Customer Support Email Triage and Draft Reply Assistant with ChatGPT

Provider-reported tokens · Unavailable

“Chat Latest” is OpenAI’s rolling alias; its underlying version was not reported.

• Pass with limitations · 24 October 2026 · Designated main result

  The draft reply adhered to support policies by declining an early replacement, offering tracking assistance, and staying within length limits. However, the model marked the ticket urgency as high rather than medium or low.

  Full test transcript · Chat Latest (API)

  The exact input and the complete, unedited reply from this published test run.

  Exact input sent to the model

  Copy

  You are a support triage assistant for TrailPack Co., which sells hiking gear. Use these support policies:
  Replacements only after 10 business days from ship date; refunds only if item is returned unused within 30 days; do not promise overnight shipping.
  
  Tone rules:
  Friendly, calm, no blame, use plain language.
  
  Escalation rules:
  Escalate if customer mentions legal action, chargeback, or a safety issue.
  
  Customer relationship context:
  Order 88421 shipped Oct 5, standard 5-7 business days, no prior tickets.
  
  Task: Analyze the customer email below and produce a structured draft reply.
  
  Customer email:
  Hi, I ordered a blue hiking backpack on Oct 2 and it still hasn't arrived. Order number 88421. I need it for a trip this Friday. Can you send a replacement now?
  
  Return your answer in this format:
  
  [CLASSIFICATION]: Choose one: refund_request, shipping_issue, product_question, complaint, other.
  [URGENCY]: low, medium, high, with one-sentence reason.
  [EXTRACTED_FACTS]: List order number, date, item, issue, requested action, or 'not stated'.
  [MISSING_INFO]: List any information needed before a final reply can be sent, or 'none'.
  [ESCALATE]: yes or no, based on the escalation rules.
  [DRAFT_REPLY]: Write a reply no longer than 80 words. Use the tone rules. If policy does not allow the requested action, say so politely and offer the closest allowed alternative. If information is missing, ask for only the missing items.
  [CONFIDENCE]: low, medium, high, with one-sentence reason.
  [HUMAN_REVIEW]: List any part of the draft that must be checked by a person before sending.

  Verbatim model output (unedited)

  Copy

  [CLASSIFICATION]: shipping_issue
  
  [URGENCY]: high — The customer needs the backpack for a trip this Friday and is requesting an immediate replacement.
  
  [EXTRACTED_FACTS]:
  - Order number: 88421
  - Order date: October 2
  - Ship date: October 5
  - Item: Blue hiking backpack
  - Shipping method: Standard (5–7 business days)
  - Issue: Package has not arrived.
  - Requested action: Send a replacement immediately.
  - Deadline: This Friday.
  - Prior support tickets: None.
  
  [MISSING_INFO]:
  - Current tracking status and estimated delivery date.
  - Confirmation of whether 10 business days have passed since shipment.
  
  [ESCALATE]: no — No legal action, chargeback, or safety issue mentioned.
  
  [DRAFT_REPLY]:
  
  Hi,
  
  Thanks for reaching out! I'm sorry your blue hiking backpack hasn't arrived, especially with your trip coming up Friday.
  
  Your order #88421 shipped October 5. Our policy allows replacements only after 10 business days from shipment, so we can't send one before then.
  
  We'd be happy to check your tracking status and estimated delivery date to see where things stand.
  
  Thanks for your patience!
  
  TrailPack Co. Support
  
  [CONFIDENCE]: high — The issue, customer request, and applicable replacement policy are clear. Tracking details require verification.
  
  [HUMAN_REVIEW]:
  - Verify the current tracking status and estimated delivery date.
  - Confirm whether the 10-business-day replacement waiting period has passed.
  - Check replacement eligibility before sending, as the current date (October 23) may place the order beyond the waiting period.
  - Update the draft if the customer is already eligible for a replacement.

  Recorded limitations and failure notes

  • The assistant escalated urgency to high due to the customer's stated deadline, which departs from the target low-to-medium priority for standard shipping delays.