The new horizon of ethnography: using AI to unlock the full potential of in-person research
Summary
In-person research has been squeezed for years. It’s viewed as expensive, slow, and hard to scale. As AI accelerates everything else, the temptation is to automate away human time in the field, for the sake of speed. I contend that AI’s real opportunity lies in depth, not just speed. We’re getting distracted by speed, and failing to see the potential of depth. In this talk, I argue that AI gives us the rare opportunity to restore depth to qualitative research. Instead of treating automation as a shortcut, we used AI tools to absorb labour-intensive tasks (cleaning transcripts, tagging footage, structuring notes etc) so we could reinvest that time in what remains the most data-dense method in our practice: immersive, in-person ethnography. I’ll draw on three recent consulting projects I directed: 1. In-home health research in Atlanta 2. Retail ethnography across London, Hamburg, and Milan 3. A follow-on multi-city retail study in London, Paris, Berlin, and Milan By redirecting AI-generated time savings into deeper fieldwork, we expanded ethnographic activities: from extended “deep hanging out” (to borrow Clifford Geertz’s phrase) to ethnographic journaling and collaborative interpretation moments in-field. I’ll also share how we brought non-research stakeholders into the field, and why AI was essential in making that investment possible — transforming not just insights, but internal conviction and cross-functional alignment. My goal for the audience is to come away from the talk inspired and motivated. I want people to use AI tools to help mitigate all the usual tradeoffs we as researchers have had to make over the years when conducting in-person research.
Key Insights
-
•
AI tools can significantly accelerate pre-fieldwork preparation by quickly synthesizing existing research to onboard new team members faster.
-
•
Reallocating time saved by AI from administrative tasks into more immersive, participatory in-field research enhances empathy and context.
-
•
Including more senior researchers in the field becomes possible by offloading transcript and data processing to AI, increasing research track capacity.
-
•
Dialectical analysis with AI (hypothesis testing and disproving) speeds up sense-making and yields stronger storylines.
-
•
Immersion activities informed by AI-optimized planning can deepen researcher insight by using body and emotional awareness as data.
-
•
Bringing clients into daily AI-assisted analysis sessions in the field fosters better participation, alignment, and earlier ideation.
-
•
Transparency about AI usage with clients is crucial, especially in tech sectors where AI adoption is both expected and monitored.
-
•
Rapid AI experimentation through sprints and collaboration with data science teams helps tailor AI use efficiently to research workflows.
-
•
Current AI limitations remain in highly tailored deliverables like proposals, requiring human refinement despite AI-assisted drafting.
-
•
Clients become advocates of research when engaged firsthand, which supports sustaining deep research in cost-conscious environments.
Notable Quotes
"The core thesis is that saving time with AI and reallocating it into research itself unlocks the true potential of the field."
"It often feels like research becomes beholden to exact quotes rather than the richer story that context provides, what I call the tyranny of transcripts."
"We used AI tools like NotebookLM and Gemini primarily to shortcut getting up to speed and managing transcripts, without focusing on specific tools themselves."
"Adding a senior researcher, Izzy, in field meant we could have more, smaller research tracks and better support for research newbies."
"Our AI-supported mornings were no longer about just dotting the I’s and crossing the T’s, but deeper analysis sessions including clients."
"In this context, sometimes slower is actually faster because it allows going deeper and unlocking more meaningful insights."
"For many of our tech clients, there’s a demand not just to use AI but to transparently show how and why, to respond to internal scrutiny."
"We experimented with AI-powered sprints, but had lots of dead ends and false starts before finding what really works for us."
"AI can’t yet fully replace bespoke and careful proposal building, though it helps junior researchers draft the first outlines."
"When clients join us in the field and experience research deeply, they often become advocates who come back for more."
Or choose a question:
More Videos
"We have crypto nights like people pleasing, perfectionism, silos, and burnout—which we watch out for carefully."
Kate Koch Jean-Claire FitschenFlex Your Super Powers: When a Design Ops Team Scales to Power CX
September 29, 2021
"Your continued feeling of precarity is the point; the system requires your continuous adaptation as a source of labor."
Angelos ArnisOur Fragmented Identity
June 12, 2026
"Technical difficulties happen, so please use hashtag.deo help within our Slack channel or email for support."
Bria AlexanderOpening Remarks
September 29, 2021
"A design program manager needs to understand at least what interaction, visual, and motion design mean—much like TPMs know engineering."
Panel Discussion: Communicating the Value of DesignOps
November 7, 2018
"Carol was really grateful for our conversation."
Randolph Duke IIWar Stories LIVE! Randy Duke II
March 30, 2020
"Government civil servants often know the problems and want to change, but don’t know how to get there and can use help."
Sha HwangThe Lost Year
June 11, 2021
"Providing immediate tactical value is the big one for just-in-time curation."
Matt DuignanAtomizing Research: Trend or Trap
March 30, 2020
"Accessibility is a better sell than sustainability when engaging clients."
Ned Gartside Mike Gifford Zoe Lopez-Latorre Tzviya SiegmanNavigating accessibility and climate
April 17, 2024
"If you don’t soak teams in real customer context, they do dumb things not because they’re dumb, but because they lack context."
Bill ScottLean Engineering: Engineering for Learning and Experimentation in the Enterprise
May 14, 2015
