Report highlights challenges in agentic analytics quality

A new report from MIT Technology Review Insights, produced in partnership with Snowflake and published on Nov. 2, 2026, reveals that the quality of insights generated by agentic AI remains a significant hurdle for enterprises. While generative AI has made analytics more accessible, the transition to autonomous agents that can execute workflows has exposed gaps in performance.

The research, which surveyed 400 data and technology executives, found that 48% of agent-generated insights currently lead to actions that improve business outcomes, an increase from 37% a year ago. However, organizations struggle with the volume of data produced, as 52% of respondents report generating too many insights to evaluate effectively, and only 24% possess the capacity to review all of them.

Executives identified two primary barriers to high-quality output: 55% cited a lack of business context, while 45% pointed to fragmented or inaccessible data. Carl Perry, VP of Analytics at Snowflake, noted that agents often struggle to determine the true value of their findings when they lack a sufficient understanding of business needs, requirements, and priorities.

The survey data shows a disparity in how different types of insights are perceived. While 83% of respondents rated descriptive insights as good or excellent, only 45% gave the same rating to prescriptive insights, which provide recommendations for future actions.

To address these quality issues, the report suggests that organizations should prioritize investments in semantic layers and governance rather than focusing solely on acquiring larger AI models. A small group of organizations identified as insight leaders, representing 7% of those surveyed, have achieved a 70% success rate with their agents. Looking ahead, 60% of executives predict that the orchestration of teams of AI agents will become a primary role for analysts within two years.