How to Use AI as a Thinking Partner for Qualitative Analysis

By DoReveal

• article

• AI
• Artificial Intelligence
• Qualitative Research
• Qualitative Data Analysis

When people talk about AI and research in the same breath, the conversation almost always goes to speed. How much faster can I get my transcripts? How much time can I save on the first pass of analysis? These are real and useful benefits. But I think if we stop there, we’re leaving the most interesting part of the conversation on the table.

I’ve spent thirty years in research and design, and the last ten applying AI in various capacities. What I’ve come to believe – and what I find genuinely exciting about where we are right now – is that AI’s real value in qualitative research isn’t about doing the same things faster, but about enabling us to go deeper, think more rigorously, and ask questions we wouldn’t have had the time or the methodological confidence to ask before.

That’s what I want to explore here.

Learn how AI makes qualitative frameworks more accessible:

AI as a Creative Platform

One way to think about AI is as a creative platform rather than an automation engine. If I don’t know something, I can ask questions with a very specific context – what I’m looking for and why – and get a highly customised answer. I can learn much faster through that process. And when I’m analysing data, AI enables me to go beneath the surface and gain richer insights in ways that weren’t practically achievable before.

This is what I mean when I talk about ten-times research. Not ten times the speed, though that has its own value. Ten times the depth – ten times the questions I can ask of the same data. And the key to unlocking that is something that researchers have always had access to in principle, but rarely had the time or resources to use fully: analytical frameworks.

What Frameworks Actually Do

I should be clear that I’m using the term “framework” loosely here. I mean grounded theory, jobs to be done, journey mapping, emotional laddering, persona development, means-end chain analysis – the full range of structured, systematic approaches that researchers use to direct their attention onto something specific in the data.

One way to think about what frameworks do is this: when you’re working with qualitative data, there is an enormous span of possible observations and insights in front of you. A framework helps bring some of that information into the foreground while other things recede into the background. It’s not that what’s in the background isn’t real, it’s that you’re choosing where to focus your attention, and different frameworks direct that attention differently.

The metaphor I keep coming back to is a metal detector. You’re scanning the surface, and somewhere something goes beep. You dig and find something interesting, or you don’t. Either way, you come back to the surface and keep scanning, constantly zooming in and out, bringing things forward, making determinations. That’s the nature of good qualitative analysis. And frameworks are what allow you to do that scanning systematically, rather than just following your instincts.

The Same Data, Multiple Lenses

Let me make this concrete. Say a participant in a study about a food delivery app told you this: they were ordering takeout after work when they felt exhausted, and even though it was more expensive, it made their life easier and they could justify the cost.

Read through a broad thematic lens, you’d extract something like: fatigue plays a role in the decision, convenience matters in this context, and there’s some price sensitivity that is overcome by the perceived ease of use.

Now apply a journey map to the same quote. Suddenly you’re tracing a sequence: the trigger is reaching the end of the day and not wanting to cook. The user opens the app, is potentially overwhelmed by choices, manages to place an order, and experiences relief, followed maybe by a lingering guilt about the cost. That emotional arc from exhaustion to overwhelm to relief to guilt is a genuinely different kind of insight than a thematic summary would give you.

Apply jobs to be done to the same data, and something else comes into view. The functional job is clear: access to dinner without effort, so I can recover. But the emotional job takes it a step further – the person wants to feel taken care of. They don’t have the energy, and they want the feeling of restoration that comes from not having to manage yet another thing. That’s a meaningfully different insight from the journey map, and it has different implications for design and communication.

Try a laddering exercise, and you’re asking why at every level. Why is it easier to order? Because cooking feels cumbersome – there are too many decisions and steps. Why does that matter? Because I’m already overloaded. Why does that matter? Because I don’t want to feel behind. I want to handle this so I can focus on my work and achieve what I set out to do. Now you’re touching on something about identity – how this person sees themselves and what they’re trying to protect.

Three frameworks, the same forty-word quote, three genuinely distinct, yet complementary, perspectives. Each one illuminates something the others don’t.

Where AI Changes What’s Possible

Here’s the honest challenge with everything I just described. To do each of these well, you need to know the frameworks. You need the depth to apply them correctly, and time: if every framework takes hours to apply properly across twelve interviews, you simply can’t do all of them in a typical project budget.

This is where AI significantly changes the picture.

What I’ve found is that current large language models – such as Claude, ChatGPT, and Gemini – have been trained on substantial amounts of material about these frameworks. The depth of that training isn’t fully known, since it’s not published, but if you go and ask any of these models whether they’re familiar with, for example, interpretive phenomenological analysis or Kelly’s personal construct psychology, they will tell you what it is and how to apply it. You are not starting from zero.

This matters in a few ways. First, it means you can use AI to rapidly develop your skills and confidence with frameworks you don’t know well. Ask it what framework you should use for a particular kind of sentiment analysis, to explain how personal construct psychology actually works in practice, or to generate a starting prompt for applying a specific technique to your data. That conversation, the back-and-forth, is itself a form of learning.

Second, and more directly, it means you can ask AI to apply these frameworks to your actual data. Upload your transcripts, give it the right context, and ask it to do a laddering analysis, build behavioural personas, or map the emotional journey from your interviews. It will do it. Not perfectly, and I’ll come back to that, but it will give you a substantive starting point in seconds rather than hours.

The Researcher Still Drives

I want to be careful here not to oversell this in a direction that I don’t think is accurate or useful. AI is not replacing the researcher’s judgment. In my view, it’s multiplying the researcher’s capability to exercise that judgment on more of the data, through more lenses, more quickly.

When AI applies a framework to your data and gives you a set of ladders or a set of personas, those are not the answer, they’re the raw material. They’re a signal that says something is here, go look. From there, the researcher decides what’s meaningful, what’s context-dependent, what contradicts the expected pattern, what deserves to be foregrounded and what can safely recede. That’s irreducibly human work.

What AI offers is the ability to generate that raw material much more quickly, and to have a genuine conversation about the data at the level of your inquiry – not at the level of explaining how the framework works or wrestling with how to structure the analysis.

Diversity of Thought as a Research Asset

One of the things I find most genuinely valuable in this, beyond the time savings, is what I’d call diversity of thought. Because AI understands these frameworks, you can interact with it almost as if you’re working with different colleagues – each one bringing a different lens to the same data. That gives you the ability to challenge your own biases, get ideas you might not have reached on your own, and start to triangulate across different analytical perspectives to find what’s really robust.

And that, I think, is closer to the kind of ten-times research that AI actually makes possible. Not ten times faster, but ten times more thorough and more honest about what the data contains and what it doesn’t. That’s worth getting excited about.

Learn how AI makes qualitative frameworks more accessible:

Author

Learn more about