Teradata (NYSE: TDC) on Tuesday unveiled an expansion of its Tera product into a full AI agent system designed for enterprise data workloads, with components scheduled to roll out in the fourth quarter of 2026.
The company said the system is built to operate across multiple data platforms, not solely within Teradata’s own environment. It comprises three components: a context and orchestration layer, an execution engine, and a set of pre-built agent skills.
The Tera Context Engine links databases, data platforms and catalogs without moving data between systems, according to Teradata. It incorporates Industry Knowledge Models containing sector-specific terminology and relationships, applies governance controls and tracks data lineage.
The Tera Harness serves as an execution layer that coordinates workflows across different tools and models. Built on a Go-native engine, it supports concurrent operations and includes checkpointing capabilities so long-running tasks can be paused and resumed.
Agent Skills are divided into Platform Agents, which handle operational tasks such as workload tuning and compute sizing, and Analytics Agents, which translate natural-language requests into SQL and Python code.
Teradata also provided benchmark results comparing its system against rival offerings. On SWE-bench Pro using Opus 5, the company said Tera used 73% fewer tokens than Claude Code, completed work 42% faster and cost 58% less. On data-eng-bench, Teradata reported a 53% lower cost per task compared with Snowflake Cortex Code using the same model.
The vendor will offer Teradata AI Services to assist customers with deployment, implementation support and configuration of Industry Knowledge Models. No pricing was disclosed.
Teradata is based in San Diego.
