McKinsey Warns AI Agents could Drive Up Corporate AI Costs

McKinsey & Co. is warning companies that AI spending could become more expensive and less predictable as businesses increasingly deploy AI agents.

AI agents can perform multistep tasks through different workflows, potentially creating significant variation in computing costs for businesses.

The consulting firm found that the cost of completing the same task with different agents can vary by as much as 30 times, Lakshmi Varanasi reported for Business Insider.

Unlike conventional text-based AI tools, agents are designed to complete tasks, often through multiple steps.

Different agents can take different routes to reach the same result, causing their computing and token usage to vary substantially.

McKinsey’s 2026 State of AI survey found that about one-third of organizations reported spending more than 10% of their technology and communications budgets on AI.

About 60% of respondents said they planned to increase AI spending over the following year, while roughly one in five said AI spending was beginning to constrain operating costs.

McKinsey has also found that agentic AI can substantially reduce the amount of human time spent on certain tasks.

In some transformation-office applications, the firm said the reduction can reach 70% or more.

But the potential productivity benefit must be weighed against the cost of running the systems.

The findings point to a growing challenge for businesses: measuring AI not simply by the price of individual model calls but by the cost of completing a useful task or achieving an outcome.