The Pulse
Dell Adds a Semantic Layer and Knowledge Graph for AI Agents
Dell announced a Unified Semantic Layer, Enterprise Knowledge Graph and Knowledge Agents for its AI Data Platform on October 6, 2026. The three features are scheduled for release in the first half of 2027, while separate storage and data-pr
AI.info Team · Written with AI ·
Dell says enterprise AI agents often have access to company data but lack the context to use it reliably; its proposed fix will not arrive until 2027. The company announced a Unified Semantic Layer, an Enterprise Knowledge Graph and Knowledge Agents for the Dell AI Data Platform on October 6, with all three scheduled for release in the first half of 2027. The gap between the problem Dell identifies and the delivery date leaves customers waiting for the central additions to its pitch.
Dell wants agents to share a definition of the data
The Unified Semantic Layer is designed to give structured and unstructured information consistent business meaning through rules, definitions and a searchable glossary. Dell’s example is a company where one system calls a customer a “client” and another calls the same person an “account”: the layer is meant to help AI treat those terms as equivalent. Customers can also import existing ontologies and classification taxonomies, and Dell says it is enabling NVIDIA Auto-Ontology, an open-source library for building knowledge graphs from enterprise data.
The Enterprise Knowledge Graph maps relationships among data and uses metadata, lineage and query history to adjust as activity changes. When an agent asks a question, Dell says, the platform can gather related tables, data products, multimodal information and vector indexes, subject to the agent’s access rights. In the company’s manufacturing example, an unusual sensor reading could be connected to a machine, its repair history, a supplier batch and orders that might be affected.
Knowledge Agents sit on top of that context. Dell describes each as an adviser for a particular topic, grounded in a defined portion of the graph; customers set its instructions, data permissions, quality requirements and spending limit. The company says the three capabilities will remain in the customer’s data center, rather than requiring sensitive information to leave the platform.
Lewis says access is not the same as understanding
“Data without context is just noise. Most enterprises have spent years making their data accessible. That's not the same as making it usable,” Arthur Lewis, president of Dell’s Infrastructure Solutions Group, said in the announcement. His statement captures the distinction Dell is drawing: search and retrieval can locate a record, while a shared vocabulary and mapped relationships are meant to help an agent interpret why it matters.
Dell’s release does not provide customer deployments or independent results for the semantic and graph features, which remain future releases. The platform already uses NVIDIA Nemotron Retriever models for document parsing, embedding and reranking, and NVIDIA cuVS for vector indexing and search, according to Dell. The new layers extend that existing data path rather than replacing the company’s storage and orchestration components.
Faster processing and storage changes arrive first
Dell also announced updates with nearer availability. Its Data Processing Engine enhancements using NVIDIA’s acceleration stack are due in December 2026; further acceleration using Apache Arrow is expected in the first half of 2027. Dell says PowerScale security and multitenancy updates, including support for as many as 500 tenants in one cluster, are scheduled for November 2026. Its Storage Performance Tool and AI-ready data services are available now.
The performance figures are based on Dell’s internal September 2026 tests, not an independent benchmark. On a Dell PowerEdge R770 with NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs, Dell compared GPU-accelerated and CPU-only Apache Spark runs and measured a 3.9-times average speedup, with a 20.4-times peak on a batch data-mining workload. The company says the tests used default configurations without performance tuning and cautions that actual results may vary.
For customers, the announced schedule separates the platform’s main context features from updates they can use sooner: the graph, semantic layer and agents are planned for the first half of 2027, while the storage benchmark tool is available now. Dell’s claim is that shared definitions and relationships can reduce the repeated work agents do to reconstruct business meaning; the release date, rather than a customer result, is the concrete next milestone.