AI-Assisted User Interview

AI handles the volume work in user research — guide generation, transcription, theme extraction, quote retrieval. You bring the judgment — noticing the pause, reading the subtext, sensing what's left unsaid.

---

Context

Three modes:

Mode 1 — Pre-interview: Design the guide and prepare the PM

Mode 2 — Post-interview: Analyse a single transcript

Mode 3 — Cross-interview synthesis: Find patterns across multiple sessions

---

Mode 1: Pre-Interview

Step 1 — Understand the research objective

Ask: what decision this informs, who is being interviewed, current assumptions, session length, and type (discovery / validation / problem-depth).

Step 2 — Write the interview guide

Four sections: Warm-up (5 min, rapport), Current state exploration (15-20 min, what they actually do today), Problem/need exploration (15-20 min, underlying needs not feature requests), and Wrap-up (5 min, open-ended).

Step 3 — Prepare the PM

Pre-session checklist: clear research objective, no leading questions, recording consent, transcription tool, timer, assumptions list, exact quotes not paraphrases.

---

Mode 2: Post-Interview

Step 1 — Receive the transcript

Step 2 — Four extraction passes (extract without interpreting):

Exact quotes by topic

Jobs to be done (When [situation], they want to [motivation], so they can [outcome])

Frustrations and workarounds

Surprises (contradicts assumptions — most valuable signal)

Step 3 — Separate observation from interpretation in two columns

Step 4 — Output session summary with key quotes, JTBD, frustrations, surprises, observations vs. interpretations, new questions, and assumptions checked.

---

Mode 3: Cross-Interview Synthesis

Step 1 — Receive all transcripts or summaries

Step 2 — Extract cross-session patterns: Frequency (themes in multiple sessions with counts), Contradictions (where participants disagreed), Weak signals (1-2 sessions, worth watching).

Step 3 — Output synthesis report with top findings, secondary findings, contradictions, decision implications, unknowns, and recommended next steps.

Quality check before delivering

Suggested next step: Run aipm-research-synthesis if you have more than 4 sessions to synthesise — it handles larger research sets with theme clustering and contradiction mapping.