The Six Myths Executives Tell Themselves About AI and Jobs
TL;DR
Clara Shih’s six “myths” are really a warning against passivity — the former Salesforce AI CEO argues executives and policymakers are using comforting stories about AI and jobs to avoid building training programs, safety nets, and policy now.
The labor impact is already showing up in the data, especially for younger workers — the hosts cite Stanford research saying AI remains a statistically significant driver of cuts after controlling for macro factors, and Shih points to 2.3 million underemployed recent grads being missed by headline unemployment numbers.
“AI creates more jobs” is not the whole story if supply outruns demand — Shih pushes back on the Jevons paradox narrative with the example of London black cab drivers, whose real income fell 50% after GPS and Uber even as ride demand soared.
The ‘everyone can just become entrepreneurs or go into the trades’ fallback sounds neat but breaks under numbers and reality — the podcast calls entrepreneurship “really freaking hard,” and Shih notes BLS projects only about 38,000 net new trade jobs per year versus 2.3 million underemployed grads.
Public sentiment is shifting fast, but people still think the disruption will hit someone else — the hosts cite a March poll showing 70% of Americans expect fewer job opportunities from AI, while their own survey of nearly 2,100 respondents found 71% think more jobs will be eliminated than created over the next three years, yet only 21% worry about their own jobs.
The most plausible scenario may be neither job apocalypse nor total abundance, but partial displacement we’re least prepared for — they highlight Ezra Klein’s argument that an AI shock affecting 8 million workers could be harder politically than one affecting 80 million because broad crises trigger action while targeted pain gets ignored.
Summary
Economists laughed, and now Clara Shih is forcing the conversation
The segment opens with a jab at economists: the host says he’s spent six years being “literally laughed at” for suggesting AI would affect jobs. That sets up Clara Shih — founder of the New Work Foundation and former Salesforce AI CEO — as a credible, not-at-all anti-AI voice arguing that six myths about AI and jobs are giving leaders permission to wait instead of act.
Myth by myth: the comforting stories executives keep telling themselves
They quickly run through Shih’s six myths. AI layoffs are not just a cheap-money hangover, she says; Stanford research from last fall found AI still driving statistically significant cuts, especially in entry-level roles, even after controlling for macro factors. She also rejects the Jevons paradox comfort blanket, arguing that supply can grow faster than demand and crush wages — like London black cab drivers, whose real incomes fell 50% after GPS and Uber turned their expertise into a commodity.
Why AGI debates and unemployment headlines can distract from what’s already happening
Shih’s third myth is that arguing over AGI timelines has become a substitute for the harder conversation about what to build now. The fourth is that headline unemployment numbers miss the deeper damage: she points to 2.3 million underemployed recent grads and warns that this kind of early-career scarring can ripple into depressed lifetime earnings, delayed household formation, and weaker consumption.
“Just send them to trade school” and other too-easy answers
The hosts linger on myth five because it’s one of the most common executive fallback lines. Shih notes the Bureau of Labor Statistics projects only about 38,000 net new trade jobs per year nationally, which is nowhere close to absorbing 2.3 million underemployed grads, and the host adds that the other favorite answer — “everyone can be entrepreneurs” — ignores how brutally hard entrepreneurship actually is.
The PR push for abundance is ramping up
Paul says it feels like tech leaders have stepped up their messaging lately, selling a story that everything will be fine and abundance is around the corner while glossing over job losses. He points to a widely shared Jensen Huang interview making the bullish case that AI will be “amazing for jobs,” then contrasts that with his own effort to keep searching for serious counterarguments because, as he puts it, “I want to be wrong.”
The polling says people are worried — just not about themselves
They bring in Ezra Klein’s New York Times piece, which cited a March poll showing 70% of Americans think AI will mean fewer job opportunities, up from 56% a year earlier, and 30% worry about their own jobs. Then comes their teaser for upcoming State of AI for Business research: among almost 2,100 respondents, 71% said more jobs will be eliminated than created over the next three years, only 13% said more will be created, and yet just 21% were concerned about their own role — classic “it’ll happen, just not to me.”
The real danger may be a medium-sized shock no one treats like an emergency
The hosts walk through Ezra Klein’s more measured position: economists remain skeptical of mass joblessness, and University of Chicago economist Alex Mas argues the real question is what stays scarce and what humans will still want humans to do. But the line that lands hardest is Klein’s warning that a world where AI displaces 8 million workers might actually be harder to handle than one where it displaces 80 million, because total crisis forces systemic action while partial pain gets treated like somebody else’s problem — much like COVID only triggered urgency once everyone felt it.
Ending in the messy middle, not certainty
The close is notably not apocalyptic. They say the truth is probably somewhere between “glorious future” and total collapse: maybe tech leaders are right, maybe they’re not, but at least the discussion is broader and happening weekly now instead of barely at all six or twelve months ago.
Was This Useful?
Share
Keep Reading
Make Alcreon Yours
Tune your feedFive quick questions, and the feed ranks what matters to you first.Or just get notified
The weekly Echo. Signal worth keeping in your inbox.
Every new piece, announced on X.
Read Next
See all
Playbook
The Retirement Email Isn't a Warning
Model retirements now arrive every few weeks; the config-eval-rehearsal loop turns each deprecation email from a fire drill into an afternoon swap.
Playbook
The Cheapest Model That Passes
OpenRouter lists 400 models behind one API. The fix for choosing isn't a better leaderboard, it's a four-step protocol that ends in a real eval.
Playbook
Cheap Models, Hard Tasks
Most agent workflows route every step to the frontier model by default. The bill scales with how chatty the agent gets, even when most steps don't need that brain.