Why Your Data Lakehouse Is the Core of Your AI Strategy

The future of enterprise AI is here. Business adoption of AI isn’t a question of if, or even when, but how. For AI to succeed at scale, it needs to be rooted in the core foundation of modern business–data. As enterprise AI moves from experimentation to reality, the stakes have never been higher. Organizations are seeking solutions that turn ideas into action and innovation into production. 

At Starburst, we believe AI is best realized using the Icehouse architecture, built on Apache Iceberg and Trino. This versatile foundation unifies data access, collaboration, and governance, bringing your AI to your data.

Starburst AI features

Today, we are proud to announce the release of Starburst AI features: 

• Starburst AI Agent 
• Starburst AI Workflows
• Starburst AI Assembly Line

It’s a monumental release, and we are immensely proud of our Engineering and Product teams for the hard work they’ve put into making this a reality. The result is an AI platform designed to empower enterprise AI with governed access to real data, both on-premises and in the cloud, across hybrid environments. 

Let’s explore how Starburst makes enterprise AI a reality, leveraging an approach that we call Lakeside AI. 

 

Bring AI to your data, using Lakeside AI

Why look to the Data Lakehouse, powered by Icehouse architecture, for AI? The Starburst approach to AI architecture is based on continuity with the data architecture that already powers our analytics and application workloads—the Starburst Icehouse architecture. 

This is rooted in a single truth. Data migration is difficult, and AI can’t wait. 

Why AI can’t wait for a data migration

Data migration has always been costly and time-consuming, and that’s just as true of analytics as it is of AI. The reality is that most organizations can’t adopt AI overnight. Meanwhile, AI timelines aren’t waiting. By leveraging the data lakehouse, AI becomes achievable and deliverable. 

This is exactly what Lakeside AI is designed to achieve: AI workloads built on a robust, Open Data Lakehouse foundation. 

Introduce AI to your data

For Lakeside AI, we did something unexpected. We leveraged data lakehouse architecture to build the foundation for tomorrow’s AI. 

Using this strategy, your existing data stack can adapt to AI, not the other way around. The approach is designed to meet our customers where their data lives. By adopting Lakeside AI, enterprises can leverage their existing data stack to support future AI workloads. 

Enter the Starburst AI Agent. 

The Starburst AI Agent: Bringing Agentic AI to the Enterprise

Agentic AI is changing the way that businesses operate. Tomorrow’s AI agents will do more than assist. They’ll be an integral part of how enterprises run their core business.

To meet this moment, Starburst is launching the Starburst AI Agent, a natural language interface that accelerates insights and performs agentic tasks across multiple domains. With the Starburst AI Agent, enterprises can build and scale AI applications faster, with reliable performance, lower cost, and greater confidence in security, compliance, and control. 

Powered by Lakeside AI

This entire approach is built using Lakeside AI. By leveraging the power of Icehouse architecture, Apache Iceberg becomes the backbone for secure, scalable AI agents and applications. 

What can you do with the Starburst AI Agent?

The Starburst AI Agent extends the capabilities of the Starburst platform by embedding natural language interaction directly into the data product lifecycle. This functionality enhances data access, collaboration, and governance at every level. It enables intuitive data exploration and generates reusable assets that support downstream applications, creating an accessible flywheel for data product insights.

Let’s look at the AI Agent workflows one by one.

Create data products using AI

Starburst AI Agents simplify the process of creating and documenting data products, saving both time and effort. Users can interact with the Starburst AI Agent to produce structured data product documentation by combining catalog metadata with descriptive input. The agent supports LLM-assisted drafting and revision through conversational or direct editing, ensuring consistency.

Generate AI-Powered data insights

Once data products are defined, users can query them using natural language. To do this, the agent translates the input into SQL, executes the query, and returns the results in an insight-rich format. Users have complete transparency into the generated SQL and can iteratively refine prompts to explore data more deeply. 

Build Agentic workflows using data products 

Data products created with AI Agents are not static. They can be reused as building blocks across a wide range of downstream use cases, including by external agents and automated workflows. This makes them a central, governed component of broader AI and analytics initiatives.

Build your own Air-Gapped AI Agent

Just like other Starburst workloads, with AI, optionality is always yours. Agentic AI workloads are designed to draw from on-premises, cloud, and hybrid data sources as needed. This means that whatever data stack you use, it can be AI-ready with Starubrst. 

Combined with the ability to access and ingest structured and unstructured data and store it in Iceberg tables, this feature unlocks an abundance of possibilities while helping to meet enterprise data where it already lives. 

Today, customers in highly regulated industries are particularly excited about using Starburst as the core data delivery platform for AI agents, even in fully air-gapped, on-premises environments with custom models and local data. 

Beyond the Starburst AI Agent

The built-in Starburst AI Agent is powerful, but we’re going even further. With the new AI Workflow features detailed below, we’re unlocking something even more transformative: the ability for customers to build their own AI agents, tailored to their data, use cases, and governance needs.

 

What are Starburst AI Workflows?

Starburst AI Workflows is a new suite of capabilities designed to accelerate AI from experimentation to production by making governed, proprietary data instantly usable. Integrated with the Starburst platform itself, AI Workflows combine vector search, SQL-based Large Language Model (LLM) functions, and model access governance—all without the need for pipelines or data movement. 

