As the era of agentic AI dawns, business and technology leaders are increasingly embracing this transformative technology. AI agents are swiftly being integrated into organizations, and executives are recognizing their vast potential to enhance operational efficiency. However, many companies encounter challenges in achieving the anticipated return on investment (ROI) from these systems. A key barrier lies in inadequate infrastructure and insufficient data, which remain significant hindrances to effective AI implementation.
The functionality of agentic AI introduces new demands on existing enterprise data frameworks. Unlike traditional systems that primarily respond to inquiries, AI agents require comprehensive access to both structured and unstructured data from across the organization, all contextualized for effective decision-making. To operate efficiently and in real-time, these agents must seamlessly interact with various operational systems, including those related to supply chain logistics, point-of-sale transactions, and human resources. Unfortunately, many legacy data systems, even those that have undergone recent upgrades, are ill-equipped to meet these evolving requirements.
The urgency to upgrade data systems is amplified as AI agents become more integrated into business operations. According to Gartner’s projections, by 2027, AI agents will be integral to augmenting or automating half of all business decisions. Organizations need to address existing bottlenecks or risk limiting the capabilities of their AI agents in making timely and accurate decisions. A recent report surveying 300 data and technology executives reveals that while a few organizations—designated as “data leaders”—successfully navigate legacy system limitations, the majority struggle with inadequate data access. The findings highlight that data leaders enjoy better outcomes with their AI agents, demonstrating the necessity of establishing a robust data environment for scaling trusted AI systems.
The report uncovers several critical insights, including that on average, AI agents currently access only 45% of enterprise data across all surveyed organizations. This percentage drops to 30% or less in organizations identified as “data laggards,” while data leaders ensure that over 70% of their data is accessible. Furthermore, trust in the decisions made by AI agents correlates strongly with data readiness; while only about half of organizations express confidence in their agents’ accuracy, all data leaders trust their AI systems, indicating that reliable AI relies on a solid data foundation. Additionally, two-thirds of data laggards cite legacy systems as a barrier to scaling AI capabilities, whereas data leaders report minimal constraints. With plans to fully integrate agentic AI within two years, organizations are prioritizing improvements in data access and governance to enable AI agents to thrive.
Source: Scaling AI agents with trustworthy data via MIT Technology Review