My fellow pro-growth/progress/abundance Up Wingers in the USA and around the world:

Perhaps the biggest source of US economic uncertainty this year is the job market. The trend isn’t great. Monthly employment growth has fallen from about 175,000 to near zero, notes Moody’s Analytics, driven by tight labor supply from immigration restrictions and weaker labor demand from government budget cuts, tariffs, and business uncertainty.

What’s more, unemployment is creeping higher, and upcoming BLS data revisions are likely to show job growth was overstated, suggesting the labor market may be even weaker than it currently appears, the firm’s chief economist, Mark Zandi adds.

And as the bank Goldman Sachs sees it, “the most difficult question about the labor market is to what degree companies will replace workers with artificial intelligence, or at least hold off on hiring.”

That the GS-identified question is even a “question” would likely surprise AI doomers and many like-thinkers in the media. Their frequent assumption is that considerable dislocation is already happening. Yet the data so far suggest the near-term impact of AI looks less like an ongoing — or even looming — worker bloodbath and more like quieter, slower disruption at the job-market margins.

For three years, of course, the launch of ChatGPT has been cast as the opening act in a white-collar reckoning. If that negative narrative were correct, the labor market should be flashing numerous warning lights, including collapsing wages and the welling ranks of unemployed professionals.

But not so much, it turns out. Employment and pay in much of white-collar America continue to rise. Routine clerical roles are shrinking, but hybrid technical-and-judgment jobs are expanding. So far at least, AI appears to be reconfiguring tasks more than erasing entire occupations.

By the numbers, lots of numbers

Consider what the data compiled by The Economist show, via an excellent new piece:

• Since late 2022, the US has added roughly 3 million white-collar jobs, spanning management, professional, sales, and office roles.
• Across more than 100 large white-collar occupations, employment has risen by about 4 percent while real wages have increased by roughly 3 percent.
• Several professions widely portrayed as early AI casualties are still expanding, including software developers (+7 percent), radiologists (+10 percent), and paralegals (+21 percent) over the past three years.
• Real wages in professional and business services have climbed by about 5 percent since late 2022, while office and administrative workers are earning roughly 9 percent more.
• Roles that combine technical expertise, judgement, and coordination are growing fastest, with project managers and information-security specialists up roughly 30 percent and “business operations specialists, all other” up nearly 60 percent since late 2022.
• Emerging technical categories are also expanding quickly, as “other mathematical-science occupations” have grown by roughly 40 percent, accompanied by around 20 percent real wage growth.
• Where job losses do exist, they remain narrow and task-specific, with employment down about 13 percent for insurance-claims clerks and down about 20 percent for secretaries and administrative assistants.

A job rejiggering, not recession

I wouldn’t think that this is what a labor-market collapse looks like. If anything, it resembles the start of a grand re-sorting during a technological inflection point — rewarding roles that combine technical skill, judgment and coordination, while trimming more routine clerical work.

Where hiring has cooled — such as in computer science — there’s new evidence the slowdown began in early 2022, months before ChatGPT launched, aligning more closely with tighter monetary policy and the post-pandemic tech correction than with an AI shock.

Likewise, the Dallas Federal Reserve finds “little indication of widespread labor-market disruption due to AI.”

Part of the explanation is that AI adoption remains early days. Evidence from Anthropic suggests only about 4 percent of occupations use AI for three-quarters or more of their tasks, and almost no jobs can be fully automated. GS reckons that while AI has the potential to eventually automate a large share of job tasks, that works out to displacing just 6-7 percent of current jobs.¹

Of course, as more companies adopt generative AI tools and figure out how to best use them, the disruption and revaluation of skills² will increase (as will the productivity impact). GS predicts:

AI adoption increased last year, especially in areas where work tasks are more exposed to AI automation, and many companies appear eager to implement AI to reduce labor costs. As a result, we expect to see a more meaningful decline in employment in these industries of at least 20k per month in 2026, though this should be partly offset by AI-related job gains in other areas.

Goldman’s “most difficult question” may turn out to have a reassuring answer: AI is mostly changing what workers do, not whether they work at all.

1 The disruption that does exist is showing up less in layoffs than at the hiring gate. Employment among 20-to-24-year-olds in highly AI-exposed roles has declined modestly since late 2022, while employment among prime-age workers in the same roles has remained largely stable. Firms appear to be raising hiring standards, preferring candidates who can contribute immediately. Entry is getting harder, while exit remains rare.

2 Reminder: Technology automates some tasks, complements others, and creates entirely new tasks and jobs. Too often we focus mostly on the first effect.