How to Follow Up on NPS Scores With AI Interviews
Qualitati Research Team · 2026-09-22 · 9 min read
Short answer: To follow up on NPS scores with AI interviews, invite a sample of detractors, passives, and promoters to a short conversational interview within days of their rating. Start with their own verbatim comment, ask about a specific recent episode, and probe what would change their score. Then code the transcripts by segment so the "why" behind the number reaches the teams that can act on it.
Last updated: September 22, 2026
Following up on NPS scores with AI interviews solves the oldest complaint about Net Promoter Score: the number tells you that loyalty moved, not why. Most programs attach one open-ended "why?" box to the rating, collect a few words per respondent, and then stall. This guide shows how to turn that one-line comment into a real conversation at survey scale, with a sampling plan, a follow-up interview template, and an analysis workflow you can run next week.
Key takeaways
- NPS is a useful tracking metric, but on its own it is a weak diagnostic. Qualitative follow-up is where the causes live.
- Interview across all three segments. Passives are often the most informative and the most ignored.
- Anchor every interview in the respondent's own comment and a concrete recent episode, not in the score.
- Never ask people to justify or raise their number. That is leading, and it contaminates both the interview and the metric.
- Report themes by segment with quotes, and route each theme to an owner.
Why the NPS "why?" box is not enough
NPS asks customers how likely they are to recommend a company on a 0 to 10 scale. According to the Net Promoter System site maintained by Bain & Company, promoters score 9 or 10, passives 7 or 8, and detractors 0 to 6; the score is the percentage of promoters minus the percentage of detractors. Fred Reichheld introduced the metric in Harvard Business Review in December 2003.
The metric has well-known limits. A longitudinal study by Keiningham and colleagues in the Journal of Marketing (2007) did not find NPS to be a better predictor of company growth than other satisfaction measures. Nielsen Norman Group's June 2024 guide to NPS for UX teams notes that the score says little about specific parts of an experience and recommends pairing it with behavioral and qualitative data that explain why people rated as they did.
The open-ended "why?" field is the usual answer, but it has three practical problems:
- It is short. Many comments are a few words ("pricing", "support is slow") with no episode, no context, and no severity.
- It is one-shot. Nobody can ask "slow compared with what?" or "what happened the last time you contacted support?"
- It is skewed. Angry detractors and delighted promoters write; passives often leave it blank.
How to follow up on NPS scores with AI interviews: a 5-step workflow
An AI-moderated follow-up interview is a short, asynchronous conversation (text or voice) in which an AI moderator asks a planned set of questions and generates neutral probes from each answer. Because it runs without scheduling, you can talk to far more respondents than a human team can call back.
Step 1: Decide what decision the follow-up serves
Write one sentence: "We will use these interviews to decide ___." Examples: which onboarding problems to fix first, why passives in one market are not converting to promoters, or whether a pricing change drove the latest drop. A follow-up with no decision attached produces a pile of quotes and no action.
Step 2: Sample across all three segments
Invite respondents from every segment rather than only detractors. A simple starting quota is roughly equal numbers of detractors, passives, and promoters, then oversample any segment or market tied to your decision. Invite within a few days of the rating, while the experience is fresh, and state the time commitment honestly.
Step 3: Build the interview from the respondent's own words
Pipe the score and the verbatim comment into the opening so the moderator starts where the customer left off. Keep the guide to five or six core questions; the value comes from probing.
Step 4: Set probing rules, not just questions
Tell the moderator how to probe: ask for a specific recent example, ask what they did next, ask what they compared you with, and reuse the participant's own terms. Forbid evaluative language and any request to reconsider the score.
Step 5: Code by segment and route to owners
Code transcripts into themes, then compare theme frequency and severity across detractors, passives, and promoters. A theme that appears in all three segments is structural; one concentrated in detractors is a likely churn driver. Assign each theme an owner and a date to revisit.
