A new McKinsey survey of enterprise technology leaders finds that AI spending is outrunning what most companies budgeted for it. As organizations move from isolated pilot projects to enterprise-wide deployment, AI spend increases nearly fourfold, according to the analysis. Sixty-two percent of surveyed organizations have moved beyond experimentation into active AI deployment, yet 93 percent report exceeding their AI budgets, and a majority expect their AI spend to climb by at least 25 percent over the next 12 months.
McKinsey attributes much of the overrun to fragmented, hard-to-track consumption. Business units purchase AI tools independently, employees build AI-powered workflows outside central IT oversight, and usage tied to tokens and API calls behaves far less predictably than traditional software licensing. The same task can generate dramatically different costs depending on which model or agent chain handles it, and token usage for a single task type can vary by as much as 30 times, according to the report.
Despite the overruns, McKinsey found real savings available to companies that manage AI consumption deliberately. Organizations that actively optimize spending, through steps such as routing tasks to lower-cost models, capping response lengths, and reusing static prompt context, have captured savings of 20 to 30 percent. Reusing cached prompt context alone can cut repeated input costs by roughly 90 percent for certain workloads. Still, only 20 to 25 percent of companies currently have mature AI financial-operations practices in place, meaning most organizations remain without the visibility needed to catch overspending before it happens.
Source: McKinsey and Company - https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-cost-of-intelligence-how-cios-can-manage-ai-demand-at-scale