Same transcripts, two synthesis passes. Where do you agree, and what does each of you miss?
What
Synthesise a small set of interviews yourself, then run the same transcripts through an AI tool and compare the two.
Why
AI synthesis looks finished, which makes it tempting to trust. Comparing it with your own analysis shows you exactly where it helps and where it flattens the nuance. Doing yours first means the AI can't anchor your thinking.
How
1Pick 4 to 8 transcripts and anonymise them.
2Do your own affinity map or thematic analysis without looking at any AI output.
3Give the tool the same transcripts and ask:
Identify the key themes across these interviews. For each theme, give 2 to 3 direct quotes as evidence. Then list any points where participants contradicted each other or themselves.
4Put the two sets of findings side by side.
What to watch for
- ○Which themes overlap?
- ○What did the AI miss, especially the gap between what people said and what they did?
- ○Did it find something you'd unconsciously dismissed?
- ○Are the quotes real and in context, or paraphrased into something nobody said?
Tried it?
Would you use AI as a first pass, a second opinion, or not at all?
Tell us what happened, good or bad. Other members are trying it too.
Discuss in our LinkedIn Group