NPS Follow-Up Interview Template
This is an original Qualitati template. Adapt the wording to your product and language.
| Stage | Question | Probe rule |
|---|
| Anchor | "You mentioned [verbatim comment]. Can you tell me more about what you had in mind?" | Use their exact words; do not paraphrase into your own framing. |
| Episode | "Think of the last time you used [product] for [job]. Walk me through what happened." | Ask for sequence and outcome, not opinions. |
| Comparison | "What did you use or consider before, or alongside, us?" | Probe what the alternative does better or worse, in their terms. |
| Impact | "How did that affect your work or your plans with us?" | Ask for consequences: time lost, workarounds, renewal thoughts. |
| Recommendation context | "If a colleague asked you about [product], what would you tell them?" | Captures the actual word of mouth behind the score. |
| Change | "What, if anything, would have to be different for your experience to feel better?" | Never ask "what would make you give us a 10?" |
Decision matrix: which follow-up method fits?
| Method | Depth | Scale | Best for |
|---|
| Open-ended "why?" field only | Low | All respondents | Tracking broad topics over time |
| Conversational survey with AI follow-ups | Medium | High | Adding one or two clarifying probes to every comment |
| AI-moderated follow-up interview | Medium to high | Dozens to hundreds | Explaining segment differences and finding root causes |
| Human callback interview | High | Low | Key accounts, sensitive complaints, relationship repair |
What to do with the results
Keep the number and the narrative together. A useful NPS readout pairs the trend with three to five themes, each with its segment distribution, two or three participant-anchored quotes, and an owner. If you already code open-ended survey text, see our comparison of LLM and machine-learning coding for open-ended answers and our guide to sentiment analysis vs. thematic analysis. When detractor themes point to cancellation risk, move into dedicated churn interviews.
Where Qualitati fits
Qualitati is an AI user research platform for product, UX, and customer insights teams. For NPS follow-up, you can run AI-moderated interviews in text or voice, or a conversational survey with AI-driven follow-up questions and branching logic, in 10 languages. ThemeLens then performs thematic analysis across up to 100 transcripts at once, mapping codes to your research questions and synthesizing themes with participant-anchored quotes, and the QDA Workspace supports inductive and deductive coding if you want to code by segment yourself. The free tier includes 30 credits on signup with no credit card required, and usage rates are published on the pricing page.
Qualitati does not replace your NPS survey tool; it handles the qualitative follow-up that explains the score.
Limitations and methodology concerns
- Self-selection. People who accept a follow-up invitation differ from those who do not. Report response rates by segment and avoid treating theme frequencies as population estimates.
- Consent and contact rules. Only invite customers who agreed to be contacted, and explain that the interview is for research, not a sales or retention call. See our note on informed consent in AI-moderated research.
- Score contamination. If follow-ups feel like pressure to change a rating, future NPS responses become less honest. Keep research follow-up separate from service recovery.
- AI moderators can still lead. Review early transcripts for evaluative or suggestive probes before scaling fieldwork.
- Automated themes need review. Treat AI-generated themes as a first draft that a researcher checks against the transcripts.
Who this is for — and when not to use this approach
This workflow suits product managers, CX and customer insights teams, and UX researchers who already run NPS and want to explain movements in it. Skip it when you have very few responses (call them yourself), when a complaint is an urgent service issue (route it to support), or when you need a validated measure of satisfaction rather than explanations.
FAQ
Should I interview promoters, or only detractors?
Interview all three segments. Promoters show what to protect, and passives often reveal the small frictions that keep them from recommending you.
How soon after the NPS survey should I follow up?
Within a few days is a practical target, so the experience behind the rating is still easy to recall.
How many follow-up interviews do I need?
It depends on how many segments and markets you compare. Plan enough per segment to reach thematic saturation for your decision, and check whether new interviews still add themes.
Can I use the NPS comment as an interview question?
Yes. Quoting the respondent's own comment is a strong, neutral opener. Avoid adding your interpretation of it.
Is it acceptable to ask what would raise their score?
No. Ask what would make the experience better. Asking about the score invites people to bargain with the number instead of describing their experience.
Bottom line
NPS tells you where loyalty is moving; interviews tell you why. Following up on NPS scores with AI interviews lets you talk to all three segments at scale, starting from customers' own words and ending with themes that have owners. Start free with 30 credits to run an AI-moderated NPS follow-up, or view transparent pricing. For question wording, see our guide to avoiding leading questions.
Human-review note: the template and decision matrix are editorial frameworks from the Qualitati Research Team, not a validated instrument. Adapt them to your study design and local contact rules.