en.Wedoany.com Reported - Pinecone announced that its AI knowledge engine product, Pinecone Nexus, has integrated with Microsoft OneLake, providing enterprise AI agents with a fast response method based on existing data, complete with citations and structured output. This integration transforms the previously slow and expensive retrieval process into rapid, structured responses, with data sourced directly from the Microsoft ecosystem.

During runtime, common AI agents spend most of their time and tokens searching for relevant information from accessible data. They need to retrieve raw data, piece it together, and then send it to frontier large language models (frontier LLMs) for understanding. When scaled in production environments, this process sees a 60% drop in task completion rates, while token consumption and costs become unpredictable.

Pinecone Nexus is a knowledge engine built specifically for AI agents. Instead of having agents assemble raw data at runtime, it completes this work in advance, dynamically building structured, task-optimized contexts called artifacts. Each artifact targets a specific task scope, extracts the correct data, applies appropriate permissions, and formats the results for direct agent use. Agents query artifacts using KnowQL, a query language designed for knowledge retrieval. KnowQL queries specify what the agent needs, the output format, citation requirements, and latency budget. Early data shows that this approach reduces token usage for frontier LLMs by over 95%, improves task execution speed by more than 30 times, and achieves task completion rates exceeding 90%.

For users of Microsoft Fabric, their data—including documents, tables, and Power BI semantic models—is already unified in OneLake. The Nexus integration connects directly to OneLake without manual import steps. When an agent needs to complete a task, Nexus queries OneLake, builds an artifact tailored to that task and the user's permission scope, and returns a structured, cited response via KnowQL. Every answer is traceable to its source, data does not exceed role-based access control (RBAC) permissions, and personally identifiable information (PII) is flagged at ingestion and centrally managed.

Currently, every team building AI agents invents its own retrieval interface, leading to fragmentation. KnowQL defines a common language where agents can specify the question, output structure, citation standards, and latency budget, while the knowledge engine compliant with KnowQL handles the rest. "The data needed to power enterprise AI agents already exists in Microsoft OneLake," said Ash Ashutosh, CEO of Pinecone. "Nexus builds task-specific artifacts from this data and provides AI agents with a clean, structured, cited interface via KnowQL, delivering over 30 times faster speed at a fraction of the cost of traditional retrieval methods." Dipti Borkar, Corporate Vice President and General Manager of Microsoft OneLake and Fabric Ecosystem, stated: "Microsoft OneLake provides a unified data foundation for AI applications and agents. Pinecone Nexus pre-completes the heavy lifting of extracting, assembling, and reasoning over data from OneLake, so our customers' agents spend less time on tool calls, consume fewer tokens, and get accurate answers faster."

Nexus is designed for enterprises with compliance requirements. Artifacts are assembled per task, scoped by RBAC and attribute-based access control (ABAC) permissions, and version-controlled. Every answer cites its source. PII tagging and LLM processing rules can be configured once and applied consistently. Token consumption is tracked across users and workloads via a unified dashboard. Pinecone Nexus integrated with OneLake is now available for early access. Pinecone is the AI knowledge infrastructure provider trusted by over 9,000 customers and 800,000 developers worldwide, with a mission to give AI knowledge.

This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com