Agentic AI places a fresh set of data demands on enterprises

Techday NZ

09-07-2025

Organisations have invested significantly in data for their generative AI journeys, but may need to put in additional work to meet the demands of a new generation of AI technology.
Generative AI taught organisations a lot about their data: about the preparation, cleanliness, lineage, and quality required to function relatively error-free in the AI era. Or, put another way, it did not take long for language models to expose any shortcomings in an organisation's data management practices, leading many to embark on improvement projects as a starting point for their AI journey.
Many have since done well in creating tangible value and cost savings.
But AI is a fast-moving space, and generative AI is no longer the pinnacle of an enterprise AI journey. That title now belongs to Agentic AI. Acting as a self-directed 'agent', the technology is capable of making informed decisions by drawing on multimodal data and underlying algorithms, which then 'learn' from the outcomes of those decisions.
Even more compelling is Agentic AI's ability to carry out tasks entirely on its own. This capacity to adapt, plan, and deliver complex operations without human direction is what sets it apart from previous AI iterations.
Organisations have not wasted their time with generative AI. They've achieved value, developed skills and capabilities, and set up resources such as data platforms to support their work. While some of this is transferable to Agentic AI, not all structures will be completely fit-for-purpose in this new context.
As a result, organisations may need to review and/or renew their data infrastructure and architecture to support their emerging Agentic AI strategies and plans.
How agent data consumption differs
The efficacy of an organisation's Agentic AI deployments will depend on several factors, not least the data and context the agents rely on to execute their part of a process or transaction.
Task-based agents plan and change actions depending on the context they're given. They delegate subtasks to the various tools available through a process often referred to as "chaining", where the output of one action becomes the input for the next. This means that queries (or tasks) can be broken down into smaller tasks, with each requiring access to data in real-time, processed iteratively to mimic human problem-solving.
To trust an agent to complete sophisticated tasks based on multiple retrieval steps, the value of the data needed to support the decision-making process multiplies significantly compared to generative AI.
Explainable and accurate data retrieval is required at each step of the chain. This is important to provide users with visibility of the data source upon which the decision is based and ensure that every agent output can be analysed, to determine if it was, in fact, the best possible outcome. It's important for agents to learn from any dissatisfactory results they produce, which could point to an underlying problem in the data or its retrieval by an agent.
The utility of knowledge graph technology
Making reliable enterprise data available to agents is key to success with Agentic AI. Graph database technology offers a way of achieving this.
Gartner already identifies knowledge graphs as an essential capability for GenAI applications, as GraphRAG (Retrieval Augmented Generation), where the retrieval path includes a knowledge graph, can vastly improve the accuracy of outputs.
The unique structure of knowledge graphs, made up of 'nodes' and 'edges', is where higher-quality responses can be derived. Nodes represent existing entities in a graph (like a person or place), and edges represent the relationship between those entities – i.e., how they connect to one another. In this type of structure, the bigger and more complex the data, the more previously hidden insights can be revealed. These characteristics are invaluable for presenting data in a way that makes it easier for AI agents to complete tasks reliably and effectively.
It's not only efficient data retrieval where knowledge graphs excel. The technology is also useful in multi-agent environments, where graphs serve as a well-structured memory for large language models (LLMs), to aggregate what they know, what questions they've answered, and what sort of information they've collected, and then use that as an interface to other agents in a system. In other words, the LLMs themselves are leaning on graphs as a form of memory and communication in agentic frameworks.
For organisations with multi-agent AI ambitions, it is important to ensure that their data is as rich, connected, and contextually aware as possible, so it's fully accessible and usable by these intelligent agents. Taking this step helps unlock the true value of your data, enabling agents that are not only more accurate and efficient but also more explainable in their actions.
This is where the integration of Agentic AI with knowledge graphs becomes a genuine game-changer. Connected data provides the deeper context agents need to reason more effectively, produce smarter results, and deliver greater real-world impact.