“…greater than 80% of IT leaders believe their enterprise data is not ready for agentic AI…” – Gartner, from “Innovation Guide for AI Agents”
Agentic AI holds great promise, but as this quote from Gartner makes clear, success has been uneven, as 80% seem to be struggling. Why is that? Agentic AI has some special requirements, beyond some of the more typical needs of technology implementations. In this post, I’ll cover these requirements and the best way to address them.
What AI Agents Need
AI agents live in real time. Similarly, operational/transactional systems create data in real time as the business is conducted. A point-of-sale system records transactions as a customer buys goods, while reducing the available inventory of those goods in the inventory management system. Such systems generate and manipulate live data, and AI agents need to be able to operate on this data; otherwise, they will leverage older, out-of-date data for their decisions. Live data is the key ingredient that can make AI agents most informative, helpful, and successful.
AI agents also need a certain amount of enterprise context, so that they can fully understand, and can act on, all available data. When semantics differ across different systems, even the most accurate data can be misleading for an AI agent, causing them to make less-than-ideal decisions or hallucinate outright.
Finally, AI agents also need guardrails, to keep them from acting out in erratic, disruptive, or counter-productive ways, or even potentially go rogue. Like any data-centric initiative, they need effective governance and security controls that keep them in check across all applicable data sources.
How Best to Deliver These Capabilities?
The most flexible way to deliver them is through a logical data management layer, as such layers operate on real/live data directly in the operational systems. Logical data management enables the provisioning of live data to AI agents with contextual semantics and safeguarded by data governance and security.
A logical data management layer can use data agents to intelligently deliver the right data, at the right time, to the right AI agent, enabling effective multi-agent systems (MAS) synchronized by sophisticated Agent2Agent (A2A) protocols.
Logical Data Management in Action
I’d like to share a powerful example with you. A financial regulator for a top-10 global financial center leveraged logical data management to enable real-time compliance with agentic, AI-powered, autonomous oversight. This regulatory agency deployed agentic AI to continuously monitor financial activity across 60+ financial institutions and other government agencies, automating audits and ensuring real-time compliance alerting and remediation. As a result, the company is now being recognized as a global regulatory innovator.
The Necessary Ingredient for Successful Agentic AI Projects
Live data is the key ingredient for successful agentic AI, but delivered with semantics and governance. Logical data management can deliver this, for the best chances of success with agentic AI. For more information about logical data management, see O’Reilly’s new book, The Rise of Logical Data Management – An Essential Data Strategy for Transforming Your Business in the Age of AI.
- How to Get Your AI Agents to Do Your Bidding (and Keep Them from Going Rogue) - February 19, 2026
- Why Many Leading Analyst Firms Named Denodo a Leader Over the Years? - December 11, 2025
- O’Reilly Releases First Chapters of a New Book about Logical Data Management - January 21, 2025