A new MIT Technology Review report finds that fewer than half of analytics agents’ insights actually lead to business improvement — and the top reason for the failures isn’t model size, but a lack of business context. (Credits: MIT Technology Review)
On average, 48% of agent-generated insights at surveyed companies lead to actions that improve the business — up from 37% a year ago, but still meaning most AI-generated insights fail to help. A cohort of “insight leaders,” however, gets more than 70% of insights converting into improvement, largely because their agents have better context.
Agentic Analytics: Context, Governance and Semantic Layers
When insights fall short, 55% of surveyed executives blame insufficient business context, while 45% point to fragmented or inaccessible data. The report also flags “insight overload”: at 52% of organizations, agents generate too many insights to be properly evaluated, and only 24% can review all of them.
Companies are responding by investing in what the report calls the real levers: stronger AI and data governance (cited by 60%) and adding a semantic layer to the data architecture (45%) — rather than simply buying bigger models. Notably, while descriptive insights get praised, prescriptive insights — the ones that trigger actual actions — draw mixed reviews, raising the stakes for getting quality right before agents act on flawed conclusions.
Also Read: The Agentic Web Is Here: MIT Technology Review Says AI Agents Are Quietly Reshaping the Internet
Source: MIT Technology Review — read the full story there.
