Flipping the Script

Start with Themes, Not Codes—Here’s Why

Author

Susanne Friese
June 12, 2025

Hi

In today’s newsletter, I want to shed some light on theme building with LLMs. Before the LLM era, coding was essential for managing and making sense of qualitative data. You got to know your material through the coding process, and built a structure—categories, subcodes, and sometimes higher-level themes—to organize it all. Without coding, it was simply too difficult to hold everything in your head and analyze it systematically.

But now, LLMs can generate themes (or higher order concepts) in minutes. And you might be wondering: how is that even possible? Are these real themes, like the ones I used to craft after weeks or months of coding?

Some of our users asked the same question—and here’s what they discovered:

“QInsights was able to identify the same themes that took me months to create in a fraction of the time. It’s a clear, effective, and efficient tool. I’m genuinely impressed by how much detail it preserves.”

“The comparison of my first QInsights analyses coincides with the results I got when using a traditional QDAS – it was just much faster ;-).”

Another user uploaded a document they had already spent two days coding. The results from QInsights? Nearly identical themes—returned in under two minutes.

Impressive? Yes. But also a bit unsettling. If themes appear this quickly, you may feel disoriented—like you’ve skipped the part where you really work with the data. If you haven't helped shape the themes, it may feel like something crucial is missing.

And that’s exactly the point.

Themes aren’t the end. They’re the beginning.

LLMs are great at generating overviews. That’s their strength. But if you stop there, you’ve only skimmed the surface. You’re missing the context, the voices, the nuance.

So instead of the traditional workflow:

  1. Code data

  2. Build a code system

  3. Interpret and extract meaning

You flip the script:

  1. Upload your data

  2. Let the LLM surface potential themes

  3. Use those themes to guide your inquiry

  4. Build meaning in dialogue with your AI assistant

  5. Reconnect with the data to test and illustrate your ideas with original quotes

With QInsights, you can do all of this:

  • Extract themes.

  • Ask follow-up questions in Themes Analysis to explore what lies behind a theme

  • Or start a conversation with the data, guided by the theme

  • Or bring a theme into a Guided Analysis to dive in step-by-step guided by Q, your AI-assistant in QInsights.

👉 If you want to dig deeper and understand better how LLMs work, I just published a new blog post – a start of a new series - Conversations from Inside the AI Mind. This first post looks at what is really happening when an AI codes data.

On the horizon:
I’ll be back in your inbox soon with exciting news—two long-awaited features are finally here, and I can’t wait to show you what’s new.

Warm regards,
Susanne

Dr. Susanne Friese
Rebel Methodologist
Founder of Qeludra and Co-Founder of QInsights

Converse of Insights