Scaling AI agents with trustworthy data
2026-08-20 · MIT Technology Review
Scaling AI Agents with Trustworthy Data
The Data Foundation Imperative for Agentic AI
Business and technology leaders need no convincing that the age of agentic AI is upon us. Organizations are rapidly adopting AI agents, and few executives doubt the technology’s transformative potential. However, many are discovering that achieving desired return on investment hinges on having the right foundational infrastructure, with inadequate data systems emerging as a primary obstacle.
Agentic AI imposes significant new demands on enterprise data platforms. Moving from answering questions to taking actions requires agents to access data across the entire enterprise — in both structured and unstructured forms — enriched with relevant business context. To make decisions and act in real time, agents also need frictionless connectivity to operational systems holding supply chain, point-of-sale, and HR data. Even legacy systems refreshed only a few years ago are struggling to meet these requirements.
Legacy Systems Are Constraining Agent Performance
As AI agents become more deeply embedded in enterprise operations, the urgency to overcome limitations of legacy data systems is growing. If Gartner’s forecast that agents will augment or automate 50% of business decisions by 2027 proves accurate, organizations must remove these bottlenecks or risk depriving agents of the data they need to decide and act at speed.
This report, based on a survey of 300 data and technology executives, examines how legacy systems are limiting the effectiveness of AI agents in most organizations. It identifies a small group of “data leaders” who are achieving significantly better results with agentic AI and experiencing far fewer data-related constraints. These leaders provide a roadmap for building the data environment necessary for agents to scale reliably.
Key Findings
Limited Data Access Remains Widespread
Across all surveyed organizations, AI agents have access to just 45% of company data on average. This figure falls to 30% or lower among “data laggards.” In contrast, data leaders provide their agents with access to over 70% of enterprise data — and are seeing markedly better outcomes.
Trust in Agent Decisions Reflects Data Readiness
Today, only about half of organizations trust that the decisions made by their AI agents are accurate and relevant. By comparison, 100% of data leaders express full trust in their agents’ decisions. This stark difference strongly indicates that reliable AI requires a reliable data foundation.
Data Leaders Scale Agents More Easily
Two-thirds of data laggards report that legacy systems are limiting agent scaling (66%) and preventing agents from making decisions at speed (68%). Having largely overcome these legacy constraints, data leaders report these roadblocks at just 8%.
Universal Push Toward Agentic AI Increases Pressure
Within two years, 100% of respondents plan to be using agentic AI, with 69% expecting to deploy it widely. Without addressing data system limitations, organizations risk failing to capture the speed and efficiency gains that agentic AI promises.
Priority Initiatives
The top priority cited by respondents for enabling agent scaling is improving access to both structured and unstructured data. Closely following is enhancing data and AI governance with business context. Data leaders are also placing heavy emphasis on automation of data management.
Conclusion
The report makes clear that successfully scaling trustworthy AI agents requires modernizing data estates specifically for agentic workloads. The experience of data leaders demonstrates that dramatically increasing data accessibility, embedding business context, and automating data operations are essential steps to building trust in agent decisions and removing the primary obstacles to scale and speed.
*This content was produced by MIT Technology Review Insights in partnership with Google Cloud.*