Testing the Limits of PandasAI (Part 1): What It Can (and Can’t) Do to Help Data Scientists

Using AI to optimize your analysis workflow

Testing the limit (Part 1). Image by Author

I remember when I first heard of PandasAI back in late 2023.

ChatGPT-4 had just been released, and we were seeing the first waves of people trying to bolt LLMs onto real data workflows

What an exciting time!

Naturally, I had to try PandasAI myself and see what all the hype was about. I mainly wanted to see if you could really “talk to your data“ and finally solve my problem of forgetting Pandas and Matplotlib syntax.

Long story short, PandasAI was cool, but it wasn’t very useful in practice.

And for us Data Scientists, who already juggle countless tools and syntax, adding another step is only worth it if it truly brings value.

Fast forward to 2025, LLMs have gotten MUCH better, especially at logic and coding tasks, and PandasAI has, of course, continued to evolve.

So a few months ago, I decided to give it another shot as I’ve been integrating AI more into my data science workflows (more about this at the end of the article 😉).

Now I want to share where PandasAI stands today, and most importantly, what it can (and…