Teradata adds agentic, multimodal tools to vector store

Teradata has added agentic and multimodal capabilities to its Enterprise Vector Store, expanding support for unstructured content such as documents, images, and audio alongside structured enterprise data.

The update tightens integration with Unstructured, a specialist in processing unstructured data. Teradata said the combined approach reduces reliance on multiple point products and streamlines preparation of content for retrieval and agent workflows.

What is new

Teradata Enterprise Vector Store now includes automated ingestion and processing for documents, PDFs, images, and audio, with video support planned. It also adds hybrid search, combining semantic and keyword search with metadata-based techniques.

The release expands support for multimodal embeddings across text, images, and audio. It also increases supported embedding dimensions to 8K, which Teradata says can improve retrieval quality for complex queries.

The update also adds LangChain integration, which Teradata positions as a way for developers to connect agent workflows with data in the vector store and the broader platform. Teradata also references LangGraph in examples of orchestrated workflows.

Why it matters

Enterprises are pushing generative AI systems into production while struggling with fragmented data estates. Unstructured content such as PDFs, chat logs, audio recordings, and images often sits outside core analytical systems. That separation can limit what AI tools can access and add complexity when teams build retrieval-augmented generation systems and AI agents.

Teradata cited Gartner research estimating that unstructured data is growing three times faster than structured data. It also pointed to external research that it said shows many companies are deploying AI agents and forecasting strong returns, while still facing barriers such as data silos, scalability, and unified access to content.

Vector databases have become a common component for AI retrieval, but large organisations often require governance, access controls, and deployment flexibility across cloud and on-premises infrastructure. Teradata is positioning Enterprise Vector Store as a vector layer that fits within its broader data platform and governance model across hybrid environments.

Scale and governance

Teradata said its vector store can ingest millions of documents and process thousands of files per hour, depending on configuration and data characteristics. It also said the system has scaled to billions of vectors and supported more than 1,000 concurrent queries without performance degradation.

These claims matter for organisations applying AI tools to large internal data collections, especially where audit trails and controls are needed when agents take actions based on retrieved information.

Forrester has noted that supporting tens of billions of vectors at high scale remains difficult. Teradata points to that context to argue it can differentiate on scale and operational characteristics, particularly for customers already running its data warehousing technology.

Use cases

Teradata outlined examples in regulated and data-heavy industries. In insurance, it described claims adjudication agents that process damage photos and policy PDFs alongside structured claims data. The workflow extracts information from images and documents, then cross-references coverage rules and claim history. Teradata said the result is faster, explainable decisions that support audit compliance.

In financial services, Teradata pointed to loyalty and discount eligibility queries that require both policy text and structured transaction or customer data. It said firms can build governed agents that combine unstructured policy definitions with structured business data to answer questions that SQL alone cannot.

Healthcare was another example, using visual question-and-answer scenarios that combine patient records with clinical notes, medical images, and audio dictations. Teradata described an orchestrated workflow using vision models, multimodal vector search, and grounding in trusted documentation.

Teradata also highlighted defence intelligence scenarios, including processing images of camouflaged assets alongside terrain patterns and threat signatures, with agents providing guidance based on retrieved context.

Executive comments

Teradata framed the release as part of a broader shift toward agents as a front end for enterprise information systems.

"We're entering an era where AI agents will become the primary interface for enterprise intelligence-autonomously orchestrating workflows, making decisions within defined governance frameworks, and uncovering insights across every data type," said Sumeet Arora, Chief Product Officer, Teradata.

Unstructured said the integration keeps data processing inside Teradata's environment.

"Enterprises shouldn't have to choose between data security and AI readiness. By embedding Unstructured natively inside Teradata Enterprise Vector Store, Teradata customers get production-quality, AI-ready data at scale, with no external tools, no data leaving the platform, and no compromise on governance," said Brian Raymond.

The new agentic and multimodal capabilities for Teradata Enterprise Vector Store are due to become generally available to Teradata customers from April 2026.