AI Business Value Drift: When Model Quality Holds but the ROI Quietly Disappears

This article discusses the phenomenon of 'AI Business Value Drift,' where an AI system's return on investment deteriorates over time even though model quality remains technically sound. It highlights factors like provider pricing changes, increased prompt complexity, higher retry rates, more tool calls, and growing human review needs that can erode cost-per-outcome. The author proposes a 'living AI value contract'—a versioned document capturing outcome units, baselines, costs, quality thresholds, and triggers for reassessment. The piece emphasizes that AI value is not static and requires continuous monitoring, citing frameworks from NIST and Microsoft. It is relevant for IT leaders, data scientists, and business stakeholders managing AI deployments.