AI assistant that connects to all my work tools

Summary

• An AI assistant that connects to all your work tools should answer in plain language over both your structured data and your documents, and it should respect the permissions you already have rather than opening a side door around them.
• On Databricks, AI/BI Genie is the natural-language assistant over governed data: you ask a question, it writes and runs the SQL against curated tables, and it returns an answer with a chart.
• Knowledge Assistant grounds answers in unstructured sources like documents, while Lakeflow Connect ingests SaaS application data (from CRM, HR, IT service management, and collaboration tools like Confluence) into Unity Catalog so the assistant can use it.
• Because everything runs on Unity Catalog, the assistant sees only what the signed-in user is allowed to see, analytical queries are read-only, and every interaction is logged; Model Context Protocol (MCP) connections extend it to external tools under the same governance.

AI assistant that connects to all my work tools
The assistant you want does three things well: it reaches your data and applications, it answers in plain language over both structured tables and unstructured documents, and it respects the access rules you already have. That last part is what separates an enterprise assistant from a demo. On Databricks, AI/BI Genie is that natural-language assistant over governed data, Knowledge Assistant grounds answers in your documents, and Lakeflow Connect plus Unity Catalog connections and Model Context Protocol (MCP) bring your work tools into one governed place.

What to look for in an AI assistant that connects to your work tools

Capability	Why it matters
Connects to your sources and apps	The assistant is only as useful as the data it reaches. It should ingest from your SaaS applications and query your existing tables, not just a single silo.
Natural language over governed data	You ask in plain language; the assistant writes and runs the query against curated, trusted datasets and returns a clear answer, ideally with a visualization.
Structured and unstructured	Real questions span dashboards and documents. A good assistant answers from tables and from knowledge locked in files.
Permission-aware	The assistant should inherit each user's access rights, so people see only what they're entitled to. A shared, over-privileged bot is a security problem.
Read-only and auditable	Questions and answers should be logged, and analytical queries should be read-only, so you get an audit trail for compliance.
Open and extensible	Support for open standards like MCP lets the assistant connect to external tools without custom one-off integrations.
Governed in one place	Access control, lineage, and business definitions should live in one catalog so every answer uses the same trusted numbers.

How Databricks approaches this

AI/BI Genie is the natural-language assistant for business users. You ask a question in plain language, and Genie translates it into SQL, runs it against curated tables, and returns the result with a visualization. Data teams set up Genie Agents, domain-specific environments where they register trusted datasets in Unity Catalog, add sample queries, and write the business rules and terminology so answers match how your organization talks about its data.
Knowledge Assistant, part of Agent Bricks, grounds answers in unstructured sources such as documents, so the assistant can pull from business knowledge stored in files, not just tables. Agent Bricks builds and governs these agents on your enterprise data.
To bring your work tools in, Lakeflow Connect provides managed connectors that ingest data from SaaS applications and databases into Unity Catalog. Documented connectors span business systems such as CRM, human resources, IT service management, and web analytics, plus collaboration tools like Confluence, so the records from the tools your teams use become governed tables the assistant can query. For tools and services that expose their own interfaces, Databricks added built-in support for Model Context Protocol (MCP) connections, and the Genie Agent MCP server lets agents query a Genie Agent in natural language without writing SQL. MCP connections are managed and governed through Unity Catalog, so they stay simple to use and well governed.
The foundation under all of it is Unity Catalog. It provides centralized governance for data and AI assets, controls access, tracks lineage, and enforces consistent business definitions through Unity Catalog Business Semantics, a single place to define metrics, dimensions, and rules that dashboards, SQL, and AI agents all share. Genie operates inside those access policies: users see only the data they're authorized to see, analytical queries are read-only, and every interaction is logged for audit. For business users, Genie One offers a simplified home to find and use data, dashboards, and apps without navigating workspace structure, with connections to external sources such as Google Drive and SharePoint. See the Genie Agents documentation and the next-generation Genie blog for details.

Use cases

Self-service business questions. A sales or finance lead asks why a number moved and gets a governed answer from trusted tables, without waiting in an analyst queue.
Answers from documents. A support or operations team asks a question, and Knowledge Assistant returns an answer grounded in policies, runbooks, and other files.
Agents that act, under governance. MCP connections let agents reach external tools and take actions, with Unity Catalog enforcing who can do what.

FAQs

Does the assistant respect my existing permissions?

Yes. Genie runs inside Unity Catalog access policies, so a user sees only the data they're authorized to see. Analytical queries are read-only, and every interaction is logged for audit.

Can it answer from documents, not just tables?

Yes. Knowledge Assistant, part of Agent Bricks, grounds answers in unstructured sources such as documents, so the assistant handles both structured data and business knowledge stored in files.

How does Databricks connect to my SaaS work tools?

Lakeflow Connect provides managed connectors that ingest SaaS application data, from systems such as CRM, HR, IT service management, and web analytics, plus collaboration tools like Confluence, into Unity Catalog, where the assistant can query it as governed data.

What is MCP and why does it matter here?

Model Context Protocol (MCP) is an open standard for connecting AI agents to tools and data. Databricks supports MCP connections managed through Unity Catalog, and the Genie Agent MCP server lets agents query a Genie Agent in natural language.