The 30-Year Trend That Predicted AI
David Deming at Harvard published one of the most important labor market studies of the century — and almost nobody outside of economics has heard of it.
Deming tracked employment and wage trends across occupations from 1980 to 2012 and found a striking pattern: jobs requiring high social skills grew by 12 percentage points as a share of the U.S. labor force. Jobs that were math-intensive but socially isolated shrank by 3.3 percentage points.
But here's the really interesting part: the highest employment and wage growth occurred in occupations requiring both high cognitive skills AND high social skills. Not one or the other. Both.
In other words, the labor market has been telling us for 30 years that the future belongs to people who can think AND connect. People who can analyze AND persuade. People who can strategize AND build trust.
AI accelerates this trend exponentially. Because AI is essentially infinite "thinking" capacity at near-zero cost. The thinking part is becoming cheap. Which means the connecting, feeling, trusting, caring part becomes the scarce — and therefore valuable — resource.
The "Feeling Economy" Framework
Roland Rust and Ming-Hui Huang at the University of Maryland coined a term for this: the Feeling Economy.
Their framework is elegantly simple. Economic activity, they argue, requires three types of human capability:
Physical (doing). This was largely automated by industrial machines in the 19th and 20th centuries.
Thinking (analyzing, computing, processing). This is being automated by AI right now.
Feeling (empathizing, connecting, caring, intuiting). This is what's left. And it's far more valuable than we've given it credit for.
As AI handles more of the "thinking" work, humans increasingly handle the "feeling" work: genuine empathy, emotional recognition, nuanced communication, authentic care, trust-building, and the ability to navigate the messy, irrational, beautiful complexity of human relationships.
This isn't a soft, hand-wavy prediction. It's an economic inevitability. When thinking capacity is abundant and cheap, feeling capacity becomes scarce and expensive. Supply and demand.
What "Feeling" Skills Actually Are
Let me be specific, because "soft skills" is a garbage term that means everything and nothing. Here are the specific capabilities that are becoming premium:
Emotional intelligence in context. Not just "being empathetic" in the abstract, but reading the room in a specific meeting, sensing when a client is losing confidence, knowing that the silence from your colleague means something different than you think. AI can analyze sentiment in text. Humans can read the unspoken in a room.
Trust-building over time. This is irreducibly human. Trust is built through consistency, vulnerability, follow-through, and shared experience. An AI can be reliable, but it can't earn trust in the way that a human who's shown up for you through three crises can.
Navigating ambiguity together. When the answer isn't clear — and in leadership, the answer is almost never clear — the ability to sit in uncertainty with other people, hold space for different perspectives, and move toward a decision together is a profoundly human skill.
Persuasion and influence. Not manipulation — genuine persuasion, where you understand what someone values, frame your argument in terms that resonate with those values, and move them toward a position that serves everyone. AI can generate arguments. Humans deliver them with credibility.
Conflict resolution. When two smart people disagree about the right path forward, the person who can help them find common ground without either feeling diminished is worth their weight in gold. This requires understanding egos, histories, fears, and aspirations — all human territory.
Creative collaboration. The best creative work almost always emerges from groups of people bouncing ideas off each other, building on each other's half-formed thoughts, and creating something none of them could have created alone. AI can contribute to this process, but the chemistry between human collaborators is where magic happens.
The Measurement Problem (And Why It's Actually Good News)
One reason these skills have been undervalued is that they're hard to measure. You can test someone's coding ability. You can assess their financial modeling skills. But how do you quantify "reads the room well" or "builds trust with difficult stakeholders"?
This measurement difficulty has historically worked against feeling-skill professionals. In a world that rewards what it can count, uncountable skills get discounted.
But here's the shift: AI is making thinking-skills increasingly measurable and automatable, which makes them less valuable as differentiators. As the measurable skills become commodity, the unmeasurable ones become premium.
The fact that feeling skills are hard to measure is actually their moat. It means AI can't easily replicate them, employers can't easily commodity them, and the people who have them command increasing leverage.
What This Means for Career Strategy
If you're naturally gifted at interpersonal skills — if you're the person people come to when they need to talk, if you can walk into a room and sense the dynamics, if you build relationships easily — you are sitting on a career goldmine.
The traditional career advice for "people people" has always been slightly condescending: "You'd be great in HR!" or "Have you considered sales?" As if interpersonal mastery is a consolation prize for not being analytical enough.
That's changing. In the AI age, the ability to build trust, read rooms, navigate conflict, and inspire action is the most valuable set of capabilities an organization can deploy. Because everything else — the analysis, the data processing, the report writing, the modeling — can be augmented or automated. But the human connection part? That requires a human.
Don't apologize for being a people person. Invest in it. Deepen it. Treat it as the strategic asset it is.
And if you're naturally more analytical than interpersonal? Start investing in the feeling side. Not because analysis doesn't matter — it does — but because the premium is shifting. The analyst who can also build client trust will outperform the analyst who can only build models. Every time.
The future of work isn't human versus machine.
It's human feeling plus machine thinking.
And the feeling part? That's all you.