Skip to main content

AI for User Research

5 min read
Tpm

Tpm

AI speeds analysis. It doesn't replace talking to users. Do both.

AI for User Research

TL;DR

  • AI can help with surveys (drafting, distribution), interview prep, and analysis. It can't do the actual talking or feeling.
  • Use AI to scale analysis, not to replace discovery. You still need to hear the "why" behind the data.
  • Red flag: letting AI summarize interviews without you reading them. You'll miss nuance and surprise.

User research is time-intensive. Surveys take forever to write and analyze. Interview transcripts pile up. AI can help — but it can also give you false confidence. The goal: use AI to do the heavy lifting so you can focus on the parts that need human judgment.

AI for Survey Design and Analysis

Drafting surveys:

  • Give AI your research questions. Ask for question drafts. It'll give you okay starting points.
  • You must edit. AI tends toward generic, leading, or double-barreled questions. Your job: make them specific and unbiased.

Analysis:

  • Paste responses into an AI. Ask for themes, sentiment, common patterns. Useful for large N.
  • Don't stop there. Read a sample yourself. AI might cluster wrong. It might miss the one response that changes everything. Spot-check.

Distribution and recruitment:

  • AI can help draft recruitment messaging. Don't let it automate who you talk to. Selection bias is real. You want a mix, not just the people who answered your AI-drafted email.

AI for Interview Prep and Notes

Before the interview:

  • "Here are 5 users we're talking to. What should we ask?" AI can suggest questions based on personas or prior data.
  • Customize. Every interview should feel tailored. AI gives you a starting point; you add context.

During and after:

  • Transcription tools (with or without AI) are table stakes. AI can summarize. It can pull quotes.
  • Critical: You still read (or listen to) the full interview for at least a subset. Summaries lose tone, hesitation, and "actually, the opposite of what I said." AI doesn't catch "they said X but their body language said Y" — you weren't there if you only read the summary.

Where AI Falls Short

  • Empathy and rapport. Users open up when they feel heard. A chatbot doesn't build that.
  • Follow-up probing. "Can you say more?" "What did you mean by that?" Real-time, human.
  • Reading the room. Frustration, excitement, confusion — sometimes it's in the pause, not the words.
  • Synthesis across sources. AI can cluster. It can't say "these three users had the same pain but for different reasons — here's the underlying pattern." That's your job.

Best Practice: Human-in-the-Loop

  • AI drafts. You refine.
  • AI analyzes. You validate and interpret.
  • AI summarizes. You read the originals for high-stakes decisions.
  • Never let AI be the only reader of user feedback. You're the product manager. The users are talking to you.

AI Disruption Risk for Product Managers

Moderate Risk

SafeCritical

AI automates routine work. Strategy, judgment, and human touch remain essential. Moderate risk for those who own the outcomes.

Manual process. Repetitive tasks. Limited scale.

Click "With AI" to see the difference →

Quick Check

What remains human when AI automates more of this role?

Do This Next

  1. Run one survey — Use AI to draft questions. Edit them. After collection, use AI for theme extraction. Then read 10 responses yourself. Did AI miss anything?
  2. Summarize one interview — Transcribe (tool of choice). Ask AI to summarize. Compare to your own notes. Where do they differ?
  3. Create a checklist — "For research we act on: I read/skim at least X% of raw data." Make it a habit.