There are useful comments in the client feedback folder, but no consistent way to discuss them. You ask AI for the main themes and receive a neat story about what customers want. The story sounds broader than the evidence you supplied.
Use AI-assisted feedback analysis to propose themes linked to actual comments. Define the collection first, preserve disagreement and separate theme counts from claims about your whole customer base. Treat the result as a structured reading aid, not a survey you never conducted.
Photo by Jakub Żerdzicki on Unsplash.
State what the collection represents
Record where the comments came from, the period, the number of contributors and how they were selected. A folder of complaint emails answers a different question from responses to a consistent post-project questionnaire.
Do not blend sources without keeping their origin visible. A website review, an account-manager note and an interview quotation have different context. They may all be useful, but they are not interchangeable observations.
Only use material the approved tool is permitted to process. Remove unnecessary identifying details and keep the original evidence in its authorised location.
Create a theme card
For each proposed theme, ask for:
A plain-language description.
Comment identifiers supporting it.
Exact excerpts where appropriate and permitted.
Contradictory or qualifying comments.
Cases that fit more than one theme.
A question the business should investigate next.
Require the assistant to leave a theme unsupported if it cannot identify the source comments. Do not allow it to create representative quotations to make the report read better.
An illustrative set of mixed comments
Suppose several respondents mention slow onboarding. Some describe waiting for a welcome email. Others say the document request was confusing. Combining them under Slow service may hide two different problems.
Keep the subthemes separate long enough to understand the mechanism. The practical next step might be reviewing email timing and rewriting one request, rather than changing the entire service.
A positive comment about thorough explanations can coexist with a complaint about too many messages. Neither cancels the other. The useful question is which information helped, and at what stage it became excessive.
Count only what you have defined
If one person submits three comments, decide whether the unit is a comment or a contributor. Say whether a comment can appear in several themes. Otherwise totals can look more precise than they are.
A theme appearing often in this collection does not establish its prevalence among every client. Keep that limitation beside the finding, particularly when sources were self-selected or gathered for another purpose.
Turn themes into a small investigation
Select one theme, inspect its supporting examples and speak with the relevant process owner. Record what is known, what remains an interpretation and which change would address the underlying concern.
Keep the theme structure stable when comparing periods, or explain any revision. A renamed category can otherwise look like a change in customer experience when only the analysis changed.
For related practical guidance, see collecting useful client feedback and investigating recurring process problems.
Hili Consulting can help organise customer feedback into traceable themes with honest sample limits. Arrange a free 30-minute consultation and bring one example of the work you want to improve.
