Raising the Quality of Insights in Agentic Analytics
Bridging the gap between automated agent outputs and actionable business improvements.
1The Agentic Analytics Gap
Generative AI has transformed enterprise analytics, yet agentic systems struggle with output quality. Currently, fewer than half of agent-generated insights lead to tangible business improvements. While descriptive insights receive praise, prescriptive outputs remain inconsistent, creating risks of financial or reputational damage when incorrect actions are triggered based on flawed automated data analysis.
2Slow Progress in Effectiveness
Data indicates that 48% of insights generated by analytics agents now lead to business-improving actions. This represents a measurable increase from the 37% reported just one year ago. Despite this growth, over half of all generated insights still fail to provide actionable value, highlighting a significant efficiency bottleneck within modern enterprise workflows.
3Characteristics of Insight Leaders
A select cohort of organizations, identified as insight leaders, achieves superior performance compared to the general average. In these companies, over 70% of agent-generated insights successfully drive business improvements. These organizations consistently rate the quality of their agents higher, demonstrating that high-performing AI integration is achievable through specific strategic and operational adjustments.
4Contextual Deficits in AI
The primary driver of substandard insight quality is a lack of sufficient business context, cited by 55% of surveyed executives. Additionally, 45% of respondents point to fragmented or inaccessible data as a major barrier. Insight leaders report fewer struggles with context, suggesting that data accessibility is a critical differentiator for successful agentic analytics.
5Managing Insight Overload
At 52% of surveyed organizations, agents generate an excessive volume of insights, making manual evaluation nearly impossible. Currently, less than one-quarter of companies can evaluate every insight produced. Most firms rely on human-intensive, case-by-case review processes, which limits the scalability of agentic systems and prevents real-time decision-making across the entire enterprise.
6Governance as a Priority
Companies are actively prioritizing initiatives to enhance insight quality, with 60% of organizations focusing on strengthening AI and data governance. Establishing robust governance frameworks is seen as the foundational step to ensure agents operate within defined parameters. This shift reflects a broader organizational move toward formalizing AI oversight to mitigate operational risks.
7The Semantic Layer Necessity
To improve agent performance, 45% of companies are integrating a semantic layer into their existing data architecture. Experts emphasize that this layer is crucial for providing agents with the necessary context to interpret data accurately. By standardizing definitions and relationships, organizations can significantly reduce errors and improve the relevance of automated outputs.
8Strategic Path Forward
The transition from descriptive to prescriptive agentic analytics requires addressing both data accessibility and semantic clarity. Organizations that successfully implement governance and structural data improvements see higher returns on investment. Future success depends on moving beyond simple automation toward high-quality, context-aware insights that reliably trigger actions leading to measurable business growth.
Original article: MIT Technology Review — Raising the Quality of Insights in Agentic Analytics