CFOs Scrutinize AI Spending Amid Low Productivity Gains
Corporate executives are questioning high AI expenditures as companies report massive budget overruns without significant measurable increases in developer productivity.
Chief financial officers are increasing scrutiny of corporate AI spending as organizations struggle to link high token consumption to measurable productivity gains. Budget overruns have become acute, with Uber Technologies Inc exhausting its entire 2026 AI budget by April due to the use of Claude Code. Another unnamed client reportedly spent approximately $500 million on Claude in a single month.
Infrastructure usage has scaled rapidly, though output gains remain modest. Google reported that its monthly token processing grew from 9.7 trillion two years ago to 3.2 quadrillion in May 2026. Despite this volume, Gartner found that more than half of developers believe AI has improved team output by 10% or less.
Industry experts suggest a shift in how companies measure value. Mark Williamson, CTO of Undo, argues that focusing on token costs is a mistake and that organizations should instead prioritize the cost per solved problem. He recommends optimizing return on investment by matching specific models to task complexity and improving runtime context to reduce hallucinations and overall spending.
