Hey there,

So I was definitely late to the whole AI thing.

While everyone was posting screenshots of ChatGPT doing amazing stuff, I was sitting there thinking "okay, but how does this actually help me teach data?"

Plus, a lot of us educators are nervous and cautious about AI. If tools can write code and analyze data, will students still learn the fundamentals? How can we make sure they do?

But earlier this year, I finally decided to stop being skeptical and actually try it.

What Changed My Mind

The breakthrough wasn't AI writing better code or queries. It was realizing it could simulate the messiness that happens at work (which is rarely addressed in textbook examples).

For example, how do you teach students to handle a manager who says "I need customer insights" and then gets frustrated when you ask what kind? I used to try role-playing in class, but it felt forced and awkward.

But AI can actually play that confused manager.

I can tell it to be a marketing director who thinks "engagement" just means social media likes. Then I watch my students figure out the right follow-up questions.

Or have it play a VP who says "our data shows we're losing customers" without any details about which customers or when.

Suddenly my students are practicing skills that matter just as much as knowing syntax or the next feature.

The Reality Check

I got excited and started using AI for everything. But I did hit some walls pretty fast.

Some stuff worked great - like generating fresh practice problems instead of using the same tired datasets.

But then I started wondering: should students learn to write code independently first?

What happens when they can't tell if AI gave them the right answer?

Because AI does make mistakes - there's literally a warning about it in every chat conversation.

How do students develop the critical thinking skills to spot when a query looks right but actually isn't? Or when an analysis seems plausible but misses something fundamental?

Without that solid foundation, how can they use these powerful tools effectively and responsibly?

What I'm Working On

This whole experience made me realize other educators and trainers were probably hitting the same questions.

So I started writing down what worked and what didn't.

That became a newsletter called Teach Data with AI.

Some of the recent issues cover:

• Creating practice emails from managers (so students can practice translating vague requests)
• Building datasets with realistic errors that mirror actual workplace problems
• Teaching chart reasoning instead of just chart rules (and teaching when to bend the rules)

Here's what I mean by specific prompts.

Instead of "create a practice problem," I use something like:

The specificity matters because students get practice with real problems, not textbook perfection.

It's still a work in progress. Or an experiment in progress. It's a moving target because I experiment with the prompt and tools, but the tools keep on changing as well.

One thing's for sure - I'm not trying to replace teaching with AI. I'm using it to handle the tedious setup so I can focus on the thinking parts.

What's your experience so far? Are you experimenting with AI too? I'd love to hear what's working (or not working) for you.

Talk soon,
Donabel

​

Most of my lessons are still completely manual. Some things still need a human who remembers being confused.

​Here is an example issue on generating realistic data exercises.​
​