Goldman Sachs put a number on a fear a lot of people already had. Artificial intelligence could eventually displace around 9 percent of US workers, roughly one in eleven, and the bank’s economists added a sharper point. The biggest wave of those job losses would most likely come during the next recession, because companies have historically saved their automation cuts for downturns. That warning spread fast, and it collapsed two separate questions into one. Whether AI will disrupt the labor market, and whether AI will cause a recession. Those are not the same question, and keeping them apart is the only way to read this clearly.
What Goldman actually said
The core finding is about timing, not a prediction of imminent collapse. Goldman’s economists looked at how past waves of automation played out and found the labor disruption tends to be cyclical. Companies automate and shed workers when the economy slows, not while it’s growing, because a downturn gives them both the pressure and the cover to cut. So their conclusion was that the full labor-market impact of AI might not show up until a recession hits, at which point AI could deepen the job losses a normal downturn already produces.
Two things in the report cut against the panic reading. AI-attributed layoffs have actually fallen sharply from their peak earlier in 2026, not risen. And Goldman expects most displaced workers to eventually find other jobs, with the net rise in unemployment staying modest over time. This is a warning about how the next recession might play out, not a claim that one is here or that AI is about to trigger it.
The order of operations matters
Here’s the distinction that the headlines blur. Goldman’s thesis is that a recession would accelerate AI layoffs. That is the recession causing the layoffs, not the layoffs causing the recession. The job cuts, in this framing, are a consequence of a downturn that started for other reasons, amplified by companies reaching for automation once times get hard.
That ordering is the whole point. If AI layoffs mostly arrive during a recession, then they are a lagging feature of the cycle, something that shows up after the economy has already turned, not a leading cause you could use to see the turn coming. Which puts AI layoffs in the same category as ordinary layoffs. By the time they’re happening at scale, the recession is already underway.
Why layoffs lag and don’t lead
This connects to something true of employment in general, AI or not. Layoffs are one of the last things to happen in a slowdown, not one of the first. Cutting staff is expensive, disruptive, and hard to reverse, so companies try almost everything else first. They pause hiring, trim hours, cancel projects, and delay investment long before they start cutting people. By the time mass layoffs arrive, the weakness that caused them has been building for months in places that moved earlier.
So reading the labor market, including AI-driven cuts, as your early warning for a recession gets the timing backwards. Employment confirms a downturn. It doesn’t forecast one. The forecast lives in the signals that move before companies start cutting, the yield curve, credit spreads, business investment, and the direction of the market. Those bend first. Jobs break last.
Could AI still cause a recession?
It’s a fair question, and the honest answer is that the more plausible AI-driven risk isn’t layoffs at all. It’s the spending. Enormous amounts of capital are flowing into AI, data centers, chips, infrastructure, and that investment is currently a real engine of economic growth. The risk worth watching is not that AI quietly replaces workers. It’s that the AI investment boom slows or reverses, because a boom that large going into retreat would pull a lot of economic activity down with it, the way any major investment bust does.
If that happened, it would show up in the leading signals. Business investment would roll over. Credit conditions would tighten as lenders reassessed. The market would reprice. In other words, even an AI-triggered downturn would announce itself through the same channels every other downturn does, which is exactly why those channels are what a recession model watches. The mechanism might be new. The signature in the signals would be familiar.
What the signals say right now
Set the forecasts aside and look at what’s actually measurable today, and the leading signals are calm. The yield curve is positive, nowhere near inversion. Credit spreads are narrow, so lenders aren’t pricing stress. Business investment has been holding, and unemployment has stabilized in a low range rather than deteriorating. AI-attributed layoffs, by the trackers Goldman itself cites, have eased from their spring peak rather than accelerating.
That describes near-term risk that is low, with some risk further out on the horizon, the same shape the leading signals have shown for months. The AI disruption Goldman describes is a real long-term force worth taking seriously. It is not, right now, something the recession signals are flashing about. The warning is about what the next recession could look like, whenever it comes, not a sign that one has started.
How to read warnings like this
Big institutional warnings deserve attention, and they also deserve to be read precisely. A few questions keep them in proportion. Is the warning about something leading or lagging, something that forecasts a downturn or something that happens during one? AI layoffs are the second. Is it about now or about a hypothetical future recession? Goldman’s is explicitly about a future one. And does the measurable data back the alarm, or run against it? Here the live layoff data has been easing, not climbing.
None of that makes AI disruption fake or unimportant. It makes it a long-term structural story about the labor market, not a recession signal you can read today. The thing that would tell you a recession is actually coming is the leading signals turning, and when they do, they’ll say so whether the cause is AI, oil, rates, or something nobody’s named yet.
See where the signals stand
The recessionodds.com model updates monthly with the probability of a U.S. recession from 3 to 24 months out, read from five leading public signals, tested out of sample across more than fifty years and eight recessions. Subscribe to get the reading in your inbox when it moves.
recessionodds.com is published for informational purposes only and is not financial, investment, or legal advice.