Turn transcripts and PDFs into themes and evidence tables

Use Claude Code to summarize interviews, build evidence tables, and draft an annotated bibliography from a study folder.

Done with Claude Code· Intermediate

What they did
A UX researcher organized interview files into a single folder, with one transcript per file and consistent filenames. They kept a CLAUDE.md file alongside the transcripts with the study questions, a codebook, and a rule that quotes must stay verbatim and include source filenames. Claude Code then processed the materials in small batches or with parallel subagents, wrote a themes.md file and evidence tables, and searched each quote back in its original transcript. PDFs were kept in a separate papers folder so Claude could pull together the bibliography.
How it went
The workflow cut down the time spent hunting through transcripts; the guide says one typed question can save about an hour of Ctrl+F. The researcher used this approach for studies with 50 transcripts. One audit found that 7.7% of quotes labeled as verbatim were not actually present in the source files.
Tips
Keep the codebook fixed before using parallel agents, or the theme lists can drift apart. Always do a quote-check pass, and remove identifying details before uploading because the file contents are sent to Anthropic.

Try it yourself with Claude Code

Review the interview transcripts in [transcript folder] and the related PDFs in [papers folder] for my study on [research question]. Use the codebook in [CLAUDE.md path] to group findings into themes, then create: 1) a theme summary, 2) an evidence table with one verbatim quote and source filename per theme, and 3) an annotated bibliography for the PDFs. Treat any theme backed by fewer than [number] sources as tentative. Before making any irreversible change, such as deleting, submitting, or sending files, stop and ask me to confirm. Save each output as a separate file in [output folder].

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Read the original on ccforeveryone.com ↗

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.