Nexus integration
Pinecone has integrated its Nexus product with Microsoft OneLake, linking its AI knowledge engine with data stored in Microsoft's Fabric environment.
Agent querying
The integration is intended to let AI agents query enterprise data in OneLake and receive structured responses with source citations, rather than assembling raw information at runtime.
Pinecone said many AI agents now spend most of their processing effort locating relevant information across the data they can access. It argued that this approach can become costly and less reliable at scale because systems retrieve raw data, combine it, and then pass it to a large language model for interpretation.
Data assembly
With the new integration, Nexus connects directly to OneLake without manual imports or uploads, Pinecone said. When an agent submits a task, Nexus queries data in OneLake, assembles what Pinecone calls an artifact for that task, applies user access controls, and returns a formatted response through its KnowQL query language.
The arrangement is designed to move more data preparation upstream. Pinecone said the system builds task-specific contexts in advance so AI agents can work from pre-assembled material instead of repeating retrieval and reasoning for each request.
The company positioned the product as a way to address a broader problem in enterprise AI deployments. Teams often create their own retrieval methods for different agents, Pinecone said, which can lead to inconsistent governance, fragmented interfaces, and weaker cost control across systems.
KnowQL is Pinecone's proposed common interface for these requests. An agent can specify the information it needs, the output format, the citation standard, and the latency budget, with the knowledge engine handling retrieval and assembly, according to the company.
Pinecone also made performance claims tied to early use of Nexus, saying customers had cut frontier large language model token use by more than 95%, run tasks 30 times faster, and raised completion rates above 90%.
Microsoft Fabric users already store a range of enterprise data in OneLake, including documents, tables, and Power BI semantic models. The integration is meant to use that existing data foundation rather than require organisations to rebuild pipelines or maintain separate retrieval systems for AI workloads.
Access control and governance are central to the pitch. Pinecone said artifacts are created for each task and limited by role-based and attribute-based permissions, while personally identifiable information is tagged on ingest and governed centrally.
Each response also links back to its original source, Pinecone said. Token use across users and workloads is tracked through a single dashboard.
Executive statements
"[The data enterprises need to power their AI agents already live in Microsoft OneLake]," said Ash Ashutosh, Chief Executive Officer, Pinecone. "Nexus builds task-specific artifacts from this data, and gives AI agents a clean, structured, cited interface through KnowQL, 30x+ faster and at a fraction of what traditional retrieval approaches cost."
For Microsoft, the integration adds another way for customers to apply AI systems to data already consolidated in Fabric. OneLake serves as the company's unified data layer across analytics workloads, and Pinecone's link is intended to make that data easier for AI agents to use under existing governance rules.
"Microsoft OneLake offers a unified data foundation for AI applications and agents," said Dipti Borkar, Vice President and General Manager, Microsoft OneLake and Fabric Ecosystem. "Pinecone Nexus does the hard work of fetching, assembling, and reasoning over OneLake data up front, so our customers' agents spend less time making tool calls, burn fewer tokens, and get accurate answers faster."
Pinecone said more than 9,000 customers and 800,000 developers use its products worldwide.