Analytics agents are making it easier for people across an organization to get insights from data, and they promise to automate workflows and inform actions. Yet the findings in a new report show that useful outcomes remain far from routine: fewer than half of the insights generated by agents at surveyed organizations lead to business improvement.
More insights are helping, but most still fall short
On average, 48% of agent-generated insights at surveyed companies lead to actions that improve the business. Respondents say that figure was 37% one year ago, suggesting progress. Still, more than half of the insights do not lead to an action that benefits the business.
The report distinguishes between descriptive insights, which respondents generally rate positively, and prescriptive insights, which inform what an organization should do. Reviews of those prescriptive outputs are mixed. That difference matters because an insight used to guide action carries consequences beyond simply describing what the data shows.
When an agent’s recommendation is sound, it can help an organization act on information. When it is flawed, acting on it can cause financial or reputational damage. The report frames insight quality as central to whether companies get the return they want from agentic analytics.
A small group reports stronger results
The report calls a small cohort of organizations “insight leaders.” More than 70% of agent insights at these companies lead to actions that improve the business, substantially above the overall average.
These organizations also give their agents’ insights higher quality ratings than respondents overall. The gap is especially wide for prescriptive insights. The findings associate stronger insight quality with a clearer business advantage, though they do not suggest that simply deploying an agent guarantees better decisions.
The comparison points to a practical challenge for other companies: improving the usefulness of outputs, particularly recommendations that influence what people do. Higher volumes of answers or broader access to analytics do not by themselves ensure that agents understand what the business needs.
Agents need business context to interpret data
The most commonly cited reason for substandard insight quality is a lack of business context, named by 55% of surveyed executives. Fragmented or inaccessible data is another major obstacle, cited by 45%.
In plain terms, an agent may be able to produce an output from available data while still missing the organizational meaning needed to make that output useful. If relevant information is split across sources or hard to reach, the agent’s view can also be incomplete. The report says fewer insight leaders report difficulty giving their agents access to context.
This helps explain why the report highlights data foundations and governance alongside AI capabilities. A stronger model alone would not resolve the stated problems if an agent still lacks the context to interpret information or cannot access relevant data.
Too many insights can slow action
Insight quality is only part of the challenge. At 52% of surveyed organizations, agents generate more insights than people can properly evaluate and act on. Only 24% say they are able to evaluate all the insights their agents produce.
Most organizations review insights case by case, and the report describes that process as human-intensive. When outputs arrive faster than people can assess them, potentially useful recommendations may not receive attention. A growing stream of agent insights can therefore add work without reliably translating into better decisions.
The findings suggest that organizations need to consider whether their processes can keep up with the outputs agents create. The report does not offer a universal method for sorting or prioritizing insights, but it makes clear that evaluation capacity is a constraint for many respondents.
Governance and semantic layers are priorities
Surveyed companies are focusing on governance and data meaning to improve agent outputs. Strengthening AI and data governance is cited by 60%, while 45% are adding a semantic layer to their data architecture. Experts interviewed for the report describe that layer as crucial for high-performing agents.
A semantic layer can help connect data to shared business concepts, while governance initiatives address how AI and data are managed. Together, these priorities align with the barriers executives identify: agents need access to information and enough business context to produce insights people can assess and use.
The report’s central finding is that agentic analytics is beginning to deliver more business-improving insights, but quality remains uneven. For organizations seeking better results, the emphasis is on context, accessible data, governance and semantics, as well as the ability to evaluate outputs before acting on them.