Agentic AI has been sold to enterprises as the next leap beyond dashboards and chatbots. The pitch is that autonomous agents will not only surface insights but act on them. New survey data, however, suggests the reality is far less impressive: most agent-generated insights still fail to improve the business.

What You Need to Know

Agentic analytics tools are being deployed widely, but their output quality remains uneven. On average, only 48% of insights generated by analytics agents lead to actions that improve the business, up from 37% a year earlier. The main obstacles are a lack of business context, fragmented data and sheer volume of insights that overwhelms human reviewers. Organizations that invest in governance and semantic layers are pulling ahead of the pack.

The Gap Between Agent Output and Business Value

The report, based on a survey of executives at organizations deploying analytics agents, paints a picture of a technology that is improving but still far from reliable. On average, 48% of insights generated by analytics agents lead to actions that improve the business. That is a marked advance from the 37% average respondents reported one year ago.

The improvement is real, but it also means more than half of agent-generated insights still fail to help businesses. That failure carries a cost. Organizations that do not see a return on their agentic analytics investment lose the value they expected. Worse, flawed insights that trigger incorrect actions can cause reputational or financial damage.

A small cohort of organizations, which the report calls insight leaders, performs far better. In this group, more than 70% of agent insights lead to actions that improve the business. These leaders also rate the quality of their agents' insights higher than the overall respondent pool, with the widest gap appearing in prescriptive insights, the recommendations on which actions are based.

Why Context and Volume Undermine Agent Performance

When analytics agents deliver substandard insights, the main culprit is a lack of business context, cited by 55% of surveyed executives. Fragmented or inaccessible data is the other major factor, according to 45%. Fewer insight leaders report that their agents struggle with access to context, suggesting that data foundations separate the best performers from the rest.

Insight overload compounds the problem. At 52% of surveyed organizations, analytics agents generate too many insights to be properly evaluated and acted upon. Less than one-quarter, or 24%, are able to evaluate all agent insights. Most do so on a case-by-case basis, a process that remains human-intensive.

  • Lack of business context: Cited by 55% of executives as the main reason for substandard agent insights.
  • Fragmented or inaccessible data: Named by 45% as a major factor limiting insight quality.
  • Insight overload: At 52% of organizations, agents generate too many insights to evaluate properly.
  • Limited evaluation capacity: Only 24% of organizations can evaluate all agent insights.

The pattern is clear: agents are producing more output than organizations can absorb, and much of that output lacks the grounding needed to be trustworthy. The result is a bottleneck in which human reviewers become the limiting factor in an otherwise automated pipeline.

Governance and Semantics Become the New Battleground

Surveyed companies are responding by prioritizing initiatives that enable better agent insights. Strengthening AI and data governance is the top focus, cited by 60% of respondents. Adding a semantic layer to their data architecture follows at 45%.

Experts interviewed for the report agree that the semantic layer is crucial to ensure high-performing agents. A semantic layer gives agents a shared, governed vocabulary for business metrics and dimensions, reducing the ambiguity that leads to flawed recommendations. Without it, agents may retrieve the right data but interpret it incorrectly.

The emphasis on governance reflects a broader realization: agentic analytics is not just a model problem. It is a data architecture and organizational problem. Companies that treat it as a pure software deployment are likely to see the same disappointing returns that plagued earlier waves of enterprise AI.

Why This Matters

The stakes extend well beyond individual vendors. If agentic analytics cannot reliably produce insights that improve the business, the entire category risks a credibility crisis similar to the one that followed early enterprise AI pilots. Buyers will demand proof of outcomes, not just automation claims.

For data teams, the report shifts the priority from model selection to data readiness. Semantic layers, governance frameworks and context engineering are becoming the differentiators that separate insight leaders from laggards. Organizations that fail to invest in these foundations will keep generating insights that never reach the bottom line.

For vendors, the message is equally direct. The next phase of competition will be fought over context, not raw model capability. Companies that can help customers close the context gap and manage insight volume will win the enterprise budget. Those that cannot will find themselves explaining why their agents produce so much and deliver so little.