Deb the DBA

It’s another T-SQL Tuesday! Our host this month is Taiob Ali (b). (Thanks for hosting this month, Taiob!) He’s asked us to talk about how AI is changing our careers.

The pressure is definitely there to use AI every day and embrace it or be gone. When I think about my relationship with AI, I think these vignettes may help explain it better:

---

My coworker is helping me with a problem. It’s the classic “the script works on their machine but not mine.” They suggested we use the AI tool all the developers are using. I had coincidentally remembered to finally install it that morning in my VS Code; I don’t actually don’t use it that often. My coworker walks me through the prompts and we hit run.

It’s actually kind of cool. In the side window, the tool is telling me its thought process as I see the changes in the output terminal. I left the code compare window up so I can see the changes through the different iterations. A few minutes later, it’s done. The final output script looks pretty good and the initial problem is fixed. Right?

Until the coworker looks at how the tool modified the code. Turns out that we can’t use it because it removed other code that needed to be there and created a new file that won’t work, along with a bunch of other problems.

Alright, back to the drawing board….

---

Did you see my video on getting started with Vector Searches in SQL Server 2025? I was able to take the code examples that have been shared publicly with us and walked through the steps with my own database and got it to work. I was even able to walk through the documentation to set up Azure AI Foundry to create a model. While it’s an oversimplified explanation of what vectors are and how they work, I’m proud of myself for being able to understand the basics of what things are in a way that hopefully makes sense to others to help them get started as well.

The next question is where do I go from here? I have some ideas of what I want to try next. Hmm… can AI help me find time for that?

---

The vendor showed us their AI offerings. Afterwards, we’re standing around talking about what we just saw and thinking about how it all works. Someone says something and next thing you know, we’re debating what this particular functionality is, how it works and is it really what we think it is. I’m fairly confident I understood it, but I still had to go home and research one of the pieces…

I think the reason for the debate came down to there’s a lot of nuances and differences between the various types of AI and their use cases. It’s nice when vendors can do the demos, waiving all of the hands to make it look good. But when talking to technical people, we should stop the magic show and lift the table cloth to show what’s going on with the architecture underneath. We need to understand the specifics of the tech we’re working with and the differences between them all. Everything is based off of a “model”, but there are so many different types and we use them all in different ways. It’s easy to be confused.

This is where I feel I should throw in a shout out to Muazma Zahid‘s LinkedIn post series, #LearnWithMZ. I so appreciate when people are evangelists not just for the product and company they work on but the underlying tech behind it.

---

It’s not just Muazma. Many people are blogging about the different use cases they’re coming up with on how to integrate various types of AI. I’m reading them and am curious to see where these all go. I’m attending various sessions at conferences or online on how people are using AI or just learning the fundamentals. There are so many different areas that AI can be applied – from summarizing documents, to extracting text from unstructured data like images or sound files, to identifying manipulated images which may indicate fraud, to helping write or analyze code, or even to just have better search abilities with data we already have. There’s a sense that we’re all trying to learn at once because it’s early stages of its maturity. In the meanwhile, it’s great to read some of the cool things that are going on, along with the cautionary tales like Replit. We all know where our Responsible AI Guidelines are, right? RIGHT?????

That reminds me – Our office is on a “AI all the things!” kick. We have dedicated resources and slack channels just for talking and sharing all things AI. I need to remember to add some of these blog posts there…

---

Y’all, I am really tempted. What would it write for me? I’ve been so overcommitted – how am I going to have time to pull together a new post? My poor blog has been neglected! Is this the answer?

But my brain always turns to my friends who are professors and I know they know how to identify work that their students did not do themselves. AI tools are also known to steal copywritten work and other content without permission or accreditation or just make up things that don’t exist. People are submitting what is generated by AI tools without confirming what was generated was right. If you think people don’t know that your output is from AI, you’re wrong; tools exist to identify these things. And it’s not just because there’s an em dash in there somewhere. These tools may not always be right, but do you really want to take a chance?

This doesn’t even begin to cover the deepfake videos that are getting better in terms of quality. It’s quite scary when you realize how much harder it’s becoming to recognize what’s real and fake.

It’s not to say that I haven’t used ChatGPT to help me with a session title or to massage the body of an abstract or to get an idea of how to construct what I want to say. I know people use it to write their self reviews at work or refine their comments to better corporate speak to align with the higher ups. It’s more useful in those situations than I may care to admit.

But when it comes to my blog, the English major in me has too much pride in being able to say that these written words are mine. Every last misused word, awkward phrase, and grammatical mistake. (Apologies to all my English teachers and professors.) Not just that, I want to ensure that what I’m saying is correct and reflects what I want to say. I guess I’m probably going to have a big “trust but verify” – or rather “don’t trust, just verify” – caveat when it comes to AI, especially when it comes to the content I’m creating.

---

So what’s the TL;DR version of how am I using AI right now and how’s it changing my career? It depends….