Databricks Adds AI Coworkers

Databricks is aiming to bridge the gap between raw data and actionable AI insights with a trio of new offerings: Genie One, Genie Agents, and Genie Ontology. The company announced these tools are designed to bring AI into the daily workflows of business users, moving beyond simple chatbots.

Genie One is positioned as a "data-smart AI coworker." It aims to connect insights across an organization's entire data estate and integrate with everyday business applications like Slack and Microsoft Teams. This allows users to ask questions in natural language and receive answers grounded in their company's specific data, rather than generic inferences. Databricks is also launching Databricks Genie Sparks Media Personalization, showcasing its potential in specific industries.

The system offers full co-working capabilities, including monitoring, document creation, and custom skills, extending its utility beyond analytics. Sales leaders, for instance, could get daily briefs compiled from various data sources and communications.

For users on the move, Databricks is releasing iOS and Android apps for Genie One. The platform also accommodates existing AI agents through the Genie MCP App, aiming for seamless integration into current workflows. The company also highlighted how Databricks Genie accelerates pharma launches, demonstrating its cross-industry applicability.

From Prompt to Autonomous Action

Genie Agents represent an evolution from Databricks' previous Genie Spaces. These are domain-specific AI agents that can take autonomous action, completing multi-step workflows without direct oversight.

Agents can reason over both structured and unstructured data, grounding their actions in documents, files, and knowledge sources alongside traditional tables and views. Creating an agent is intended to be as simple as describing the desired outcome in a prompt.

This allows domain experts to scale their knowledge by transforming trusted rules and workflows into reliable team resources.

Context is King with Genie Ontology

Powering both Genie One and Genie Agents is Genie Ontology, an automatic context layer. It extracts knowledge snippets from various data sources and connected applications, organizing them into a "living graph" that represents how a company operates and what its data signifies.

This layer provides Genie with context on where to find information, what sources are trustworthy, and how to formulate answers that align with the company's actual data usage. It includes metric definitions, business terms, and relationships between concepts.

Genie Ontology uses a system similar to PageRank to determine the authority of information sources, weighing factors like origin, author reputation, usage frequency, and recency. Permissions are enforced, ensuring users only see content they are authorized to access.

Internal benchmarks suggest Genie Ontology significantly boosts agent performance on complex enterprise data tasks, achieving higher accuracy and lower latency compared to general-purpose coding agents.

Governed and Secure by Design

Databricks emphasizes that Genie One includes comprehensive admin governance tools. Permissions are enforced by default, and the Unity AI Gateway provides a centralized governance point for administrators.

The company states that security and governance are foundational to the Genie platform, ensuring enterprise readiness.