000, even as AI has automated substantial portions of knowledge work. What changes is the composition of work — jobs involving relational tasks like trust, empathy, and long-term human relationships will be the last to be automated, if ever.
How do I know if my job is in the relational sector?
Ask one question: does the person receiving this service care that a human did it, beyond the quality of the output? If yes for any part of your role, that part is relational. Most jobs contain a mix of relational and non-relational tasks. Identifying which of your tasks are relational — and deliberately strengthening those — is one of the most practical career strategies in an AI economy.
What skills are most valuable in a world with advanced AI?
Skills rooted in human connection are becoming more valuable as AI absorbs technical and analytical tasks: deep listening, trust-building, judgment in emotionally complex situations, and understanding what people need beyond what they explicitly ask for. Communication, leadership, mentorship, and long-term relationship management are all areas where human value is likely to endure.
Does appreciating AI mean ignoring the risks of AI?
No. Appreciation is not the same as uncritical acceptance. It means understanding AI clearly — its genuine capabilities, its limitations, and the ways it changes the economy and society. That clarity is what makes thoughtful adaptation possible, rather than reacting from fear or denial.
Is the human preference for human interaction just a temporary bias that AI will eventually eliminate?
Possibly, but not certainly. Research suggests the preference for human connection has deep evolutionary roots. Alex Imas's experiments show people pay a measurable premium for human-made goods and services across multiple contexts. Whether that premium persists as AI becomes more sophisticated is one of the genuinely open questions in economics — which is exactly why building skills around human connection is a sound strategy under most plausible futures.