Home · Business · Customer Service · Complaints & escalations

PARTLY

As of 13 August 2026, AI can only partly respond to a public social media complaint.

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 ita colleague

What the alternative costsA support agent such as Intercom Fin resolves support conversations from your help content; its price is not provided here.

If this goes wrong: the public reply exposes private details, makes an unsupported promise or turns a manageable complaint into a wider reputational problem.

What to actually do

  1. Use a tool built for this

    The route this page recommends

  2. Do it yourself

    Second choice

    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 public post and copy the complaint, including the visible context and any replies that affect what the customer is asking.
    2. Open the relevant customer record, order or case file and gather only facts that a colleague has confirmed, removing account numbers, addresses and other private data.
    3. Open your current complaints, refunds, privacy and social-media policies, then paste the relevant sections into the prompt with the outcome you are authorised to offer.
    4. Paste the complaint, verified facts, policy text and preferred next step into the prompt and ask for the public reply, private message and escalation note.
    5. Compare every factual statement, apology, promise and proposed remedy against the case record and current policy, deleting anything unsupported.
    6. Ask a colleague who can approve public customer communications to check the tone, privacy, escalation decision and proposed remedy before publishing.
    7. Publish only the approved public reply, move case details to a private channel, and record the complaint and response in the normal case system.

    Prompt

    Draft a public reply to this social media complaint.

Complaint:
[PASTE THE PUBLIC COMPLAINT]

Verified case facts:
[PASTE ONLY FACTS CONFIRMED BY OUR RECORDS]

Our relevant policy or service information:
[PASTE THE CURRENT POLICY, PRICE OR SERVICE DETAILS]

Our preferred next step:
[FOR EXAMPLE, ASK THE CUSTOMER TO SEND A PRIVATE MESSAGE, PROVIDE A CASE REFERENCE, OR WAIT FOR A PERSON TO CONTACT THEM]

Write:
1. One public reply of no more than 60 words.
2. One private follow-up message of no more than 100 words, if private contact is appropriate.
3. A short internal note listing any fact you cannot verify, any promise the reply would make, and whether the complaint should be escalated.

Use a calm, plain and respectful UK customer-service tone. Acknowledge the customer's experience without admitting fault unless the verified facts support it. Do not invent facts, compensation, deadlines, refunds, legal conclusions or policy exceptions. Do not repeat personal data or ask the customer to post order details publicly. If the complaint involves safety, discrimination, threats, serious harm, legal action or a vulnerable customer, say that a trained colleague should handle it instead of drafting a final public reply.

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

  3. Hand it to a person

    The distant third

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

What it gets wrong

  • AI cannot see the full customer history, internal conversations or operational constraints unless you provide them.
  • AI cannot decide reliably whether an ambiguous complaint needs a refund, a formal investigation or senior escalation.
  • AI cannot judge the relationship with a particular customer or the reputational effect of publishing a response.
  • AI cannot take responsibility for an inaccurate admission, disclosure or promise made in the organisation's name.
  • AI cannot safely handle serious allegations, threats, safety issues or vulnerable-customer situations without human ownership.

What caps this at PARTLY: judgement under ambiguity, stakes of error and relationship.

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 AI write a reply to a complaint on social media?
Yes, it can draft a clear public reply and a private follow-up from the facts and policies you provide. A colleague should check it before publication because your organisation remains responsible for every public claim and promise.
Should I let AI reply directly to customers on social media?
Usually not for complaints without a human approval step. AI can miss relevant case history, reveal private information or offer a remedy that your organisation has not authorised.
What should an AI-generated response to a public complaint say?
It should acknowledge the customer's experience, state only verified facts and move account details into a private channel. It should not invent explanations, admit fault without evidence or promise a refund, deadline or exception that nobody has approved.
Can AI handle an angry customer complaint?
It can help draft a calm first response, but anger alone does not tell you what action is appropriate. Escalate complaints involving safety, discrimination, threats, legal action, serious harm or vulnerable customers to a trained colleague.

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.