Teradata this week revealed it will deliver in the third quarter a platform designed specifically to enable artificial intelligence (AI) agents to safely discover and access data at scale across highly distributed computing environments.
The Teradata Autonomous Knowledge Platform has been designed from the ground up to unify the management of structured and unstructured data in a way that provides the business context AI agents need to repeatedly sense, decide, and act reliably, says Sumeet Arora, chief product officer for Teradata.
Rather than requiring organizations to add a separate overlay to achieve that goal, Teradata is embedding those capabilities directly into its data management platform, he added.
The core platform provides AI agents with a workspace, dubbed Tera, to access execution environments that are invoked via a natural language interface. Tera supports multiple modes for data analysis using Tera Analyze, coding with Tera Code, and multi-agent system automation and orchestration using a forthcoming Tera Claw tool. Organizations can also make use of a Teradata AI Studio tool to build, activate and govern AI agents.
There is also a set of pre-built agents that perform tasks such as managing infrastructure or driving operational efficiency to ensure cost optimization using the Teradata Cloud platform.
Finally, the company later this year plans to make available a turnkey Teradata Factory option for deploying the Teradata Autonomous Knowledge Platform in an on-premises IT environment to ensure strict data residency and regulatory requirements are maintained.
At the core of the Teradata Autonomous Knowledge Platform is the massively parallel processing (MPP) architecture the company developed, which in turn is layered on top of a unified block and object storage platform that supports both the Apache Iceberg and Delta Lake formats. As a result, every data and model interaction can be governed via the metadata that Teradata already tracks.
That approach is fundamentally different from other agentic AI platforms because the data fabric that provides the guardrails for governing AI agents is built into the platform, says Arora. “There’s no need to waste time moving data around,” he says.
In effect, the knowledge an organization has can now be instrumented in a way that enables an AI agent to safely act on it, adds Arora.
It’s not clear to what degree organizations will be revisiting the platforms used to process and analyze data in the age of AI. A recent Futurum Group report, however, forecasts the global data intelligence, analytics, and infrastructure (DIAI) market will grow at a 17% compound annual growth rate to exceed $1.2 trillion by 2031.
The one thing that is certain is that data is going to be the key to any successful agentic AI initiative. After all, the more effort there is to ensure that only data of the highest quality is exposed to AI agents, the less likely it becomes there will be an unexpected outcome. Of course, given the probabilistic nature of any AI agent, there will always be a need to validate the output, so each organization will ultimately need to determine what degree of actual autonomy they want to provide to their AI agents.
