McKinsey warns AI agent costs can vary by 30x as enterprise AI spending pressure grows

McKinsey is warning companies that AI spending could rise further as AI agents become more common. Unlike text-based AI tools, agents must actually complete multi-step tasks, and the same goal can be achieved in different ways, with costs varying by as much as 30 times.

McKinsey's "State of AI 2026" survey shows about one-third of organizations spend over 10% of their tech and communications budgets on AI. 60% of respondents plan to increase AI spending next year, and about one-fifth say AI costs are already pressuring operating expenses. Senior partner Tanguy Catlin noted that the scale of AI spending is becoming quite significant.

Software development teams using agents for automated coding face especially heavy cost pressure. Amazon previously shut down an internal leaderboard tracking AI token usage after some employees ran tasks just to climb the rankings. Companies like Coinbase and Salesforce have also started limiting AI use.

Another McKinsey report says agentic AI can cut manual time for some transformation office tasks by 35% to 40%, and sometimes by over 70%. But companies need to weigh whether the value created justifies the cost.