“Management skills are real and measurable, and they are seriously underpriced by the labor market.” - David Deming
A good store manager at one South American retail chain was worth between $180,000 and $240,000 a year in added profit - and was paid a small fraction of it. In the final part of his series on work in the age of superintelligent AI, David Deming - labor economist and Dean of Harvard College - explains where the money in social skills comes from. Specialization made the modern economy rich, but its gains depend on coordination, and coordination is expensive: meetings, handoffs, misunderstandings, work that melts away between people. Social skills reduce the melt. Along the way: Adam Smith’s pin factory, the 19th-century trade that shipped Massachusetts pond ice to India, why the people who most want to be managers tend to be worse at the job, why the skill of managing humans is nearly identical to the skill of managing AI agents, and why the person who can make the sale may end up the best-paid link in an AI economy.
Listen on YouTube, Spotify, and Apple Podcasts
This week concludes my three-part series on the future of human work in an age of superintelligent AI. Part One used comparative advantage to explain why AI being better than us at everything won’t necessarily put the human race out of work. Part Two staked my claim that social skills are the human advantage most likely to endure, because we evolved to learn about other people in ways that brute-force pattern recognition can’t replicate. This week I answer the question sitting underneath both arguments: why are social skills valuable in the first place? You understand people. You can read a room. So what? Why would anyone pay you for that?
The episode ends with the million-dollar question: even if humans keep working, will the work pay? The honest answer is that nobody knows, including me. But the bleakest forecasts depend on assumptions about how human and AI labor combine in production, and those assumptions may not hold. If most of a production process gets automated and one human bottleneck remains, that bottleneck becomes more valuable, not less. If the bottleneck is trust, judgment, and making the sale, the people who do those things best will be massively productive precisely because AI is leveraging them up.
One thing you can do while you listen: at the Skills Lab we built the Assignment game, a test of the economic decision-making skill that predicted managerial success in the studies I discuss in this episode. You can take it at here and find out whether you would have been one of the good managers.
As always, let me know what you think at david@thecontextwindow.com.
Chapters
[00:00] Cold open - You understand people. You can read a room. You’ve got, as David’s students would say, some rizz. So what? Why would anyone pay you for that?
[00:21] The question beneath the series - Part one: comparative advantage keeps humans working. Part two: our edge is social learning. Part three: why any of it is worth money.
[01:23] Goldin and Katz, hanging in the Louvre - Back to the task-trade model from part one, where collaboration yields six papers a year instead of four.
[02:57] Adam Smith’s pin factory - Smith’s famous example of how division of labor transforms productivity.
[03:50] “Wheels are my love language” - Why specialists only emerge where the market is big enough, from carpenters to cardiologists.
[05:15] Why isn’t the economy one big team? - Becker and Murphy’s answer: coordination costs limit specialization.
[06:50] Iceberg trade costs - How trade economists model the value that gets lost in transit.
[07:47] The great American ice trade - The 19th-century export business that made the metaphor literal.
[08:21] When collaboration stops paying - At what point do coordination costs outweigh the gains from working together?
[09:29] Social skills mean less melt - The core idea of the episode: social skills lower the cost of working with other people.
[10:29] Get in where you fit in - Why good teammates adapt their role to the team.
[12:07] The same logic runs up the hierarchy - What makes someone a good manager.
[12:52] The Assignment Game - The Skills Lab’s test of economic decision-making, which you can take at skillslab.dev.
[14:25] Managers matter, and the volunteers are worse - What the lab experiment showed about who actually manages well.
[15:23] From the lab to the shop floor - The test predicted real promotions, and store performance at a South American retail chain.
[16:48] Seriously underpriced - A good manager is worth $180,000 to $240,000 a year and is paid a fraction of it, and Virginia Minni’s evidence on managers as matchmakers.
[17:25] Managing humans, managing AI - New research finds the two skills are nearly the same.
[18:29] Good leaders ask more questions - What hidden-profile tasks reveal about how good managers lead.
