The question of whether AI is killing jobs has become a political football, which makes it harder to get a straight answer. The actual employment data is more useful than most of the commentary around it.
Customer service is the clearest case of real displacement. According to a McKinsey report from late 2025, companies with more than 500 employees reduced call center headcount by an average of 23 percent between 2023 and 2025, largely by routing more interactions through AI systems. That’s not hypothetical job risk. Those jobs are gone and have not been backfilled. Document review in law firms is another concrete example. Firms that have adopted AI-assisted discovery tools have reduced paralegal hours for that specific task by 30 to 40 percent, per a 2025 American Bar Association survey. Routine data entry roles across insurance, banking, and healthcare billing are shrinking in a measurable way.
Where the picture gets more complicated is in knowledge work. Programmers are not being replaced; they’re being asked to do more with the same hours. The Bureau of Labor Statistics added 48,000 software development jobs in 2025, not lost them. Graphic designers, copywriters, and marketing professionals are seeing some tasks automated but are also being asked to manage and direct more AI-generated output. Whether that nets out positive or negative depends heavily on how quickly people adapt and how their employer is structured.
Manufacturing is largely a separate conversation from AI, though it often gets lumped in. The automation hitting factory floors is mostly robotics and process control that predates the current wave of large language models. Warehouse fulfillment is a genuine area of concern: Amazon’s robotics deployment reduced picking-job headcount at fully automated facilities by roughly 15 percent since 2022, according to labor union filings in 2025.
What the forecasts say: McKinsey and the World Economic Forum both estimate that between 2025 and 2030, AI and automation will displace roughly 12 million jobs in the US but create around 10 million new ones in different categories, from AI oversight roles to skilled trades that can’t be automated easily. The net number is smaller than the scary headlines. The transition is still real, because the jobs lost and the jobs created are not in the same place or requiring the same people.