Starburst AI Workflows are designed to address the key challenges enterprises face when scaling AI initiatives, particularly in terms of access, usability, and control. To do this, they bring critical capabilities into a unified, governed platform that supports enterprise-grade AI delivery in three key ways:

Starburst AI Search

Starburst AI Search enables enterprises to convert structured and unstructured data into vector embeddings stored in Iceberg tables, making them ready for AI use. 

This approach enables AI agents to retrieve relevant insights based on human-level semantic meaning, not just keywords. Overall, AI Search enhances the discoverability of unstructured data by making it available to analytic insights. 

Starburst AI SQL Functions

Starburst SQL Functions brings the power of generative AI directly into SQL, enabling analysts to apply classification, sentiment analysis, translation, data masking, and more to unstructured text without leaving their workflows. 

Using this feature, teams can analyze and transform raw text faster than ever, all without the need for data movement. A built-in prompting function offers complete control, allowing users to guide LLMs with custom prompt engineering that can match the most complex, domain-specific tasks. 

It’s a flexible, secure way to unlock AI-driven insights—natively, at scale, and in SQL.

Starburst AI Model Access Management 

Starburst AI Model Access Management provides a framework for managing access to proprietary and third-party AI models within a robust data governance framework. Available models include: 

• GPT-4o on OpenAI
• Anthropic Claude on Amazon Bedrock
• Llama3 hosted on-premises behind Ollama or vLLM  

Using this feature, organizations can govern who can access specific models and restrict which models can and cannot leverage their most sensitive data. Any model compatible with the OpenAI API can be connected to Starburst. In this way, organizations can bring their own models while maintaining strict data governance, enforcing usage policies, and controlling operational costs.

Together, these capabilities make Starburst AI Agents a powerful toolset for turning enterprise data into actionable intelligence—securely, at scale, and without added complexity.

 

The Starburst AI Assembly Line

We didn’t stop there. Starburst is committed to being the leading data platform for apps and AI, helping our customers adopt future-proof architecture as quickly as possible. 

In addition to the features already outlined, we are introducing several new features across both our fully managed Starburst Galaxy and our self-managed Starburst Enterprise offerings. Collectively, these make it easier to build an Open Hybrid Data Lakehouse and leverage Lakeside AI. 

These features also help support both the Starburst AI Agent and the main suite of Starburst AI Workflows. Let’s look at them one by one. 

Starburst Data Catalog

An enterprise-grade metastore designed to replace the Hive Metastore. It offers native Iceberg support, seamless migration, and future multi-engine integration, reducing metadata sprawl and improving governance. 

Build a Hybrid Iceberg Lakehouse on Starburst Enterprise

We continue to enhance the self-managed Starburst Enterprise platform with additional features, and these new Iceberg capabilities also unlock our AI Workflows. Advanced Iceberg features like automated maintenance, auto-refreshing materialized views, and full support for Data Products on Iceberg.

New Starburst-Native ODBC driver

Starburst is pleased to announce the release of a new, native ODBC driver. This high-performance ODBC driver enhances Business Intelligence (BI) integration with secure, scalable access and support for various tools, including Tableau and Power BI.

Fully Managed Iceberg Pipelines

Managed Iceberg Pipelines delivers a fully managed, end-to-end Icehouse experience within Starburst Galaxy, streamlining the entire process of turning raw data into analytics-ready Iceberg tables. From ingestion to optimization, this set of features automates the most challenging aspects of building and operating a modern data lakehouse.

Whether data arrives as files in S3 or  Kafka events, Galaxy handles ingestion, hydration into raw tables, transformation into live tables, and continuous optimization—automatically. There’s no need to stitch together pipeline tools, run manual maintenance, or manage infrastructure. Your data stays performant, governed, and always ready for fast analytics, AI, and data products.

With Managed Iceberg Pipelines, Starburst Galaxy becomes the shortest path from raw data to actionable insight—open, interoperable, and AI-ready from day one.

Scaling Iceberg Workloads 

For customers who want more control outside of the fully managed offering, we continue to release new features to make it easy to work with Iceberg and Starburst. Automated Table Maintenance across deployments, AWS S3 Table support, and nanosecond timestamp precision enhance performance and analytics at scale.

AI-Powered Auto-tagging

With intelligent column-level tagging powered by LLMs, Starburst Galaxy makes it easier than ever to govern sensitive data using Attribute-Based Access Control (ABAC) by unlocking secure, self-service access. Teams can classify PII and other sensitive data, scale access policies, and collaborate confidently without relying on data scientists or the labor-intensive work of manually reviewing data.

Automatic Query Routing 

Role-based and deployment-aware routing boosts efficiency and resilience for large-scale, high-concurrency environments.

Data-to-AI Readiness Blueprint 

Strategic service offering to evaluate, modernize, and align enterprise data infrastructure for scalable, governed AI adoption.

 

What Starburst + AI means for you

These capabilities reflect a simple idea: AI is only as good as the data that fuels it. Enterprise AI efforts will stall if the data used for agentic AI or LLM models is siloed, outdated, or inaccessible. 

Starburst is built to solve that problem—giving you open access, intelligent workflows, and robust governance across your entire data estate. Our platform supports your complete data journey from interactive analytics to autonomous AI agents.

We’re confident that using the data lakehouse to power AI architecture provides the perfect gateway to enterprise AI adoption. In short, Lakeside AI is the fastest path to enterprise AI.