[19:19] Trust is the long game - Why social skills matter even more in long-term relationships and incomplete contracts.
[21:16] The million-dollar question - Employment is one thing; wages are another.
[21:46] Could wages fall below subsistence? - The assumption behind the bleakest forecasts.
[23:05] The O-ring theory - Michael Kremer, the Challenger disaster, and production processes where a single mistake ruins everything.
[24:14] When automation raises wages - Gans and Goldfarb on why partial automation can make human work more valuable.
[24:43] You make the sale - Why relational judgment may become the highest-value input in the chain.
[25:15] The caveat - What happens if AI gets good at social interaction, too.
[25:40] Next week - Tino Cuéllar joins David to talk about law, immigration, AI, and governance in higher education.
Mentioned in this Episode
Papers
Gary S. Becker and Kevin M. Murphy, “The Division of Labor, Coordination Costs, and Knowledge“ (The Quarterly Journal of Economics, 1992) - specialization is limited by the cost of coordinating it, not just by the size of the market.
David J. Deming, “The Growing Importance of Social Skills in the Labor Market“ (The Quarterly Journal of Economics, 2017) - the task-trade model of social skills as lower coordination costs, minus the diagrams and Greek letters.
Ben Weidmann, Joseph Vecci, Farah Said, David J. Deming, and Sonia R. Bhalotra, “How Do You Identify a Good Manager?“ (The Quarterly Journal of Economics, 2026) - the assignment game, the lab experiment, the overconfident self-nominators, and the retail-chain field study behind the $180,000–$240,000 figure.
Virginia Minni, “Making the Invisible Hand Visible: Managers and the Allocation of Workers to Jobs“ (The Quarterly Journal of Economics, 2026) - good managers move the right workers to the right jobs, raising firm productivity and workers’ careers.
Ben Weidmann, Yixian Xu, and David J. Deming, “Measuring Human Leadership Skills with AI Agents“ (NBER Working Paper 33662, 2025) - the 0.81 correlation between managing humans and managing AI agents.
Yiqun Chen, “Team-Specific Human Capital and Team Performance: Evidence from Doctors“ (American Economic Review, 2021) - shared work experience between the doctors who perform heart procedures and those who provide follow-up care reduces patient mortality.
George P. Baker, Robert Gibbons, and Kevin J. Murphy, “Relational Contracts and the Theory of the Firm“ (The Quarterly Journal of Economics, 2002) - agreements that hold not because the terms are complete but because both parties value the option to transact again.
Martin Brown, Armin Falk, and Ernst Fehr, “Relational Contracts and the Nature of Market Interactions“ (Econometrica, 2004) - experimental evidence on trust and repeated relationships in markets.
Michael Kremer, “The O-Ring Theory of Economic Development“ (The Quarterly Journal of Economics, 1993) - production processes where one mistake causes catastrophic failure, named for the Challenger disaster.
Joshua S. Gans and Avi Goldfarb, “O-Ring Automation“ (NBER Working Paper 34639, 2026) - why partial automation can raise, rather than lower, human labor income.
Articles
Matthew Barnett, “AGI Could Drive Wages Below Subsistence Level“ (Epoch AI, Gradient Updates, 2025) - the viral post arguing the bleak case for human wages.
Books
Adam Smith, An Inquiry into the Nature and Causes of the Wealth of Nations (1776) - the pin factory, and the division of labor limited by the extent of the market.
Gavin Weightman, The Frozen Water Trade (Hyperion, 2003) - how New England pond ice became a global export.
Try It Yourself
The Assignment Game, Skills Lab’s test of economic decision-making skill
Further Reading and Listening
Part One of this series: “Superintelligent AI Won’t End Human Work. Here’s the Math.“
Part Two: “Why Social Skills Will Be Increasingly Valuable in the Age of AI.“
Credits
Host: David Deming, Danoff Dean of Harvard College
Executive Producer: Denise Koller Consulting Producers: Tim Smith and Jonathan Palumbo
