Teradata aims to make agentic execution of multistep data work more efficient
73% fewer tokens. Teradata's new execution layer cuts the hidden cost of agentic workflows — and makes agent budgeting predictable for the first time.
Why it matters
Teradata is shipping agent cost-control infrastructure (Tera Harness) that optimizes inference efficiency by planning before LLM calls and pruning unnecessary steps. This addresses a real pain point: agentic workloads are unpredictable cost-wise and prone to wasteful reasoning loops. Enterprises scaling agents now have a platform lever beyond just picking cheaper models.
The key facts
7 to knowTera Harness: execution-planning layer that batches tasks and drops non-advancing model calls before inference
73% fewer tokens than Claude Code on SWE-bench Pro; 42% faster task completion; 58% lower total cost on same Opus 5 model
Applies 84 execution patterns before inference to reduce repeated LLM reasoning
Cuts unnecessary model and tool calls while preserving business context
GA target: December 2026
Part of Teradata's Autonomous Knowledge Platform (Tera, introduced May 2026)
Key tradeoff: aggressive pruning can drop steps that matter; requires careful result verification and shift to outcome-definition work for developers
Go to the source
CIOcio.com
Publisher excerpt: Teradata is adding a context engine, an execution layer, and reusable agent skills to Tera, its AI-powered workspace for enterprise data and AI tasks, in order to make agentic execution of multistep workflows more efficient. Tera was initially introduced in May as part of Teradata’s Autonomous…