Starburst’s New Platform Aims to Close AI’s Biggest Gap — the Data Layer

As companies scramble to deploy AI in production, a new dynamic has emerged: data infrastructure is now the bottleneck. Models are only as good as the information they can access, and most organizations still cannot close the gap on making their data available, governed, and fast enough to keep up.

Fragmented systems scattered across a range of clouds and regulatory zones make that connection slow and unreliable. For organizations, progress depends on the ability of data platforms to intelligently bring governance into place without breaking compliance or repeating the same work.

That set the context for Starburst’s reveal at AI & Datanova 2025. The company announced an AI-ready data platform that extends its lakehouse foundation to include multi-agent workloads and vector-native capabilities. This is part of Starburst’s vision for what it calls the Agentic Workforce—a model where humans and AI agents work side by side to reason, decide, and act across data-driven workflows.

The approach combines several trends redefining enterprise AI: model-to-data access, which lets agents query information where it resides; federated governance that keeps control local; and new visibility tools tracking how models are used across an organization. Starburst’s move is part of a broader shift away from focusing on building models and instead toward strengthening the data layer that determines whether AI can run safely and at scale.

Starburst is presenting this new release as more than another feature update. It calls it a reimagining of how data infrastructure enables AI at scale. Instead of building new tools apart from existing systems, the company is returning to its roots and focusing on the foundation that allows data and intelligence to connect more easily. The result is a scale-up of its lakehouse architecture that includes features allowing enterprise data to be directly usable by AI systems.

At the center is model-to-data access, which allows AI agents to ask questions of governed information where it lives. This helps minimize duplication, reduce latency, and maintain control over privacy and compliance in complex, distributed environments. Multi-agent interoperability comes with early support for the rising Model Context Protocol (MCP), which enables separate AI agents to share context and perform tasks within that governed data layer.

A third addition, open vector access, links Starburst with Iceberg, PGVector, and Elasticsearch for retrieval-augmented generation across structured and unstructured sources. This convergence enables AI to draw from both enterprise-level and contextual knowledge, creating more meaningful insights without sacrificing governance or transparency. Combined, these capabilities position Starburst’s platform as an architecture built for a new era of enterprise-scale, data-aware AI.

Matt Fuller, VP of AI/ML Products at Starburst, told BigDataWire that the company wants enterprises to “rethink how data architecture serves AI by minimizing data movement, and bringing compute to the data, not the other way around.” He explained that solving data fragmentation has been part of Starburst’s design from the beginning. “Instead of centralizing everything in a single warehouse, Starburst’s federated query engine lets AI workloads access governed data directly across clouds, regions, and on-prem systems without replication,” he said.

“For AI use cases, that means models and agents can query, enrich, and retrieve the information they need from distributed sources through governed data products,” Fuller shared. “It reduces latency, improves efficiency, and ensures compliance when working with sensitive or regulated data.”

(nuddss/Shutterstock)

When asked how Starburst fits within the growing field of ‘AI-ready’ platforms, Fuller told BigDataWire that Starburst “enables model-to-data orchestration, letting AI, analytics, and agents query governed data in place, without replication or movement.” He added, “We’re the control plane for intelligent systems, ensuring every model, agent, and user operates from the same governed source of truth.”

Starburst is also turning its attention to what happens after the data connection is made. The company added new monitoring features so teams can see how AI models are being used, set limits on activity, and keep an eye on costs as projects scale. Every interaction can be tracked through dashboards that make it easier to spot issues early and stay within policy.

The agent itself has grown more visual. It can respond not only with text but with charts or graphs that help people understand the story behind the numbers. That small shift pushes the platform closer to everyday business use, where clarity often matters more than complexity.

Starburst has also put extra weight on compliance. Its policy engine uses metadata to keep track of lineage and meet standards such as GDPR and Schrems II while keeping data under local control. The system is built on Trino and Apache Iceberg as part of what the company calls its Lakeside AI architecture. Starburst said the new capabilities are expected later this year, marking another step in the long effort to bring trust and intelligence into the same space.

Building Intelligence into the Database Layer

Voltron Positions Data Flow as the Next Frontier in AI Performance

The post Starburst’s New Platform Aims to Close AI’s Biggest Gap — the Data Layer appeared first on BigDATAwire.