Goldman Sachs's Chief Economist Said 'The Big Story in 2026 in Labor Will Be AI.' He Said Job Losses Could Be Pulled Forward. Nobody Put That on the Front Page.

In the second half of 2025, job growth in the United States began decelerating in ways that proved durable. By November, unemployment had risen to 4.6% — up from 4.1% in June. Goldman Sachs tracked what was happening and flagged it directly: management consultants, call center workers, and graphic designers were seeing early displacement. Entry-level tech workers were absorbing the worst of it. Their unemployment rate jumped 3 percentage points in one year — a rate Goldman's own economists called 'much larger than we've seen in the tech sector more broadly and a larger increase than we've seen for other young workers.' George Lee, co-head of the Goldman Sachs Global Institute, put it plainly: 'Young employees for this period of time are a little bit the casualty of that.' This is not a projection. It is what already happened in 2025. The question for 2026 is whether it broadens.

The Frontloading Problem: Why the 10-Year Estimate Is the Optimistic Scenario

Every analysis of AI job displacement quotes the headline number: Goldman Sachs projects 6-7% of the US workforce displaced over the next decade. That number is actually the good scenario. It assumes adoption plays out gradually over 10 years, giving workers and employers time to adapt. Joseph Briggs explicitly described the alternative: 'If it's more frontloaded, the impacts on the economy are much larger.' And then Briggs said, in April 2026: 'If we see some job losses pulled forward, that sets stage for potential underperformance relative to our forecast, and that may lead the Federal Reserve to cut rates.' He named AI as 'the big story in 2026 in labor.' The Fed cutting rates because of AI-driven job losses would be a macroeconomic signal that the frontloading scenario — not the comfortable 10-year scenario — is playing out. Watch for that signal.

The Numbers Behind Who Is Actually at Risk

• 2.5% of US employment — approximately 4 million workers — faces direct displacement risk from AI capabilities that already exist today, according to Goldman Sachs's near-term model. This is the floor, not the ceiling.
• 15.1% of US employment — 23.2 million workers — already have 50% or more of their tasks automated, per SHRM's 2025 survey of 20,262 US workers. These workers have crossed the automation threshold without most of them knowing it.
• 6.1 million workers specifically identified by Brookings face both high AI exposure AND low adaptive capacity — meaning limited savings, limited transferable skills, and limited reemployment options. These are the workers who cannot absorb a job transition the way higher-income workers can.
• 13.7% of US workers, per National University research, say they have already lost a job to robots or AI-driven automation. That is approximately 22 million people.
• 37% of business leaders anticipate replacing human workers with AI by the end of 2026 as pilot programs scale, per HRDive survey data.
• 55,000 job cuts were directly attributed to AI by company announcements in 2025 — out of 1.17 million total layoffs. Amazon eliminated 14,000 corporate roles citing AI. Workday cut 8.5% of its workforce to 'reallocate resources toward AI investments.'

The Displacement Risk by Job: What the Data Actually Shows

The Demographic Data That Mainstream Coverage Ignores

The distribution of AI displacement risk is not equal across gender, and the data is striking in ways that most AI journalism has failed to engage with honestly. Research compiled by DesignRush's 2026 analysis found that 79% of employed US women work in high-automation-risk jobs, compared to 58% of men. More specifically, 86% of workers in the highest-risk administrative and clerical roles are female. This is not a coincidence — it reflects the labor market structure where the clerical, administrative, and customer service roles that AI is automating most aggressively are disproportionately held by women. The Brookings Institution specifically identified approximately 6.1 million workers at the intersection of high AI exposure and low adaptive capacity. These workers are disproportionately lower-income, have less savings, and are in geographic labor markets with fewer alternative opportunities. If AI displacement front-loads — which the early employment data suggests is already beginning — these 6.1 million workers absorb the worst of it with the fewest resources to recover.

The AI Skills Premium: The Most Actionable Number in This Article

The AI skills salary premium reached 56% in Q1 2026. Workers who can demonstrate verifiable AI skills — not casual ChatGPT use, but documented workflow automation, AI-assisted analysis, specific tool proficiency that employers can verify — command salaries 56% higher than equivalent workers without those skills. This single statistic changes the calculus. The 6-week investment in building demonstrable AI skills for your specific industry produces a salary premium that compounds annually. It is not hyperbole to say it is the highest-ROI professional development investment available right now for most American workers. The skills that earn the premium are domain-specific — a financial analyst automating quarterly reports earns it, a healthcare worker integrating AI diagnostic tools earns it, a marketer automating research and brief generation earns it. Generic AI use does not.

What To Do Starting This Week: The Specific Actions

• This week — run the task audit: List the 10 tasks that constitute 80% of your working week. Mark each one: Is it rule-based and information-processing (high displacement risk)? Or does it require physical presence, unpredictable real-world judgment, or accountability structures that cannot be delegated to software (lower risk)? Most people overestimate the judgment required in their role and underestimate how automatable it actually is. Be specific and be honest.
• This week — find your augmentation tasks: In every role, there are tasks where AI makes you dramatically more productive without replacing you. These are tasks you currently do manually that AI can accelerate 5x — research synthesis, first-draft generation, data pattern analysis, report compilation. Identify yours. AI that makes you 5x faster at billable tasks is not a threat — it is the reason you keep your job and get the raise.
• Next 30 days — build demonstrable domain-specific AI skills: Not 'I use ChatGPT.' Build something specific to your field: automate a reporting workflow, build an AI-assisted analysis pipeline, document a process you've streamlined using AI tools. Something you can show in an interview or a performance review. The salary premium goes to people who can demonstrate impact, not people who can explain what a prompt is.
• Next 30 days — document every efficiency gain: Track what you automate, quantify the time saved, note the quality improvement, and make your manager aware. In a world where AI-driven restructuring is affecting entire departments, 'I reduced our quarterly reporting cycle from 8 days to 1.5 days using AI workflow automation' is the sentence that keeps your position. Silence about your productivity gains helps no one.
• Next 90 days — expand your professional network: The workers most at risk of sustained hardship are those with narrow networks concentrated within a single employer or industry. AI displacement often plays out as a wave — entire teams restructured, not individual positions eliminated. A professional network that extends across employers, industries, and skill domains is the most durable career asset you can build. LinkedIn connections are not enough — cultivate relationships with people who would answer a referral call.

Goldman Sachs ultimately projects that AI will raise US labor productivity by approximately 15% when fully adopted — historically, productivity gains of this magnitude lead to new job creation and higher wages. The long arc is not catastrophic. But the disruption is real, uneven, and already accelerating for specific groups. The 6.1 million workers with high exposure and low adaptive capacity are not abstract statistics. They are specific people, in specific roles, in specific geographic markets, with limited ability to absorb what is coming if it front-loads faster than the baseline scenario. The actions above are not about being afraid. They are about not being in that group.