Enterprises built data systems for human users, but the rise of AI agents is exposing a gap in the underlying infrastructure.
Survey reveals widespread delays tied to data hurdles
Research from Cloudera, in partnership with Wakefield Research, surveyed 1,500 enterprise architects and cloud leaders. A striking 95% reported postponing or canceling AI projects in the past year because of data‑governance, compliance, or regulatory obstacles. The study described the situation as a “great AI re‑architecture” need.
Simultaneously, a Google and MIT study of 300 IT executives found more than half had paused AI agent deployments to address foundational data issues such as silos and lack of context. Those legacy systems also caused a “significant negative impact” on AI return on investment, with high latency slowing decision‑making.
Related: Nvidia Invests Heavily in Artificial Intelligence
Both studies note that while many organizations are eager to adopt agentic AI, the data foundation required for these agents—multimodal, context‑aware, and instantly accessible—remains out of reach for most.
Data leaders vs. laggards: the trust gap
The Google/MIT survey showed 98% of respondents either use or plan to use AI agents, with ten percent already doing so at scale. Common applications include customer‑service routing, IT systems management, and security threat detection. Future plans extend to HR, finance, and supply‑chain functions.
Enterprises classified as “data leaders” grant agents access to more than 70% of their data, and all of them report agents making mostly or consistently accurate decisions. In contrast, “data laggards” share only about 30% of data and see trust in AI decisions drop to roughly 22%.
Improving data access is seen as the top priority. Respondents highlighted the need to discover, inventory, and classify both structured and “dark” unstructured data, then apply governance before linking it to AI agents.
Related: Top power banks for MacBooks this year
Providing “frictionless access” to operating systems, multimodal data, and context helps agents align data with varied team goals. The recommendation calls for replacing batch processing with streaming and event‑driven pipelines to keep data fresh for AI workloads.
Even if an organization upgrades its data pipelines, the cultural shift required to trust AI decisions may lag behind the technical changes.
Trust in data will determine AI’s future.
