Autonomous infrastructure unlocks live data for AI agents
Everpure (formerly Pure Storage) is using discovery and classification tools across on-premises, cloud, and SaaS repositories, often tied to a configuration management database like ServiceNow, to give AI agents access to live enterprise data instead of stale copies, VP of product management Chadd Kenney said at Pure Accelerate 2026 on theCUBE. The approach extends Everpure's FlashArray and FlashBlade storage into a data-intelligence layer that builds a "knowledge map" of where data lives, so agents can reason across systems rather than being limited to a single application's view, like Salesforce alone. The conference also introduced Pure DataStream, a pipeline product co-designed with Nvidia that automates data flow from ingestion to inference. Everpure has doubled adoption of its Fusion control plane, from 600 to 1,200 customers quarter over quarter, an early indicator of whether the broader autonomous-data vision can scale.
Enterprise AI agents are only as good as the data they can see, and most enterprise data still lives fragmented across SaaS apps, on-prem systems, and cloud repositories never designed to be queried together. Everpure's pitch, discovery and classification instead of forced centralization, reflects a broader industry bet that agentic AI will fail on data access before it fails on model quality, which is worth tracking even accounting for its origin as vendor conference messaging.
What happened
At Pure Accelerate 2026, Everpure Inc. (the company formerly known as Pure Storage) vice president of product management Chadd Kenney told theCUBE that the company uses discovery and classification across on-premises, cloud, and SaaS repositories, typically tied to a configuration management database such as ServiceNow, to map data endpoints and build a contextual "knowledge map." Kenney said this lets AI agents reason over real-time data across systems rather than being limited to whatever one application, like Salesforce, exposes: without broader context, he said, an agent "would have to infer what the costs are and maybe just make up what would be profitable or not." The approach extends Everpure's FlashArray and FlashBlade platforms from raw storage into a data-intelligence layer.
Technical context
Independent analysis from ECI Research describes the broader strategy Everpure unveiled at the same conference: a three-layer Everpure Enterprise Data Cloud (a unified data plane, an autonomous-governance control plane, and a "Universal Data Intelligence" layer for classification), plus Pure DataStream, a data-pipeline product co-designed with Nvidia using NeMo for curation and RDMA networking, intended to automate the path from raw enterprise data to LLM serving without copying data into a new silo. ECI Research also reported that Everpure's Fusion control plane, the component that enables fleet-wide policy enforcement, doubled its customer base from 600 to 1,200 quarter over quarter, though the company has not yet solved governance for the semantic metadata its own agents generate.
For practitioners
This account comes from a sponsored theCUBE interview at a vendor's own conference (disclosed as a paid media partnership) plus vendor-conference analyst coverage, not independent benchmarking, so treat specific performance and adoption claims as vendor-reported until third parties validate them. The underlying architectural pattern, runtime discovery and classification instead of upfront ETL into a centralized lake, is a legitimate and increasingly common approach for agentic systems, but it requires reliable SaaS connectors, consistent metadata extraction, and clear provenance and access-control surfaces for downstream agents.
What to watch
Key adoption signals include whether Everpure hits its stated goal of getting half its customer base onto Fusion this year, how Pure DataStream's Nvidia co-design performs outside of vendor-selected examples like the reported one-hour Legal Assist build, and whether Everpure or a security partner produces a credible answer to governing the semantic metadata its own agents create, a gap ECI Research flagged as unresolved.
Key Points
- 1Everpure, formerly known as Pure Storage, uses runtime data discovery and classification to give AI agents live cross-system context instead of siloed application views.
- 2The company's new Pure DataStream pipeline, co-designed with Nvidia, aims to automate enterprise data flow from ingestion through inference without creating new data silos.
- 3Adoption of Everpure's Fusion control plane doubled to 1,200 customers quarter over quarter, an early signal of whether its autonomous-data vision can scale.
Scoring Rationale
Originally scored as a general infrastructure trend piece without weighing that its sole source was a disclosed paid theCUBE conference interview at the vendor's own event; pulled down accordingly. Kept in the 'solid' band rather than lower because independent ECI Research analysis corroborates concrete, measurable signals (Fusion adoption doubling, a technically credible Nvidia co-design) beyond the vendor's own framing.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
