As of 13 August 2026, AI can analyse themes in your customer complaints.
Most people should hand this to a purpose-built tool.
Can you do it?
5 minutesto a draft.
30 minutesto something you’d act on.
Cost, all in£0
Skill neededchat-fluent
Who has to check ityou
What the alternative costsA purpose-built AI platform such as Botpress can be configured to work with business data and build AI workflows.
If this goes wrong: important complaints are merged into a vague category or a serious pattern is missed, so you make the wrong service change or fail to escalate a customer issue.
What to actually do
Use a tool built for this
The route this page recommends
Do it yourself
Second choiceA chat interface, chat-fluent skill, and roughly 30 minutes until you can act on the result.
How to actually do it
- Export the complaints from your helpdesk, CRM or spreadsheet, including the complaint text and any useful non-sensitive fields such as date, channel, product, category and outcome.
- Remove names, email addresses, phone numbers, order numbers, addresses and any other personal data that the analysis does not need, then assign each complaint a simple internal ID.
- Open a chatbot or AI analysis tool, paste the context about your products, existing complaint categories and the date range, and paste the anonymised complaint data.
- Paste the supplied prompt and ask the model to produce the theme report, keeping the data in manageable batches if the full export is too large to submit at once.
- Read every proposed theme and compare its sample complaints with the definition, moving complaints that have been mislabelled and recording any theme the model missed.
- Recalculate the theme counts and percentages in a spreadsheet from the complaint IDs, then compare those figures with the report and resolve discrepancies.
- Ask a colleague who understands the complaint process to review the flagged cases and the final themes before using them to change service, compensation or escalation rules.
Prompt
Analyse the customer complaints below and produce a decision-ready theme report. Treat each row as one complaint. Do not invent facts, causes, customer intentions or trends that are not supported by the text. First remove or ignore names, email addresses, phone numbers, order numbers and other unnecessary personal data in your analysis. Create a clear set of non-overlapping themes, allowing one complaint to have a primary theme and up to two secondary themes. For each theme, give its label, plain-English definition, number of complaints, percentage of all complaints, representative complaint IDs if supplied, and two short anonymised examples. Separate direct evidence from inference. Flag complaints that need human review because they involve threats, vulnerability, discrimination, safety, legal action, suspected fraud, compensation disputes or unclear meaning. Identify complaints that do not fit the main themes. State what cannot be concluded from this data, including any limits caused by missing dates, products, channels or outcomes. End with a short list of practical questions for a manager to investigate, not recommendations presented as proven facts. Show the arithmetic used for counts and percentages. Data period: [START DATE] to [END DATE]. Context about the business and fields: [PASTE CONTEXT]. Complaint data: [PASTE ANONYMISED DATA].
Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.
Hand it to a person
The distant thirdA person who owns the outcome does this end to end, worth it when the failure is dear.
What it gets wrong
- AI cannot reliably decide whether two complaints have the same underlying cause when the wording and circumstances differ.
- AI cannot tell you whether a small but serious pattern matters more than a large number of routine complaints.
- AI cannot recover context that is absent from the export, such as what staff promised, what happened before the complaint or why an outcome was agreed.
- AI cannot take responsibility for changing policies, compensating customers or escalating a complaint to a regulator.
- AI cannot guarantee that sensitive details have been handled lawfully if you paste the wrong data into a third-party service.
Even on a YES, the friction has a name: judgement under ambiguity, private data access and stakes of error.
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.
| Axis | Score (0–2) |
|---|---|
| Output | 2 |
| Inputs | 2 |
| Verification | 1 |
| Liability | 1 |
| Effort delta | 2 |
| Total | 8 / 10 |
The methodology and its thresholds are published in full.
FAQ
- Can ChatGPT analyse my customer complaints?
- Yes. It can group complaint text into themes, count examples and produce a summary if you provide clean, anonymised data and explain your existing categories. Check the labels and figures against the original complaints before acting on them.
- How do I use AI to find common complaints?
- Export the complaints, remove unnecessary personal data and give the model the complaint text with useful fields such as date, product and outcome. Ask it to define themes, show counts and examples, flag uncertain cases and state what the data cannot prove.
- Can AI tell me what is causing customer complaints?
- It can identify repeated descriptions and possible patterns, but it cannot prove the underlying cause from complaint text alone. Confirm suspected causes against operational records and speak to the staff or customers involved.
- Is it safe to upload customer complaints to AI?
- Only use a service and process approved for your organisation's data, and remove personal data that the analysis does not need. Check your privacy, security and retention requirements before uploading complaints, especially where they contain vulnerable customer details or allegations.
Nearby answers
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.