Report: Only 48% of AI agent analyses prompt business action

An average of 48% of analyses produced by enterprise AI agents led to business improvement actions, according to a report published on November 2 by MIT Technology Review Insights.

The survey covered 400 senior data and technology executives across 8 industries worldwide and was conducted in July and August. Respondents rated descriptive insights more favorably than prescriptive ones: 83% called descriptive insights good, while 45% said the same of prescriptive insights.

Performance varied widely. Among the top-performing group, which accounted for 7% of respondents, 77% of analyses led to business improvement actions. A year earlier, respondents recalled, the share across the board was 37%, 11 percentage points below the current figure.

Data gaps and volume limit action

Asked what undermines insight quality, 55% of respondents cited a lack of business context and 45% pointed to fragmented or hard-to-access data. Volume is a further obstacle: 52% of companies said AI generates too many insights to review and act on all of them, while only 24% said they can evaluate every result.

Companies named strengthening AI and data governance as a priority task (60%), followed by adopting a semantic layer (45%).