T-SQL Tuesday #181 The Data Platform Present

This month, Kevin Chant has a really fun writing challenge for us. Not only is it a very inspiring call to write about SQL Server (my choice), but it’s also a crossover between T-SQL Tuesday and the Festive Tech Calendar. It’s like opening a digital present!

Kevin created a very nice image that shows the crossover quite nicely.

The original invite stated “My invite is to write about a Microsoft Data Platform announcement that you considered to be as amazing as a present. In other words, something which made you go “wow”.”.

SQL Announcement

So to get directly to the point I’d like to make; my blog is about the announcement of Copilot for SQL Server. Now I know a number of people, maybe even you, will start to mumble and grumble about bots, AI and that they’re better in writing SQL than this Copilot. And you’re probably right. Also, you are probably not be the designated audience for this Copilot. Or you probably are but haven’t realised it yet.

Copilot audience

If you’re a (very) experienced DBA, writing SQL can very well be a second nature. Still, there can be a lot of tedious typing involved. Something that can be alleviated by asking Copilot to do that for you. Yes, there can be some rewriting involved and it will never have your preferred syntax with either leading or trailing commas, tabs or spaces. But it will save you typing.

Coincidentally, I gave a presentation to a number of co-workers today. People who are involved in supporting all the different solutions we’ve deployed at customers. There are a lot of nuances in every solution and I have to salute that team for being able to manage all these differences. But this also leads to them not being specialists in any subject or tooling. They’re not the people who dream SQL code and can write pivots or recursive CTE’s by heart. They will benefit massively from this Copilot.

And it’s not just about creating SQL code, the Copilot for Azure SQL will also help in identifying low hanging fruit when it comes to tuning and optimising the database. One simple example. Copilot can leverage all the goodness that’s stored in the Query Store. When I talked about that treasure trove of data, not everyone knew about that. And that’s perfectly fine, you can’t know everything. And with the portal Copilot, it will know it for you and help you get the data from it. It will even explain the results. And with that, it will increase their efficiency and quality of work even further.

Copilot audience

The funny thing is, when ChatGPT started to go around, my first instinct was to avoid it. Surely an algorithm can’t beat me? And at that time, I was right. But the models powering ChatGPT (and Copilot) are learning much faster than I can. When I tested out the latest version of Github Copilot in Visual Studio Code (powered by ChatGPT 4o) I was really impressed by the quality, much better than earlier versions I tried out and discarded.

No, it isn’t perfect but neither are we as mere humans. And it takes time to check, verify and test the generated code. But is that really much different from our daily work? We try, check verify and test our code as well (sure, we do it on production but still… ;).

And this goes for the Azure SQL Copilot as well. It’s not perfect, it can make mistakes and, when I demoed it on a database with 22.000 tables, it choked with very general inquiries. It only means that, just like a new co-worker, you need to give it some hints and directions.

Therefore I think the biggest data platform gift this year is SQL Copilot. If it will only save the bacon of one of my support co-workers on a rainy monday morning, I will be happy. For them, for the customer and for all the people who have written blogs on the public internet that are part of the model returning the helping hand.

If you’re wondering about some more in-depth experiences, check out this link!

Thanks for reading and have a wonderful festive season!