Goldman's 15 Million AI Jobs Figure Is a Slow Reshuffle, Not a Collapse

The short answer: the same Goldman Sachs economist behind the 15 million number expects almost all of those workers to be reabsorbed, and the people who come out ahead will be the ones who saw which of their tasks were exposed and moved first. In a July 2026 Goldman Sachs Exchanges podcast, economist Joseph Briggs projected that AI could displace around 15 million US workers, roughly 9% of the labor market, over the next decade. That number deserves attention. But in Goldman's own framing, a displaced worker is usually a reallocated worker, and the change lands gradually, task by task. That is exactly why you can get ahead of it.

What Goldman Sachs actually said

On the July 2, 2026 episode of Goldman Sachs Exchanges, Briggs, who co-leads the firm's global economics team, estimated that "around nine percent of all workers in the US will be reallocated," which works out to roughly "15 million workers leaving or being displaced from their positions" over a 10-year period. Even so, he was careful about the pace: the "unemployment rate increase in any given year would be less than one percentage point." The two figures are consistent: 15 million is a cumulative total over roughly ten years, while the unemployment effect is measured one year at a time. An earlier Goldman analysis put the displacement figure at 6% to 7% of workers over a similar transition (Goldman Sachs, 2026: How Will AI Affect the US Labor Market?). In other words, Goldman is describing a slow reshuffle spread across a full decade.

The context the headlines skip

Here is the sentence that almost never makes the headline. "We're not expecting that displaced workers will be displaced over the long run," Briggs said. "We do expect that ultimately there are going to be more than enough jobs created to reabsorb workers back into the labor market." He grounded that in how the US labor market already works: "every year, we see around 30 million jobs being created, now granted, 29 million are being destroyed." Churn on that scale is normal. And over the past eight decades, Briggs noted, "around 85% of job growth has been driven by the technological creation of new positions." The technologies that displaced workers also built the jobs that absorbed them.

But "net neutral" is not neutral for you

A national net gain of about one million jobs a year is cold comfort if you are one of the 29 million on the wrong side of the ledger. "Reabsorbed" is a statistic at the economy level and a hard year at the individual level. The people who move through a transition like this smoothly share one habit. They saw which parts of their work were being automated, and they repositioned toward the safer parts before their employer made the decision for them. The macro story is reassuring, and the 2026 data still shows no broad displacement yet. Your personal story depends on timing.

Displacement happens task by task

Notice Briggs's own word: workers will be "reallocated." You cannot reallocate yourself sensibly without knowing what you are reallocating away from. Jobs are not automated in one stroke; specific tasks are. A role loses three routine tasks to an AI agent, gains two oversight tasks, and quietly becomes a different job with the same title. That is what displacement usually looks like up close, and it is invisible if you only think in job titles. Think in tasks and the 9% stops being a lottery and becomes a checklist: which of my tasks are exposed, which are safe, and which new tasks should I move toward.

What to do before the tide comes in

Three moves follow directly from Briggs's own framing. First, map your tasks and mark the ones a model could plausibly do today, because those are your exposure. Second, deliberately build the tasks that sit next to AI rather than under it: the judging, guarding, and directing work that grows when automation spreads. Third, reposition early. In a decade-long transition, the earliest movers gain the most. Year one beats year nine. None of that requires predicting the future. All it takes is a frank inventory of your own week.

The scary framing vs. what Goldman's data shows

Frequently asked questions

Will AI really displace 15 million US jobs?

Yes, but over roughly ten years, and Briggs expects reabsorption. The pace matters: with about 30 million US jobs created and 29 million destroyed in a normal year, a 15 million shift over a decade fits inside routine churn.

Does "displaced" mean my job is gone for good?

No. Goldman expects new jobs to reabsorb displaced workers, as technology has driven about 85% of job growth over eight decades. A practical first step: list your weekly tasks and flag which ones AI can already handle.

Map your risk before the headline is about you

Goldman's numbers describe the whole economy. Your exposure is specific, and it comes down to your individual tasks rather than your job title. An AI Job Risk Score breaks your role into its real tasks, shows which are most exposed to AI, and gives you the task map to reposition while you still have a decade of runway instead of a two-week notice.