The analysis by bdautomated examines 75 statistics from 18 publishers, tracing each figure back to its original report. The dataset, available in CSV and JSON formats under a CC BY 4.0 license, aims to clarify discrepancies in AI adoption data by detailing the specific methodologies and survey questions used by each source.
Data from McKinsey’s 2025 survey shows that while 62 percent of organizations experiment with AI agents, only 23 percent have scaled them, and no more than 10 percent use them in a single business function. Other reports vary significantly: PwC noted 79 percent adoption among U.S. executives in April 2025, while Capgemini found 14 percent implementation.
The U.S. Census Bureau reported that 19.8 percent of businesses of all sizes used AI in any function as of May 2026. The analysis also clarifies the MIT Project NANDA finding, noting that the 95 percent figure refers to profit-and-loss impact within six months of a pilot, rather than a failure rate for all AI projects.
Gartner’s forecast, which suggests over 40 percent of agentic AI projects will be canceled by the end of 2027, remains a prediction from June 2025. The bdautomated report excludes market-size forecasts because the underlying data is proprietary and cannot be independently verified.
