Ethnographic Insight Over Stated Preference
Observe what customers do, not what they say — AI transcript-reading is not true customer insight
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 3
- Confidence
- 86%
A caution about AI-assisted customer research. There is a difference between customer discovery (what customers say) and true insight (what they actually do). AI is excellent at transcribing and summarizing interviews, but using it to read all your interview transcripts is not customer insight — people don't always say what they mean, and may not realize they're 'lying' about why they do things. Real insight comes from ethnographic observation: watching people, instrumenting how they finish tasks, and understanding the why behind their actions.
Origin
Ezinne Udezue's lesson from running a UX lab earlier in her career, offered as a corrective to over-reliance on AI in customer interviews.
Core principles
- 01What customers do is more important than what they say
- 02Customer discovery (asking) is distinct from true insight (observing) — and AI is good at the asking part
- 03Using AI to read all interview transcripts is not customer insight; it can give you junk
- 04People don't always say what they mean, and may not realize they're lying about their reasons
- 05True intimacy with a customer comes from walking in their shoes and observing, not from what they claim they want
How to run it
- 1
Separate discovery from insight
Treat asking customers questions (discovery) as distinct from understanding what they truly do (insight). Let AI accelerate transcription and discovery, but don't mistake that output for insight.
Watch out AI reading all your customer-interview transcripts is not customer insight — it will give you junk if you stop there.
- 2
Observe behavior ethnographically
Watch what people actually do. The digital equivalent of a UX lab is core instrumentation — figure out how people actually finish tasks rather than relying on what they report.
Pro tip Ethnographic research — watching what people truly do — 'still is topnotch, the best thing you can ever do.'
Watch out What people say they do and say they want is not the true intimacy you need.
- 3
Understand the why behind the action
Whatever it takes, work out how to walk in the customer's shoes and understand the driver behind their need or problem, not just the surface behavior.
Pro tip This is a lesson people usually have to fail their way into learning — 'everyone told me they wanted this and nobody used it.'
Watch out Believing 'AI has come to save it all' on customer understanding leads to shipping things nobody actually uses.
In the wild
Early in her career Ezinne ran a UX lab doing ethnographic research — watching what people truly did. She argues the digital version today is core instrumentation: figuring out how people actually finish things, rather than reading AI summaries of what they said in interviews.
→ Observed behavior yields true insight that stated preferences and AI transcript summaries cannot.
Common mistakes
Cheating customer research with AI transcript reading
Using AI to read all interview transcripts and thinking you've figured out the customer produces junk, because people don't always say what they mean or realize they're lying about their reasons.
Trusting stated preferences over observed behavior
What customers say they do and say they want is not true intimacy; teams that rely on it ship features everyone claimed to want but nobody uses.
Is it for you?
Best for
Product teams doing customer discovery who are tempted to substitute AI transcript analysis for real behavioral observation
Not ideal for
Purely quantitative optimization work where large-scale stated-preference or survey data is genuinely the appropriate input
From the transcript
“there's a difference between what the customer says, right? Customer discovery, asking them things, right? And AI is really good at that right now”
“true customer insight comes from not using AI to read all the transcripts of your customer interviews. That isn't what it is. It's actually trying…”
“it's not what they say they do. It's not what they say they want. Those are that's not the true intimacy you want with a…”
“people think AI has come to save it all, but no, it's going to give you junk because people don't always say what they mean”
“ethnographic research watching what people truly were doing still is topnotch. It's the best thing you can ever do.”
From the episode
How AI is reshaping the product role
Oji and Ezinne Udezue