AI Is Cutting 16,000 U.S. Jobs a Month. Who's Actually at Risk?

A Goldman Sachs research note from early April put a number to something most economists have been hedging around. AI is eliminating a net 16,000 American jobs per month. The mechanics matter more than the headline. Automation is removing around 25,000 roles monthly while augmentation creates roughly 9,000 new ones. The net is negative, the distribution is uneven, and the workers absorbing most of the impact are the ones who can least afford it.

Why the pattern looks the way it does

The roles disappearing first are routine, well-defined, and heavy on data processing or structured communication: data entry, customer service, legal support, billing, and basic content work. These are also the jobs that people with no professional history tend to start in. For Gen Z workers entering the job market now, the traditional entry-level path into many industries is narrower than it was two or three years ago.

The problem isn't only that these roles are being automated. It's that experienced workers hold accumulated knowledge, about clients, processes, exceptions, and professional judgement, that makes them harder to substitute directly. Entry-level workers are the ones who haven't had the chance to build that yet. Goldman's analysis found that wage gaps between workers with high and low exposure to AI substitution are widening by roughly 3.3 percentage points, according to Fortune's reporting.

What the data doesn't support

The aggregate fears, mass unemployment, overnight labor market collapse, the end of human work as a category, aren't supported by what's happening. Unemployment remains low. Productivity is rising in sectors where AI has penetrated deepest. In an economy that adds and removes millions of positions each month, 16,000 net jobs lost is a meaningful trend, not a catastrophe. The technology also creates work. The 9,000 augmentation jobs per month Goldman identified are real positions, often better-paid than the ones they displace. AI hardware manufacturing, model operations, safety and compliance roles, and AI-assisted professional services are all growing.

What the data does support

The generational skew is real. The transition costs are real. Young workers entering the market in 2026 are the first cohort to face a job market where the entry-level rung has already been partially removed in several fields. The standard argument that automation always creates new jobs in the long run has been historically true. It is less reassuring for someone who needs a first job this year, not in the long run.

The policy response has been slow. Retraining programs remain underfunded and poorly matched to actual market demand. Education pipelines are still producing graduates for roles that are contracting. Meanwhile, 96 percent of enterprises are already running AI agents autonomously across business processes. The speed of deployment has outrun the institutions meant to prepare workers for it.

The picture Goldman's data describes is not one of mass displacement across the whole workforce. It's a more targeted story about who pays the transition costs. The answer is the people who had the least margin to absorb them.