Home · Business · Customer Service · Complaints & escalations

PARTLY

As of 13 August 2026, AI can only partly find the root cause of customer complaints.

Most people should hand this to a purpose-built tool.

Can you do it?

15 minutesto a draft.

2 hoursto something you’d act on.

Cost, all in£0

Skill neededpower-user

Who has to check ita colleague

What the alternative costsA human complaints analyst or operations team can investigate the causes using internal systems and subject knowledge; no price is stated here.

If this goes wrong: you fix a visible complaint theme while the underlying process problem continues and more customers are affected.

What to actually do

  1. Use a tool built for this

    The route this page recommends

  2. Hand it to a person

    Second choice

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

  3. Do it yourself

    The distant third

    A chat interface, power-user skill, and roughly 2 hours until you can act on the result.

    How to actually do it

    1. Export a representative set of anonymised complaint records, including the complaint text, record ID, product or service, channel, date, outcome and any available journey or process stage.
    2. Gather the current policies, scripts, process maps, product information, incident records and operational metrics that could explain the complaints.
    3. Remove unnecessary names, contact details, payment information and other personal data before pasting the material into an approved AI tool.
    4. Paste the supplied prompt and attach the anonymised complaint records and source material, then ask the model to keep evidence, inference and missing information separate.
    5. Compare each proposed root cause with the cited complaint records, process documents and operational metrics, and remove any claim that the sources do not support.
    6. Ask the relevant service, product or operations colleague to test the remaining causes against live systems and select a validation action for each one.
    7. Record the confirmed cause, rejected hypotheses, evidence and owner in the complaints process, then use the agreed action and later complaint data to check whether the cause was actually addressed.

    Prompt

    Analyse the customer complaint records below to identify likely root causes, not just complaint topics. Separate direct evidence from inference. Group complaints into distinct themes, quote or reference representative examples using the supplied record IDs, identify the customer journey stage and internal process involved, and suggest the most plausible underlying cause for each theme. Include alternative explanations, missing evidence, the records or metrics needed to test each explanation, and a proposed validation action. Do not invent facts, volumes, dates, policies or causes. Do not treat a customer's stated explanation as proven causation. Flag issues that need a human investigation, legal review or urgent operational action. Return a table with these columns: theme, likely root cause, evidence, alternative explanations, missing evidence, validation action, owner, and confidence. Complaint records: [PASTE ANONYMISED COMPLAINT RECORDS]. Relevant policies, process maps, product information and operational data: [PASTE SOURCE MATERIAL].

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

What it gets wrong

  • AI cannot see undocumented workarounds, team incentives or process failures that are absent from the material you provide.
  • It confuses a frequent symptom with a root cause unless you supply operational evidence and ask for competing explanations.
  • It cannot prove that a process change caused complaints to fall without a suitable comparison or follow-up measurement.
  • It does not take responsibility for customer harm, regulatory consequences or the cost of fixing the wrong problem.
  • It can expose personal complaint details if you paste records into a tool without following your organisation's data-handling rules.

What caps this at PARTLY: judgement under ambiguity, context depth 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.

AxisScore (0–2)
Output2
Inputs1
Verification1
Liability1
Effort delta2
Total7 / 10

The methodology and its thresholds are published in full.

FAQ

Can ChatGPT analyse customer complaints?
Yes, it can group complaints, summarise patterns and suggest likely causes from the records you provide. It cannot establish the true cause on its own, so a colleague must check the evidence against processes, systems and operational data.
Can AI identify patterns in customer complaints?
Yes. It can find repeated phrases, products, journey stages and failure themes across a set of complaints. Patterns are leads for investigation, not proof that one issue caused the complaints.
How do I use AI to find the root cause of complaints?
Give it anonymised complaint records together with relevant policies, process maps and operational data. Ask it to cite evidence, list alternative explanations and propose a way to validate each cause before anyone changes the service.
Is it safe to upload customer complaints to AI?
Only use a tool approved by your organisation and remove unnecessary personal and confidential information first. Check the tool's data-handling terms and your internal privacy process before uploading complaint records.

Nearby answers

All Complaints & escalations 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.