If you think the best use of LLMs in data science is coding, then you’re missing some of the most powerful opportunities.

Late last year, I had an important conversation coming up with a key stakeholder who I’d known for years.

And when I say important, I mean the sort of conversation that could make or break a career.

Let’s just say that this person had a very particular way of responding to situations, and I knew from experience that if I made one false step, the conversation could go sideways quickly.

So, I mentally mapped every possible direction the conversation could possibly take and came up with a game plan for each. But that wasn’t enough. I still felt there was something missing.

That’s when I decided to try something new.

I put a couple of sentences into Claude describing who this person was, how they’d typically responded in the past, and gave some examples of their behaviour. Nothing requiring a psychology degree - just my basic observations.

Then I asked Claude to role-play the conversation with me.

The result was uncanny. The AI captured exactly how this person would behave, down to their specific communication style and their likely reactions to different approaches.

My first attempt at having the conversation was a complete disaster. But after a few more iterations, I quickly developed a good understanding of the best strategy to take.

When I had the real conversation, I was prepared. I knew which arguments would resonate, which approaches to avoid, and how to frame my points in language my stakeholder would understand.

This is one example of how LLMs can be used to make you a more effective data scientist. But there are many others.

Actuary and data scientist Colin Priest has also been experimenting extensively with advanced LLM applications that go far beyond writing code faster.

In the latest episode of Value Driven Data Science, Colin joins me to share techniques that can transform how you approach stakeholder communication, data analysis, and problem-solving, including:

1. How to use LLMs to extract structured insights from messy unstructured data [02:50]
2. Advanced role-playing techniques for practicing difficult conversations [14:12]
3. Why using multiple LLMs helps reduce AI hallucinations [20:38]
4. A step-by-step approach for integrating LLMs into your workflow safely [25:52]

The most powerful LLM applications for data scientists aren’t always the most obvious ones.

Listen now on Apple Podcasts or Spotify, or click the link below:

​Episode 90: Using LLMs to Become a More Effective Data Scientist​

Talk again soon,

Dr Genevieve Hayes