Research Design in the AI Era

Last week, I explored how AI is reshaping research – driving a shift towards immediacy and 'good enough' insights. One of the biggest developments is ‘Qual at Scale’ - which I will unpack next week in our discussion on AI Moderation.

This week, rather than looking at new methods, I’ll examine how GenAI tools can enhance the research design process for qualitative projects - helping to refine research briefs and proposals, write recruitment screeners, discussion and activity guides and create research stimulus.

Shaping the Research Brief – Understanding the Problem

A well-designed research project begins with a clear articulation of the business problem and an understanding of what’s already known on the topic. Some briefs are well written, others less so – but ultimately, it’s the agency’s responsibility to extract the right information to ensure the research delivers real value. GenAI tools - whether used by researchers themselves or those commissioning research - can help develop clearer briefs. 

Firstly, let’s talk about guidance. A research brief is typically shaped around key business decisions that must be made within a defined timeframe and budget. In recognising the need for research, there will be things already known, things partially known, things felt to be true but still not fully known (hypotheses) and things unknown. There may also be constraints, parameters or protocols that needed to be followed in the research design or outcomes. A GenAI workflow can easily be created to prompt with a series of questions to create a research brief (and potentially methodological preferences).

Identifying ‘what is known’ in a complex data ecosystem that is ‘always on’ is not always easy. AI can assist in this stage by analysing large volumes of existing data, such as past studies, customer feedback, or social media content. This can help to identify knowledge gaps and be used to formulate hypotheses, sharpening the research objectives to ensure the study addresses current and relevant issues.

Once the background to the research is known, GenAI tools can help agencies create better research proposals - by framing the research context in creative and structured ways – for example by doing a Situational Analysis or using a PEST framework. The new ability to use GenAI to do ‘deep’ desk research means agencies can easily build upon the knowledge shared in the client’s brief.

Designing Discussion/Activity Guides & Screeners

Traditionally, discussion guides are crafted through experience, intuition and iteration – part of the craft of a qualitative practitioner. Indeed, some experienced practitioners may balk at the idea of using GenAI in research design, however, with the rise of DIY research, more people from a range of educational and professional backgrounds, who don’t have (much) research experience, are conducting research and will look for ‘guidance’. Even experienced practitioners can use GenAI – not to replace their expertise, but as a ‘sparring partner’ to challenge thinking and streamline processes.

GenAI can support the development of discussion and activity guides by suggesting relevant themes, topics and questions aligned with the research objectives and target audience. This doesn’t replace human creativity but provides a structured starting point, allowing researchers to focus on fine-tuning the guide to the specific cultural and contextual nuances of the target audience. Conversely, AI tools can be used to make suggestions to improve the final guide. Whether using at the front or back end (or both), better results will be obtained if the AI tool used is trained on previously used guides.

Screener design - less of an art and more about best practice - is one of the easier wins for GenAI. Its impact will be strongest when trained on a standardised library of previous screeners.

Using AI Personas for Research Design

I’m sure many of you will have heard of AI personas, and some of you have experimented with developing or using, if so, I’d love to hear about your experience.  For those unfamiliar, the basic idea is that we either:

• Take the data and insight behind an existing persona(s) and use it to train the AI model

• Create new personas from existing data

Once built, these personas can be interacted with using either text or speech (depending on the model used to build the persona). At FUEL Asia, we’ve built AI personas using a custom GPT in ChatGPT, and we see real potential in specific use cases. But there are clear limitations - while AI personas can simulate demographics and psychographics, they can’t reflect lived experience and the full cultural context.

This means they can’t replace real qualitative participants. While there are AI training schools teaching marketers how to create synthetic participants and invite them to an AI moderated focus group (and no doubt some DIYers will experiment with this), we don’t believe that AI can currently replicate the full context of people’s lives - the messiness, the contradictions, the System 1 thinking and irrationality that good qualitative research looks to uncover.

The use cases we currently see are at the front and back end of a project – to shape the research design and stimulus (which we discuss here) and to deliver insights (a future topic - the use of AI in Insight Dissemination). AI Personas can help researchers think though potential participants’ mindsets, refine hypotheses and even pressure-test very early stage ideas. In other words, help fine tune the research design. When developing stimulus, they can be used as valuable ‘co-pilots’ to fine-tune stimulus, crafting language and visuals that resonate with the intended audience. This approach can also be used to localise stimulus for global projects, where ideas have the potential to get ‘lost in translation’.

Where AI Can Add Value in Qualitative Research Design

AI is becoming an invaluable tool in research design - not as a replacement for qualitative expertise, but to enhance efficiency, structure, and strategic thinking in the research process. Whether it’s structuring a research brief, refining discussion guides, or developing stimulus, GenAI can streamline processes, provide fresh perspectives, and create frameworks that support better decision-making. I’ve shared some possible use cases for AI Personas as part of this process – firmly in the design stage, simulating humans to create a sharper design, before we talk to the real humans.

AI’s role in research design is evolving, and every researcher will integrate it differently. The key is to experiment - finding ways AI can complement your approach while maintaining the integrity of qualitative inquiry. At its best, AI is a tool to enhance thinking, streamline workflows, and free up time for what matters most: human insight.

Next week, I’ll shift focus to one of the big changes currently happening in Qual - AI Moderation (also commonly now referred to as ‘Qual at Scale’). See you then.