Q&A
Creating Simple Chat Bots with Microsoft Fabric Datastores
As generative AI becomes increasingly integrated into enterprise applications, one promising frontier is enabling conversational access to organizational data. Microsoft Fabric, the end-to-end analytics platform introduced to unify data movement, processing, and governance, is now playing a central role in that evolution. With native support for data lakehouses, warehouses, and AI integration, Fabric is making it easier for developers to build chatbots that can answer business questions directly from curated datasets.
In her upcoming session titled "Creating Simple Chat Bots with Microsoft Fabric Datastores" at Visual Studio Live! San Diego, Ginger Grant, principal data geek at Desert Isle Group, will demonstrate how to create simple yet effective chatbots using Microsoft Fabric datastores. Aimed at intermediate-level developers, the session focuses on practical, real-world implementation using data already housed in Fabric environments. Data Platform MVP Grant will walk attendees through the end-to-end process of enabling AI-powered query capabilities on Fabric lakehouses and warehouses, leveraging grounding techniques to enrich chatbot responses with organizational data.
With a demo-heavy format and a clear focus on cost-effectiveness and speed to implementation, the session promises to equip attendees with the tools to build functional prototypes in the time it takes to attend the presentation. From choosing the right data store type to understanding how to format data for optimal results, this session offers a comprehensive primer for developers interested in extending the reach of their analytics through conversational AI.
We caught up with Ginger to learn more about what inspired this session, what attendees can expect, and how Microsoft Fabric is redefining the chatbot development process.
VisualStudioMagazine: What inspired you to present a session on this topic?
Ginger: As there are a number of different methods and tools being used to provide answer to data and Microsoft is encouraging organizations to centralize all their data inside of Fabric, I thought many people would be interested in learning a solution which utilizes the app that already contains the data.
"There are various innovative ways to utilize data stored within Fabric to effectively answer questions and solve problems that people have about their data."
-- Ginger Grant, Principal Data Geek, Desert Isle Group
There are various innovative ways to utilize data stored within Fabric to effectively answer questions and solve problems that people have about their data. By sharing these insights, I hope to empower others to harness the full potential of Fabric for their own data-driven applications.
What's the main difference between using a Fabric lakehouse vs. a warehouse when building a chatbot -- is one more efficient for this use case?
The data the data stored in a data warehouse or a lake house inside of Fabric are both parquet files, the difference lies in the tool integration. Lakehouses have better integration with Spark and for that reason it is easier to transform data stored in a lakehouse and use it for a chat bot.
Once you make the choice, what's the first step?
The first step is data analysis. You need to examine how the data is stored, what data is needed to provide quality answers and if the current structure supports this kind of analysis. Next you need to format it to something that a chat bot can read. Part of the process is also going to be if you need to supplement the information with data which will provide more meaning. For example, you may need to do some synonym creation if you use a lot of abbreviations and different kinds of words because the chatbot natively is not going to know things like the abbreviation no means number.
Can you clarify what “grounding” means in the context of chatbot queries against data in Microsoft Fabric?
The process of grounding means that you are going to be using the underlying technology of an existing model such as ChatGPT or Claude and add to it multiple data sources within Microsoft Fabric. The chat bot retrieves the most pertinent information related to the query, which includes proprietary data that is used within your organization to ground or provide the answers to questions. Using the retrieved data, the chatbot generates a response that is grounded in the actual data available. This means the response is not just based on pre-trained models but is enriched with real-time, context-specific information. That's what it's meant by grounding.
Do you need to use any specific LLM provider with Microsoft Fabric, or can you integrate open-source models too?
You have your choice of selecting any of the Large language models or frameworks included inside of Azure OpenAI, which includes a number of different large language models, including some which are open-source, which can be less expensive. This versatility allows you to choose the best approach based on your technical requirements, budget, and strategic goals. And for more information about how to select the right model based on your requirements, you'll need to attend my session.
What resources can attendees use to learn more and prepare for your session?
Prior to attending the session, it would be good if you understood a little bit about Fabric architecture and or parquet files. And their method of storage and why so many data elements are stored in them. However, if you don't have this knowledge but have a good technical background, I'm sure you will be able to follow along.
Note: Those wishing to attend the session can save money by registering early, according to the event's pricing page. "Save $400 when you register by the July 11 Super Early Bird deadline" said the organizer of the event, which is presented by the parent company of Visual Studio Magazine.
About the Author
David Ramel is an editor and writer at Converge 360.