They can, unless the freed-up time goes into work that adds to profit. With AI, the same work can be done with fewer people, but the business takes on a new cost, the AI bill. McKinsey's October 2026 report, The AI value curve: the economics of applying AI to existing work, advises planning how to redeploy the people AI frees up. That needs enough higher-value work to move them to. Amar Harolikar's read is that most businesses are unlikely to grow sales fast enough to create it.
The AI bill keeps growing. McKinsey cites one estimate that company spending on large language models tripled over the 12 months to the end of 2025.
What large employers are doing. Oracle had about 162,000 staff in May 2025 and about 141,000 in May 2026, per its annual reports, and its 2026 restructuring plan cites the adoption of AI. Microsoft is offering voluntary buyouts to about 60,000 staff, 27% of its workforce, according to a person briefed on the plan cited by CNBC. Staff decide in early December and leave on 31 December.
What McKinsey itself says. Highly paid expert roles are unlikely to be replaced by AI, and its figures measure the tasks within each job that AI can take on. The job-cuts read is Amar Harolikar's. McKinsey's figures come from a model with a world view and a US view only, and they leave out one-time costs such as platforms, training and change management.
Full post with the deck: https://www.tigzig.com/post/mckinsey-ai-value-curve-job-cuts-oct2026. Related: is AI reducing jobs for young workers? and how AI has changed work inside Anthropic.
Building something like this? How I work covers the rates, the availability and what I take on.