70% of Leaders Cannot Track Where AI-Freed Time Goes, Studies Show

Multiple surveys, including research from McKinsey, Gartner, Deloitte, and ServiceNow, reveal that only about 22–30% of organizational leaders can demonstrate significant, measurable value from their AI tools. An Anthropic internal study found that despite strong adoption and self-reported productivity gains among engineers using Claude Code, actual delivery metrics showed no improvement. Experts highlight two common ways AI savings are inflated: counting freed-up hours as hard dollar savings when budgets never actually decreased, and conflating calendar time with real labor time. Freed hours only translate into genuine financial returns when redeployed on value-generating work, used to avoid new hires, or converted into increased output. The core problem is that most organizations track AI usage and token costs — both on the expense side — while failing to measure whether the work AI produced delivered defensible business value.