Home · Business · Customer Service · Feedback & NPS analysis

YES

As of 13 August 2026, AI can detect changes in customer sentiment.

This still needs a person who signs their name to it.

Can you do it?

15 minutesto a draft.

30 minutesto something you’d act on.

Cost, all in£0

Skill neededchat-fluent

Who has to check ita colleague

What the alternative costsBotpress is an open platform for building LLM chatbots and agents.

If this goes wrong: you treat a change caused by sampling, wording or a temporary event as a real customer trend and make a poor service decision.

What to actually do

  1. Hand it to a person

    The route this page recommends

    A person who owns the outcome does this end to end, worth it when the failure is dear.

  2. Use a tool built for this

    Second choice

  3. Do it yourself

    The distant third

    A chat interface, chat-fluent skill, and roughly 30 minutes until you can act on the result.

    How to actually do it

    1. Open the source containing your customer feedback and export the relevant responses with their dates, survey scores, customer group and question wording.
    2. Split the data into two comparable periods and remove duplicate, test or clearly unusable responses without changing the remaining wording.
    3. Write down context that could affect the comparison, including product changes, service incidents, campaigns, changes in customer mix and changes to the survey question.
    4. Paste the context and both dated datasets into the prompt, keeping customer names, email addresses and other unnecessary personal information out.
    5. Ask the chatbot to produce the comparison table, evidence excerpts, uncertainty labels and verification checklist specified in the prompt.
    6. Compare every reported change with the original responses and scores, checking the response counts, dates, themes and quoted excerpts.
    7. Ask a colleague who knows the service to challenge the proposed alternative explanations, then record only the findings supported by both the data and that context.

    Prompt

    Analyse the customer feedback below for changes in sentiment between the periods provided.

Use only the supplied data and do not invent missing figures, causes or customer intentions. Keep the periods separate, report the number of responses in each period if supplied, and distinguish overall sentiment from changes in particular themes. Identify:
1. The direction and size of any apparent sentiment change.
2. The themes becoming more positive, more negative or staying similar.
3. Representative examples, quoting only short excerpts from the supplied feedback.
4. Possible alternative explanations, such as a change in response volume, question wording, customer mix or a one-off event.
5. Findings that are too uncertain to support a business decision.

Present the result as a table followed by a short plain-English summary. For every finding, show the evidence used and label it as strong, moderate or weak based on the supplied data, not on assumptions. Do not claim statistical significance unless the required data and analysis are available. End with a checklist of the raw responses, dates, scores and context a colleague should verify before acting.

Context:
[Describe the product, service, customer groups and any events that may affect feedback]

Period A:
[Paste dated feedback, survey answers, scores and response counts]

Period B:
[Paste dated feedback, survey answers, scores and response counts]

    Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.

What it gets wrong

  • AI cannot tell whether a sentiment shift reflects a genuine change in customer experience or a change in who responded.
  • It cannot reliably infer sarcasm, cultural context, coded language or the importance of a complaint from text alone.
  • It cannot establish that a trend is statistically meaningful without suitable data and analysis.
  • It cannot know which operational event caused the change unless you provide and verify that context.
  • It cannot take responsibility for the customer, staffing or product decisions made from the result.

Even on a YES, the friction has a name: judgement under ambiguity, context depth and verification cost.

How we scored this

Five axes, each scored nought to two by hand: ten means AI carries the task cleanly, and the thresholds that turn a total into YES, PARTLY or NO are published in the methodology. Each axis name links to its definition.

AxisScore (0–2)
Output2
Inputs2
Verification1
Liability1
Effort delta2
Total8 / 10

The methodology and its thresholds are published in full.

FAQ

Can AI analyse customer sentiment?
Yes. It can label feedback, group themes and compare sentiment across two supplied periods. You still need to check the raw responses and whether the periods and respondent groups are comparable.
Can AI tell if customer sentiment is improving?
It can identify an apparent improvement in the feedback you provide. It cannot prove that the change represents all customers or explain the cause without reliable context and a check of the underlying data.
How accurate is AI sentiment analysis?
Accuracy depends on the wording, language, sarcasm, data quality and labels used. Treat the output as a screening and comparison aid, then inspect representative responses before acting.
Can AI analyse NPS comments?
Yes. It can group NPS comments by theme and compare the language associated with different scores or periods. It should not replace checking the score calculations, response mix or original comments.

Nearby answers

All Feedback & NPS analysis rulings

Assessed by gpt-5.6-luna (gpt-5.6-luna) on 2026-08-13, second-checked by an independent model. Wrong somewhere? Email [email protected] and it gets re-checked.

The newsletter

AI news, new answers and product picks, straight to your inbox.