[Demo] Stress-testing GenAI in user research synthesis
Summary
In this high-pressure scenario, the challenge was to conduct 17 user interviews in three days and synthesize a comprehensive report in just one additional day. I’ll explore how we used AI to streamline the research process, from transcription to synthesis, and how tools like ChatGPT contributed to efficient data processing and insight generation. We’ll reflect on the potential and pitfalls of using AI in accelerated user research, from practical aspects to more philosophical considerations on potential changes to the research process. Takeaways Practical insights into integrating AI with traditional research methodologies to expedite the research process An overview of the effectiveness of AI transcription and synthesis tools in real-world research scenarios Critical examination of AI's role in data processing and how it compares with human analysis Strategic considerations for service designers when employing AI to support rapid user research Reflection on the ethical implications and potential impact on the quality of insights and researcher well-being when relying on AI to speed up research processes
Key Insights
-
•
Generative AI speeds textual content generation and navigation but needs human guidance for meaningful interpretation.
-
•
AI struggles to track and assign specific interviewees to user archetypes without explicit human instruction.
-
•
A semi-automated workflow combining AI brainstorming and human refinement delivers faster synthesis under tight deadlines.
-
•
ChatGPT excels at brainstorming clustering parameters and generating detailed archetype descriptions when properly guided.
-
•
AI can retrieve relevant interview quotes but often requires multiple prompt iterations to produce longer, contextually fitting excerpts.
-
•
Visual generation of archetypes using DALL·E currently yields poor results and often contradicts instructions.
-
•
Token limits can constrain transcript processing, but newer models like GPT-4 improve handling of larger inputs.
-
•
AI behaves like an 'alien intern'—powerful but unaware of design research processes and simple context details unless taught.
-
•
Using AI as a creative partner helps overcome cognitive overload, especially in rapid, solo research scenarios.
-
•
Less experienced researchers require caution when using AI tools, as expert evaluation is crucial for output quality.
Notable Quotes
"Gene AI can really help to speed up synthesis moments connected to textual content generation and navigation."
"AI behaves like an alien intern because it’s not capable of making sense of some things easy for us."
"I approached the GPT as if it were my human colleague, asking very open questions."
"It couldn’t find specific interviewees for user archetypes, so I had to adapt and guide the process."
"This is a hero moment because it accelerates and simplifies exploration of clustering options."
"Text generation is the biggest superpower of ChatGPT in research synthesis right now."
"Visuals from DALL·E didn’t work well; I asked multiple times to remove cables but they still appeared."
"AI can help manage cognitive overload and speed up work when you have short timeframes and no team."
"You need a lot of annoying handholding with the AI: reminding it of how many people you have or simple things."
"I wouldn’t leave AI tools to juniors because it’s crucial to have your own experience to evaluate outputs."
Or choose a question:
More Videos
"The relationship between designers and data scientists can actually be pretty magical."
Helen ArmstrongAugment the Human. Interrogate the System.
June 7, 2023
"The New European Bauhaus combines aesthetics, inclusivity, and climate-neutral ambitions to transform built environments by 2050."
Christian BasonExpand—Rethinking Design for Public Challenges
September 14, 2022
"Prost's master plan imposed a heavy interventionist burden on the historical structure of the city."
Onur Kocan Ayhan EnsiciUnderstanding the Strategy for Civic Design in a Complex City: Istanbul
November 16, 2022
"It’s much easier to classify an answer than to generate an answer, and that’s why LLM as a judge works."
Peter Van DijckBuilding impactful AI products for design and product leaders, Part 2: Evals are your moat
July 23, 2025
"What happens if research does not report into UX? What might happen if we uncouple research from UX?"
Nalini KotamrajuResearch After UX
March 25, 2024
"Are the questions that you are asking and the stories that you are telling yourself serving you?"
Brendan JarvisFraming Tomorrow by Questioning Today
June 8, 2022
"Our job as civic designers is to bring government needs and community needs into alignment."
Bria Alexander Ariel Kennan Charlotte Lee Sarah Brooks Emily Lessard Gordon Ross Joanne DongReflect and Chart Forward
December 10, 2021
"Leaders should look for ways to get firsthand experience with these tools, not just rely on media."
Frances Yllana Jorge Arango Maria Taylor Briana ThomasThe Big Question about Impact: A Panel Discussion
September 24, 2024
"Hope is not passive. Hope is very proactive and requires facilitation that helps people dream despite their harsh realities."
Liz EbengoThe Burden on Children: The Cost of Insufficient Post-Conflict Services and Pathways Forward
December 4, 2024
