As Microsoft Fabric has continued to innovate, additional features are constantly being added including additional AI functionality in Data Warehouse. The AI implementations bring different AI functionality to two different personas, the data engineer and one for the business analyst.
Copilot in Data Warehouse
Within Data Warehouse, Copilot can generate T-SQL queries for you as it knows your table structure, so you can use the chat pane to have it write code for you. If your code is not working or you have no idea what the code is supposed to be doing, you can use Help the developer out with code, including say go ahead and use varchar(max) as that data type was added recently. You may also want to use copilot to tell you how to make the data warehouse case insensitive like SQL Server because you are tired of remembering which letters in columns are and are not capitalized and this feature is now available.
One thing to remember is that copilot does work a little differently inside of Data Warehouse than it works outside of it. Copilot does not remember the responses after you commit a change. Don’t expect it to undo changes either, which is something that it just doesn’t know how to do. And remember like all generative AI it hallucinates about 10% of the time, so you will want to validate the answers it provides.
AI Skills: Helping business users learn about the data
Another method of implementing AI is through AI skills. You can use AI to query the existing warehouse using natural language questions as it provides a chatbot to allow users to ask questions about the data warehouse content.
In order to use AI Skills, you need to first turn on the feature in the Tenant Admin portal, as you see here.
Once the feature is enabled, you can add AI Skills to your data warehouse by selecting a data source, such as a warehouse or a lakehouse. You can select all or part of the tables to be part of the skill. When the users select the AI skill, they will see a chatbot in the center of the screen which they can use to ask it a natural language question regarding information in the database and get a text answer. It is also possible to configure AI skills to include a T-SQL query to use to get the data another time.
Tailoring the chat bot for your Data
Data engineers have the ability to improve the answers that AI Skills provides by providing queries which are modified to expressly provide the answers that users are most likely to want. AI Skills provides the ability to add notes to the model to help improve the interpretation of terms. The generated AI Skills model can also be tweaked to provide better answers by allowing you to substitute custom generated queries to answer questions posed by users. This provides the ability to make the answers even more relevant to your users. If you are interested in trying this out the step by step instructions are available in Microsoft Learn.