AI has revolutionized software development, but a major transformation is also underway in design and user research. Transcription, data processing, and grouping findings have undergone significant changes over the past year.
In Fraktio’s webinar, our Design Lead Antti Lavio shared how AI can support the different phases of a user research project. We’ve gathered the key insights that anyone conducting research should take into account. You can watch the webinar recording at the end of this blog post.
Where Are We Now?
User research produces vast amounts of text content — interview transcripts, notes, analysis, and documentation. This is exactly where language models shine. They process and produce text efficiently, and over the past year, AI tools have advanced rapidly. A year ago, AI was mainly generating content but not changing ways of working. Now the situation is different. New tools enable entirely new ways of working with data.
Organizations have started to understand better where AI works well and where it doesn’t. The fear that AI would “devour the entire design process” has given way to a more realistic understanding: AI works best in different parts of the process, combined with human expertise.
Planning Phase
Good user research always begins with a clear objective. What decisions need to be made? What business questions need answers?
At this stage, AI can help discover new perspectives, identify gaps in knowledge, and review previous research or customer feedback. It can help refine the formulation of research questions and help break down an overly broad topic into manageable parts.
It’s important to remember, however, that AI doesn’t understand an organization’s internal politics, resources, or strategy in the same way a human does. It doesn’t replace professional expertise, but it can serve as an effective thinking aid.
Creating hypotheses is especially important in research. AI can help crystallize ideas and generate alternative hypotheses. During the analysis phase, you can then ask AI to look for things that contradict or confirm the hypothesis. This enables deeper analysis than a mere data summary.
GDPR and permissions are critical. Before feeding any interview material into AI, make sure you have permission to do so. Discuss with the relevant parties what tools are being used, what data can be entered into them, and how it is handled. Sometimes permission may not be granted for business reasons — in that case, stick with traditional methods.
Execution Phase
In the interview setting, AI plays mainly a supporting role. The interview itself is a human process where building trust and genuine interaction are key.
AI can help with writing invitations or creating interview guides. It can rephrase questions, check for leading questions, and even simulate an interview in advance. But the actual interview situation requires human presence.
Why Is a Human Irreplaceable in the Interview Setting?
AI doesn’t see facial expressions, gestures, or frustration the way a human does. It processes text, not behavior. A user might say “yeah” but shake their head and look frustrated. AI only processes the transcribed speech — a human observes the whole picture.
This is why it’s important to:
- Continue making your own notes during the interview
- Record observations about behavior (“the user shook their head,” “looked frustrated”)
- Record the interview for later analysis, so you can return to a specific moment
Never rely solely on AI’s automatic notes. They can be a useful addition, but without human observations, you lose critical context.
Analysis Phase
If any phase has changed radically, it’s the analysis phase. Previously, a researcher had to read through all interviews multiple times and manually search for common themes. Now the workflow has become more conversational.
Transcription is one of the biggest benefits. Tools like Condense.io automatically transcribe interviews and identify discussed themes. This saves an enormous amount of time.
Conversing with data is the new normal. You can upload material into, for example, Notebook LM, and ask:
- “What did users say about payments?”
- “What themes recur in the responses?”
- “On what topic were users in disagreement with each other?”
- “Who said what about which topic?”
You can compare multiple interviews, different rounds, and even combine usability test results into the same analysis. This enables much deeper and faster mining than before.
But beware of superficiality. AI has a tendency to please and offer readily available answers. It may:
- Emphasize things that were talked about the most — not necessarily the most important things
- Latch onto positive comments superficially, even though deeper discussion would reveal problems
- Overlook things users didn’t say but that are significant
AI does not do qualitative analysis. It produces summaries. The human performs the actual analysis — connecting the dots, thinking systemically, and understanding context.
Context and Behavior Are Not AI’s Strengths
In user research, context is everything. It determines how and why a person acts. Context can be:
- Situational context: Where and when does the user work?
- Psychological context: What are the user’s emotions, motivations, and fears?
- Systemic context: How do the organization’s processes, culture, or situation affect things?
AI is not good at combining systemic and psychological context the way a human — and especially an experienced interviewer — can.
People say one thing and do another. AI believes the transcribed text, while an experienced interviewer notices the contradiction between what is said and what is done. They see gestures, hear tones of voice, and understand when a user isn’t telling the whole story.
AI doesn’t recognize what wasn’t said. Often the most significant findings come from the fact that a user didn’t mention something important at all. This requires human observation and comparison with other interviews.
Work Methods Change, the Goal Doesn’t
The work method has shifted from simply reading through material to more conversational research, but the goal remains the same: to understand users deeply and produce knowledge that helps make better decisions. AI is not the goal itself — it is merely a means to make your own work more efficient and free up time for thinking.
Want to explore how AI can be used in user research? Don’t hesitate to get in touch, or watch the webinar recording (in Finnish, English subtitles available) for more practical tips, including tool recommendations.
