OpenText this week previewed an artificial intelligence platform that promises to make it simpler for enterprises to use their data to build artificial intelligence (AI) agents that have the context needed to successfully automate workflows.
Announced at the OpenText World 2025 conference, the OpenText AI Data Platform (AIDP) is based on version 26.2 of the core OpenText Platform that is scheduled to become available next June. It enables organizations to expose AI agents to structured and unstructured data in a way that can be centrally governed and secure to ensure compliance mandates are met.
Additionally, OpenText revealed it has allied with Databricks to provide organizations with access to an external data lake in addition to expanding a cloud alliance with Google.
Savinay Berry, executive vice president, chief product officer, and CTO for OpenText, told conference attendees that at the core of the OpenText AIDP platform is Aviator Studio, a framework for both building and deploying agents. In effect, OpenText is now providing a platform for agent software lifecycle management, he said.
OpenText AIDP builds on a knowledge graph and security controls that OpenText previously built to ensure that AI agents stay focused on specific tasks. The goal is to not only protect AI agents using a set of guardrails that thwart, for example, prompt injection attacks, but also ensure that rogue AI agents don’t access data they have not been given permission to employ.
OpenText is also adding an Aviator Eval Ops tool that leverages large language models (LLMs) to evaluate the output generated by AI agents that organizations create using OpenText AIDP.
Finally, OpenText is adding AI Aviator Services, a professional services team that organizations can tap to help build and deploy AI agents.
In general, organizations have anywhere from 100 to 150 zettabytes of data, most of which has not been exposed to an LLM, says Berry. That compares to about 10 to 15 zettabytes of data that has been used to train general-purpose LLMs. AI agents that will be relied on to automate tasks in enterprise IT environments will need to be able to access data that today resides largely behind a corporate firewall, he notes. The OpenText AIDP will enable organizations to take advantage of AI agents that both they and OpenText will use to develop domain-specific AI agents, adds Berry.
Just as importantly, the OpenText AIDP will also enable organizations to leverage a single platform to contain AI agent sprawl that will inevitably become an issue if organizations rely on a more decentralized approach to build and deploy enterprise-class AI agents, he notes.
It’s not clear at what pace and scale organizations are building and deploying AI agents, but without data there is not enough context being provided, says Berry. “Without context, AI fails,” he says.
The one thing that is certain is that far more organizations lack the data management best practices needed to successfully build, deploy and secure AI agents. The issue then becomes determining how quickly they can impose order on data management practices that, historically, have never been all that consistent.
