Intugle helps enterprises build context-aware agentic applications that connect to systems of record, activity and knowledge, unify fragmented business context and embed AI directly into enterprise workflows. As adoption grew, the company needed a scalable way to federate enterprise data, enforce runtime governance for AI agents and operationalize data products without managing complex infrastructure. With Unity Catalog, Lakehouse Federation, Databricks SQL Serverless and Lakeflow Jobs, Intugle built a governed execution layer for enterprise AI across clouds and data platforms.
Building context-aware agents for enterprise AI
Intugle is building context-aware agentic applications for enterprises. The platform autonomously extracts, unifies, and operationalizes context from systems of record, business activities, and enterprise knowledge across structured and unstructured data sources. This context layer powers AI-embedded intelligence inside business applications, enabling use cases such as CXO performance cockpits, intelligent sales command centers, supply chain control towers, customer support intelligence hubs, partner portals, and product recommendation systems.
The platform is built on a multi-agent architecture, with specialized agents responsible for metadata exploration, semantic relationship discovery, transformation generation, predictive modeling and text-to-code (SQL & Python) query generation. These agents continuously interact with enterprise metadata, lineage and business context to generate recommendations and automate workflows.
As Intugle expanded into large enterprise environments, governance and orchestration became increasingly difficult to manage. Earlier approaches using Apache Atlas and Open Source libraries required significant engineering effort to maintain lineage, access controls and metadata consistency across fragmented systems. The company also wanted to avoid forcing customers to replicate large volumes of enterprise data simply to power AI applications.
“We operate on the philosophy of copying data only when necessary,” said Prinkan Pal, Co-founder and CEO at Intugle AI. “Our goal is to unify fragmented enterprise context while allowing customers to continue operating across their existing systems.”
Governing enterprise AI with Unity Catalog
Intugle adopted Unity Catalog as the governance foundation for its structured data platform. The company uses Unity Catalog to centralize metadata management, lineage, runtime access controls and governance policies across enterprise environments.
Its agents directly consume Unity Catalog APIs, lineage APIs and metadata to understand upstream and downstream dependencies, analyze data relationships and generate recommendations. For example, Intugle AI’s metadata exploration agent can identify overlapping or redundant data products across systems and recommend consolidation opportunities based on lineage and attribute analysis.
If approved by the user, the workflow can automatically translate into Lakeflow Jobs that consolidate datasets, generate new governed data products and retire outdated assets.
Unity Catalog also simplified runtime governance for AI agents. Intugle uses role-based and attribute-based access controls (ABAC) to ensure agents inherit the appropriate permissions when generating and executing queries against enterprise data.
“In most environments, access control policies and execution engines are disconnected,” said Pal. “With Unity Catalog, Serverless SQL and Lakehouse Federation working together, those policies are enforced automatically at runtime without requiring us to build custom integrations.”
The company also uses Unity Catalog lineage capabilities to help customers understand how data products are created, transformed and consumed across downstream systems. This visibility is particularly important for enterprises building governed AI applications on top of sensitive operational data.
Federating enterprise data without unnecessary replication
A core design principle of Intugle is enabling enterprises to work across fragmented systems without centralizing or duplicating their data. With Lakehouse Federation and Unity Catalog managed connections, Intugle can connect to sources such as Amazon S3, Amazon Redshift, Oracle databases and existing lakehouses while maintaining centralized governance through Databricks.
When users interact with the platform, Intugle agents generate semantic queries that are translated into physical queries and executed through Databricks SQL Serverless across federated enterprise systems. This allows enterprises to retrieve governed data on demand while preserving existing architectures and investments.
The platform also uses Unity Catalog custom tags and metadata to persist additional business context generated by Intugle’s semantic layer. These tags help standardize business definitions, identify sensitive data such as PII and maintain consistency across large enterprise data estates.
For enterprises operating at scale, this centralized governance model became a key differentiator. Intugle supports deployments with up to 150 terabytes of data and enterprise environments with nearly 10,000 users accessing AI-powered applications built on the platform.
Accelerating enterprise AI with Databricks SQL Serverless and Lakeflow Jobs
A Databricks SQL Serverless became the runtime execution layer behind Intugle’s agent platform. The company’s agents dynamically generate Spark SQL and PySpark queries that execute in real time across federated enterprise systems.
Using Databricks SQL Serverless improved query execution times by 50–60% while reducing startup times from several minutes with traditional compute clusters to just 3–4 seconds. This responsiveness is critical for delivering enterprise-grade AI experiences that users expect to be near real-time.
Prior to Databricks, Intugle relied on Apache Airflow running on Kubernetes to orchestrate workloads. Managing and scaling those clusters created significant operational overhead because infrastructure capacity needed to be provisioned in advance.
“With Databricks Serverless SQL and Lakeflow Jobs, we can scale dynamically at runtime instead of managing infrastructure ourselves,” said Pal. “That dramatically simplified operations for our engineering teams.”
Lakeflow Jobs also helped operationalize agent-driven workflows across the platform. Intugle agents can automatically generate transformation pipelines, create governed data products and execute orchestration tasks programmatically based on user prompts and metadata analysis.
The migration to Databricks also significantly accelerated development velocity. What initially took nearly three months to build using Apache Atlas and open-source packages was implemented with Unity Catalog in roughly a week and a half.
“As a startup focused on building the agentic layer, Databricks gave us a 50–60% acceleration in getting enterprise-ready,” said Pal. “Instead of stitching together multiple open source systems ourselves, we could focus on delivering AI applications while Databricks handled the governance, orchestration and runtime infrastructure.”
