In recent years, generative AI has significantly transformed enterprise analytics by democratizing access to data insights across various organizational levels. The emergence of agentic AI further amplifies this potential, offering the ability to automate workflows and execute actions derived from insights generated by these agents. However, a critical challenge persists: the current quality of insights produced by these agents often falls short of expectations.
According to a recent report, less than 50% of insights generated by analytics agents lead to tangible business improvements within the surveyed organizations. While many acknowledge the value of descriptive insights provided by these agents, opinions on prescriptive insights—those that inform actionable decisions—are far more divided. This disparity highlights the need for enhanced insight quality, as ineffective insights can hinder organizational performance and lead to adverse financial or reputational consequences. The report delves into how companies perceive the effectiveness of their analytics agents and the strategies they are implementing to enhance the quality of insights.
Key findings from the report suggest that agentic analytics is in its early stages of delivering meaningful results. Currently, only 48% of insights generated by analytics agents lead to actions that positively impact business outcomes, a modest improvement from the previous year’s average of 37%. In contrast, a select group of organizations identified as ‘insight leaders’ report that over 70% of their agent-generated insights result in beneficial actions. These insight leaders consistently rate the quality of their agents’ outputs higher than the overall average, particularly regarding prescriptive insights. Factors contributing to the quality of insights include the availability of relevant business context and data accessibility, with 55% of executives attributing low-quality insights to a lack of context and 45% citing fragmented data. Additionally, organizations face the challenge of insight overload, where 52% struggle to evaluate the volume of insights generated effectively, leading to a reliance on manual review processes. To address these challenges, companies are prioritizing initiatives that enhance AI governance and introduce a semantic layer to their data architecture, with 60% emphasizing the importance of strong governance frameworks and 45% recognizing the value of semantic enhancements in driving higher-quality insights.
Source: Raising the quality of insights in agentic analytics via MIT Technology Review