The challenge with AI agents isn’t that they make mistakes. It’s that they can make the same mistake across a business, instantly and at scale.
Enterprises aren’t waiting on agentic AI. It’s already in production, with agents operating inside real business workflows across industries. But most organizations are building on a foundation that can’t support it. Gartner estimates that more than 40% of agentic AI projects will be canceled by 2027, often because companies moved too quickly into use cases their data and infrastructure couldn’t reliably support.
Our latest research, “Agentic AI readiness index 2026,” quantifies the gap. Nearly 60% of enterprises report investing millions in agentic AI, and 41% are already running it in production. When asked what’s holding them back, data quality and lineage top the list at 42%, followed by sovereignty and regulatory compliance at 39%, and security and privacy at 39%. Talent, strategy, and skills all rank lower.
Taken together, 85% lack the data foundation needed to support agentic AI at scale, and it’s showing up in their outcomes.
[CTA_MODULE]
This isn’t a model problem. It’s a data problem.
Only 15% of organizations say they are fully prepared for agentic AI today, and among these organizations, nearly all (98%) report strong confidence in agentic AI ROI. Among the least prepared, that number drops to 16%.
The companies seeing strong ROI from agentic AI are operating differently. Fully prepared organizations have 4 things in place:
- Automated data movement that keeps information current
- Lineage that shows where the data came from and how it changed
- Interoperability that lets data move across systems without getting trapped
- Governance that defines what agents can access, what they can do, and where human review still matters.
The result is a centralized, governed foundation, where data is consistent, traceable, and trusted across the business.
Agents don’t catch what people catch
People check dashboards, question numbers, and notice when something looks off. Agents don’t do that. They operate on whatever context they have. When a system runs autonomously, it doesn’t pause to question the data; it reads what’s there and acts on it instead.
An agent powered by low-quality or poorly governed data doesn’t improve over time; it simply makes incorrect decisions faster. What used to be small errors — a broken integration, a missing field, a conflicting definition — become system-wide failures.
So the real question is whether the data foundation underneath them can support autonomous operation at scale. For 85% of enterprises, it can’t.
What Open Data Infrastructure changes
An Open Data Infrastructure brings data out of siloed applications into a governed data lake foundation, a centralized source of truth where structured and unstructured data is stored at scale. From there, data can be consistently defined, connected, and accessed by downstream compute and AI systems across the business. Agents and analysts operate from the same complete, current, and trusted view of the business. Data moves through the system without duplication across multiple warehouses and without relying on fragile connections that break under pressure.
Governance is a core part of this. Agents need access to data, but they also need clear boundaries on what they can see, what they can do, and which decisions require human review. That is not just a compliance issue. It is becoming a condition of participation. Sixty-five percent of organizations say they would either heavily restrict or outright block vendors that cannot meet governance and sovereignty requirements. Once agents move into production, governance becomes the control layer that determines where autonomous systems are allowed to operate at all. Open Data Infrastructure defines those boundaries once and enforces them across connected systems so agents can operate reliably instead of guessing.
Without this foundation, AI delivers pockets of value. Individual use cases work, and some processes get faster. But it doesn’t scale, and it doesn’t change how the business runs. With an Open Data Infrastructure, AI becomes part of the operating model instead of just a productivity layer on top.
The companies that get the most from agentic AI are the ones that have built the data foundation to support it. That’s what separates the organizations seeing strong ROI from the ones still waiting to see it.
[CTA_MODULE]
