Stop 12 · AI & Modern Design ToolsTopics · 2/7

AI for UX Research (Synthesis & Summarization)

In short

Ten interviews can mean 300 pages of notes. AI can transcribe and group them in minutes, but it can't tell which finding will change your product.

Using AI to speed up transcription, thematic clustering, and summarization of research notes — accelerating the mechanical parts of synthesis while still requiring human judgment on what actually matters.

The jaadu sorts a pile of colourful sticky notes into baskets with a small robot helper.

Think of it like this

The dhobi's sorting pile

A dhobi sorts a mountain of clothes into whites, colours and silks before washing. A helper can do the sorting fast. But only someone who looks closely notices the one shirt with a torn pocket. AI sorts your notes; you spot what matters.

Watch it happen

Where AI helps in synthesis, and where you step in

Transcribe

Tag

Cluster

Check

Insight

1/5AI turns recordings into text. Fix names and jargon it mishears before using it.

The big ideas

What you need to know

Saves grunt work

Transcripts, first-pass tags and summaries are where AI saves the most hours.

Check against quotes

Every theme should link back to real quotes. If you can't find them, the theme may be invented.

Protect participants

Remove names and personal details before uploading. Check your company's rules on which tools are allowed.

Watch for missing voices

Summaries favour what was said most. The rare, surprising comment is often the real insight.

See it in code

Walk through the code

text

1You are helping me synthesise user interviews.2Context: 8 interviews with college students about3ordering food from the campus canteen.4Task: Group the quotes below into 4-6 themes.5For each theme give:6- a short name7- 2-3 exact quotes with participant IDs (P1-P8)8- how many participants mentioned it9Do not invent quotes. If a quote fits nowhere, list it10under "Other".11[paste anonymised quotes here]

1/3Role and context tell the model what kind of research this is.

A synthesis prompt that asks for evidence, not just a neat summary. The 'exact quotes' rule makes it easy to check the AI's work.

Save yourself the pain

Common mistakes

Oops

Copying the AI summary into your report without reading the transcripts.

Do this instead

Read at least a sample of raw notes yourself. Spot-check that each theme's quotes are real and in context.

Oops

Uploading recordings with names, phone numbers and faces into any free tool.

Do this instead

Anonymise first and get consent that covers this use. Follow your team's data rules.

Your mission

Human vs AI clustering

~40 min

Quick check

The AI gives you a theme called 'Users want dark mode' but you can't find matching quotes. What should you do?

Go deeper

1Use an AI tool to summarize a set of interview notes or a transcript.
2Compare the AI-generated themes against your own manual affinity mapping.
3Learn where AI synthesis tends to miss nuance that a human researcher would catch.