Pinecone Launches Nexus Engine: Structuring Business Context for AI Agents

Pinecone Nexus is now generally available as a knowledge engine for AI agents, converting enterprise data into a structured layer directly queryable by agents. Teams ingest and organize business context once, then reuse data across agents, improving accuracy while reducing token costs. Pinecone argues that LLMs handle general knowledge well and vector databases locate specific information, but enterprises rely on business context scattered across contracts, knowledge bases, HR documents, meeting notes, support tickets, and financial records. Having agents search each time is inefficient, causing higher token costs and slower responses. Nexus integrates distributed knowledge into a structured layer, moving token consumption from per-query retrieval to one-time ingestion. Early adopters in legal and financial services saw significant improvements: in legal tasks, Nexus completed all assigned tasks, while coding agents achieved 6% and RAG systems 66%. RAG struggled with doctrine synthesis, cross-case reasoning, and coverage questions that require integrating multiple sources. Token consumption dropped to about 1/9 to 1/15. In enterprise data management, Nexus achieved 90% accuracy vs. 65% for RAG, with a cost of $0.0038 per document. Nexus uses Workspaces as top-level containers, Contexts for specific datasets, Manifests for ingestion rules, and domain experts can embed knowledge via blueprints. Data ingestion supports local files, Box, and Microsoft OneLake; Google Drive, Slack, GitHub, Notion, Confluence, and S3 coming soon. Queries use KnowQL, usable by agents, chatbots, and recommendation systems. Nexus offers a preview experience and a BYOC deployment option for strict data residency and compliance. Competitors include Cognite, RationalAI, and LlamaIndex.