I use an AI assistant to group interview notes into themes. It is fast, and the first pass is usually close.

Then I read every quote in every group, because it gets two things wrong often enough to matter.

It merges people who disagree

Two people who both mention fees land in the same theme, even when one is angry about fees and the other picked us because of them. The theme looks strong. It is really two opposite findings.

It misses the quiet ones

Themes come from what people said most. The person who said one sentence about using a screen reader, and nothing else, does not make a theme. That sentence changed our date picker.

Where it helps

Tagging, a first draft of the theme list, and finding the quote I half remember. It saves me about a day per study, and I spend part of that day reading what it grouped.