How AI Is Transforming UX Research: From Design to Quality Control

Yogiyo's UX Research team integrated AI into Research Operations (ReOps), boosting both the speed and depth of qualitative research at the same time. By offloading the repetitive work of re-listening to and organizing countless transcripts to AI, work that used to take nearly 6 weeks was completed in 4 weeks, saving about 25% of the time.

At the research design stage, the team used ChatGPT as a tool to turn business requirements into research questions. Rather than accepting vague requests as-is, they checked the logical gaps in hypotheses and reviewed survey/interview questions for leading bias, structuring problems before kickoff.

At the data analysis stage, they used Gemini Pro to systematize large volumes of interview scripts. Rather than simple summarization, they analyzed qualitative data through the following 5-stage prompt pipeline.

• Phase 1: Break down statements into units of meaning, and remove questions and filler words to refine only the data
• Phase 2: Tag statements with Pain Point, Needs, Motivation, Barrier, and Value to quantify qualitative data
• Phase 3: Build an Affinity Diagram based on the tagged statements and derive core themes
• Phase 4: Validate hypotheses based solely on themes and raw data, while checking supporting/opposing evidence together
• Phase 5: Connect insights to product strategy and Action Items

Reports and decision-making materials were refined with NotebookLM. Thanks to a grounded generation approach that finds evidence within the uploaded raw data, they were able to build slide structures based on actual user statements rather than AI-fabricated sentences, and also used a method of generating the same material multiple times to cross-validate common keywords.

Finally, they set AI up as a senior researcher coach to check the quality of interview moderating as well. By having it self-review leading questions, points needing further probing, and lack of neutrality, they used AI not as simple automation but as a collaborator that expands the researcher's thinking.