Business leaders say AI's return on investment has fallen short of expectations, with companies cutting token spending as measurement challenges persist.

IBM Vice Chairman Gary Cohn stated Thursday that despite industry-wide 'massive over-investment,' the ROI from AI has 'not been nearly as high as people might think.' Conversations at New York Tech Week reflected CEOs attempting to justify AI costs to their boards.

The market for AI coding capabilities jumped sevenfold from 2024 to 2025, reaching $4 billion, according to Menlo Ventures. However, two startup CEOs told Semafor that companies are reducing AI spending not because tools lack usefulness, but because results are difficult to measure—a longstanding challenge in software engineering ROI.

AI companies that automate measurable work in areas like sales and customer experience are positioned to benefit in the near term, the startup CEOs said. The broader adoption of AI in software engineering remains constrained until companies develop better methods to quantify task value.