Artificial intelligence is coming for millions of American jobs. Exactly what happens next is where the consensus breaks down. Two significant pieces of research released this week paint a picture that is simultaneously alarming and more nuanced than the headlines suggest — one pointing to 15 million displaced workers, the other showing that the companies investing most heavily in AI are actually growing their headcount.
The sobering number comes from Goldman Sachs. Joseph Briggs, who leads the bank’s global economics research team, estimates that approximately 9% of the US workforce will be displaced as AI adoption spreads — a figure that translates to roughly 15 million workers who would need to leave their current positions and find new jobs. In sectors where AI tools are already in active use — technology, management consulting, and graphic design — Briggs estimates the technology is already cutting between 10,000 and 15,000 jobs from monthly employment growth figures. The June jobs report, released Thursday, showed the US economy added just 57,000 jobs — roughly half what economists had expected — with April and May also revised down by a combined 74,000.
The picture, however, is not simply one of destruction.
Why Heavy AI Adopters Are Actually Hiring More
A new study from financial services company Ramp and employment database Revelio Labs, tracking AI spending and workforce records across nearly 22,000 US companies between January 2021 and February 2026, found that firms which invested heavily in AI expanded their total headcount by an average of 10% in the two years after deploying the technology. Those making the largest AI investments saw entry-level hiring grow by 12%.
“If you are a job seeker, or you are graduating from college, and you’re choosing between two different firms that are otherwise similar, I would choose the one that’s using AI,” said Ara Kharazian, Ramp’s lead economist. “Our paper shows that that firm is going to grow faster.”
The distinction matters: the companies seeing hiring gains were not casual ChatGPT users. They were spending more than $100 per month per employee on AI and deploying advanced tools — coding subscriptions, specialised AI assistants — rather than general-purpose chatbots. Low-intensity AI adopters saw no hiring gains and in many cases reduced headcount. The finding also suggests that some corporate layoffs attributed to AI may have been driven by ordinary cost-cutting instead, with executives using the technology as cover. “When you hear CEOs talk about layoffs and they attribute it to AI, I would be skeptical,” Kharazian said.
Why the Research Does Not Fully Contradict
The Ramp study appears to conflict with a November 2025 Stanford University analysis, which found employment among young software developers had dropped nearly 20% from its late-2022 peak. Kharazian argues both findings can be true simultaneously. The Stanford study looked across the entire labour market; Ramp’s examined specifically the firms investing most aggressively in AI — many of them fast-growing, venture-backed companies actively hiring AI-native junior employees. The overall market can weaken while the most forward-looking segment of it thrives.
A separate California AI unemployment tracker adds a further complication that challenges common assumptions about who AI is actually hurting. Rather than predominantly affecting young or entry-level workers, the data shows that unemployment insurance claims among college-educated workers in high-AI-exposed jobs — customer service, software development — have been elevated since ChatGPT’s release in 2022. Claims among workers with master’s degrees and PhDs in highly AI-exposed occupations rose from a baseline of 13,000 per month in late 2022 to between 16,000 and 22,000 per month by mid-2023 and have remained there. The job losses are concentrated in the technology sector and disproportionately centred in the San Francisco Bay Area — but they span workers aged 36 to 65, not just early-career employees.
The Historical Argument for Optimism
Briggs at Goldman Sachs is not a pessimist about the long-term outcome. He points to the fact that roughly 30 million jobs are created and 29 million destroyed in the US every year regardless of AI — and that even a 5% increase in the pace of job creation would be sufficient to reabsorb everyone the technology displaces. “If we look back over the last 80 years, around 85% of job growth has been driven by the technological creation of new positions,” he said.
MIT economist Neil Thompson, who appeared alongside Briggs on Goldman’s podcast, argues the shift will move more slowly than AI’s accelerating capabilities might suggest. Technical capability is only the first step — an AI system also needs access to the right data, which is often restricted in fields like medicine by privacy regulations, and it needs to be cheap enough to justify deployment at scale. When GPS automated taxi drivers’ intimate knowledge of city routes, wages fell — but the number of drivers grew dramatically. The outcome depended on which tasks were automated, not merely that automation occurred.
Thompson frames AI as a “rising tide” that workers can see approaching and adapt to, rather than a “crashing wave” that sweeps them away without warning. Whether the June jobs report represents the first sign of that tide — or something more abrupt — is a question the data does not yet answer.
Stay informed. Subscribe to the JournalTodays Newsletter for the latest AI economy news, jobs research, and business coverage delivered straight to your inbox.
