Thorogood’s Conversational AI Accelerator Built on Databricks Genie
Accelerator Overview
Overview: Thorogood’s scalable Conversational Analytics powered by Databricks Genie
Thorogood’s Accelerator enables enterprises to rapidly design, deploy, and scale conversational analytics solutions built natively on Databricks through managed MCP servers and Unity Catalog. The accelerator combines a phased delivery model with a configuration driven architecture, allowing new Databricks tools and MCP connected services to be onboarded without significant rework. The solution connects Genie Spaces and a Databricks Vector Search index containing Power BI report descriptions, enabling the agent to link relevant reports to user queries. Through these connections, business users can query governed, consumption ready datasets using natural language, removing the dependency on SQL expertise and making data insights accessible across the organization. The framework is designed for iterative expansion, delivering measurable value within weeks while maintaining enterprise grade governance and security through Unity Catalog.
How Does it Work? Architecture: Databricks Genie Spaces & Unity Catalog at the Core
The accelerator is anchored around managed MCP servers connecting to Databricks tools, with Genie Spaces serving as the natural language intelligence layer over structured, governed datasets managed in Unity Catalog. An AI agent surfaced via Microsoft Teams connects to one or more Genie Spaces through Databricks provisioned MCP servers, enabling users to ask business questions in plain English and receive accurate, governed responses. A Databricks Vector Search index, also managed through Unity Catalog configuration tables, enables the agent to surface semantically relevant Power BI reports alongside Genie query results. The configuration driven framework means additional Databricks tools and services can be onboarded by updating configuration tables in Databricks Lakebase, making the solution highly scalable with minimal development overhead. Thorogood’s delivery approach begins with a workshop, designed to align stakeholders on business objectives, define success criteria, and identify priority use cases for the solution. This session ensures clarity on data sources, user needs, and architectural direction before moving into delivery followed by a rapid pilot, iterative improvement cycles, and production handovers.
Where did it come from?
Thorogood works extensively with our customers across industries at the intersection of business goals and technical possibilities in order to unlock real business value. We view agentic AI as the latest in the toolbox of technical possibilities which can drive real change throughout organizations. We have found across many business facing applications that an iterative approach to design, delivery, feedback, and evolution leads to more assured value, better business buy in and adoption, and overall greater initiative success. Each component of the framework from Genie Space setup and MCP server provisioning to Unity Catalog governance and agent deployment has been refined through real client engagements, ensuring the accelerator is production tested.
Reference Architecture
Interested?
Please reach out! Our consultants will be happy to schedule a free consultation with you to explore how Thorogood’s Agentic Implementation Framework can help transform your business processes and delivery real value in a matter of weeks.