Independent Veqtor8 research finds a governed AI data layer can cut AI operating costs by 68% and accelerate agent deployment threefold, highlighting the importance of trusted data infrastructure as enterprises move AI into production.
Enterprises scaling agentic AI are facing a growing challenge: connecting autonomous and semi-autonomous agents to reliable, live and governed data across fragmented IT environments. A new independent study commissioned around Denodo’s AI data layer approach suggests that addressing this challenge can deliver measurable financial and operational benefits.
Research firm Veqtor8’s report, The ROI of an AI Data Layer for Agentic AI, finds that deploying a governed AI data layer can deliver a 290% return on investment over three years, reduce AI operating costs by 68% and enable agents to reach production three times faster.
The study draws on interviews with enterprise practitioners, hyperscalers and technology partners across banking, telecommunications, chemicals and logistics. Its financial model represents an enterprise scaling from 100 AI agents in the first year to 900 in the third year, accessing data across 100 distributed sources.
According to the research, reusable data products can accelerate deployment by reducing the need to build custom integrations for every agent. The model also identifies more than 75% potential cost avoidance in integration and governance, with maintenance costs falling by 79% and policy administration effort declining by 76%.
“Enterprises cannot afford to choose between speed, governance, and cost efficiency when scaling AI.”
— Alberto Pan, Chief Technology Officer, Denodo
The findings point to a structural challenge in enterprise AI adoption. Without a unified data layer, organisations may need to build and maintain separate Model Context Protocol (MCP) servers, retrieval-augmented generation pipelines and security controls for different data sources and agents. This complexity increases as production agents interact with dozens of cloud, software-as-a-service, legacy and operational systems.
Andrew Milroy, Chief Analyst at Veqtor8 and author of the report, said that connecting individual systems is relatively straightforward, but coordinating multiple systems into consistent workflows under common security policies is considerably more complex. Denodo’s AI Data Layer, he added, allows organisations to profile data once and reuse governed data products across authorised agents.
The study also links data access to AI operating costs. By querying the data layer directly and filtering information before it reaches a model, the research model reduces context payloads from approximately 18,000 tokens to 5,760 tokens per task. This reduction can lower the compute costs associated with processing unnecessary data.
For enterprises moving beyond experimental AI pilots, the results underline the importance of treating data architecture as a core part of agent deployment rather than an integration task to be solved separately for each application.
Alberto Pan, Chief Technology Officer at Denodo, said the Denodo Platform provides unified access, consistent semantics and centralised policy enforcement across distributed data environments without requiring large-scale data relocation.
The full Veqtor8 report, The ROI of an AI Data Layer for Agentic AI, is available free of charge at https://www.denodo.com/en/document/analyst-report/research-report-roi-ai-data-layer-agentic-ai.
