By 2028, AI coding tools could cost more than an average developer’s salary, according to Gartner. Consumption-based pricing is overtaking flat subscriptions at major AI providers, and token prices are climbing. That shift makes budgeting a moving target: vendors don’t share clear formulas for token usage, so tech leaders struggle to forecast or control expenses. Gartner analyst Nitish Tyagi warns that ungoverned, autonomous AI agents in software workflows are eating into budgets faster than expected, leaving many projects underfunded before they even hit their stride. Few enterprises have a solid strategy or real-time view on what AI is costing them. KPMG data shows only about 25% of C-suite leaders know their systems’ operating expenses. In many cases, agents run in the background for days without anyone tracking their loops or throttling premium-model calls. KPMG’s Rahsaan Shears urges CIOs and CFOs to adopt a tokenomics model that ties usage to cost, assigns ownership and measures business value. He recommends consolidating AI spend across cloud platforms, copilots and coding tools, then auditing which agents are mission-critical and which can be dialed back. Without that level of visibility, a sudden price spike or feature cut by a provider could blindside the entire organization.