Signal Only about 40% of agentic AI initiatives judged successful on objective metrics

Summary

IBM cautions that having AI agents does not guarantee success. Executives report that in 2025 roughly 40% of agentic AI projects counted as successful when judged by objective metrics. The figure contrasts with high confidence elsewhere in the survey, where chief data officers say agent benefits exceed risks and many firms plan to expand autonomous agents. IBM's answer is that organisations need to elevate enterprise AI so that agents can track many signals together, analyse interdependencies and model scenarios live. The signal highlights a gap between enthusiasm and measurable results that organisations will need to close.