Companies seeking to expand the use of artificial intelligence face an increasingly relevant challenge: how to convert their investments into business results. According to a new study powered by Denodo and prepared by the research firm Veqtor8, the implementation of an AI data layer can generate a return on investment of 290% in three years.
The report, titled “The ROI of an AI Data Layer for Agentic AI,” quantifies the economic and operational advantages of implementing an AI data layer, such as the one offered by Denodo, compared to fragmented and custom-developed architectures.
Thus, the research also reveals that this approach allows AI agents to be deployed three times faster, reduce their operational costs by 68%, and estimates savings of more than 75% in integration and governance costs.
As organizations move from experimenting with pilot projects to deploying autonomous and semi-autonomous agents in their operations, having up-to-date, consistent and secure information becomes a relevant factor in controlling costs and facilitating the launch of new applications.
Scale AI through layers without multiplying costs
To quantify the economic impact, the research has been based on a three-year business financial model that includes experiences from sectors such as banking, telecommunications or logistics.
The scenario analyzed contemplates a company that goes from operating with 100 AI agents in the first year to 900 in the third, connected to 100 distributed data sources. In this context, the study identifies four main results:
- A return on investment of 290% in three years. The calculation combines the business value generated by AI initiatives with the costs that can be avoided through a more efficient data architecture.
- Deployment of AI agents accelerates up to three times faster. Reusing data products reduces the need to develop specific integrations for each new implementation.
- 68% lower AI operational costs. Accessing the data through a specific layer and filtering information before sending it to the models reduces the volume of context per task from about 18,000 to 5,760 tokens, which helps reduce computing costs.
- More than 75% savings in integration and governance costs. The report estimates a 79% reduction in maintenance costs and a 76% reduction in the effort dedicated to managing access and control policies.
Together, these figures suggest that the economic impact of artificial intelligence does not depend solely on the capabilities of the models. How companies manage information, connect their systems, and allow agents to access it also influences costs and the ability to expand its use.
Without a unified AI data layer, enterprise agents require custom Model Context Protocol (MCP) servers, RAG pipelines, and agent-specific security controls on each underlying data source. In regulated industries, where a single production agent may interact with 30 to 50 disparate cloud, SaaS, legacy, and operational systems, this approach leads to serious integration challenges and governance vulnerabilities.
Without a unified AI data layer, enterprise agents require custom Model Context Protocol (MCP) servers.
“Enterprises cannot afford to choose between speed, governance and cost efficiency as they expand the use of AI,” says Alberto Pan, Chief Technology Officer (CTO) at Denodo. “By unifying access, semantics, and centralized policy enforcement across distributed data environments without the need for massive data relocation, the Denodo Platform provides the essential foundation for reliable, enterprise-grade AI agents.”
