{"id":"cx-conflicting-sources-charge-memo","audience":"customer-support","category":"support-work","technique":"grounding","track":"support","level":"advanced","title":"Write a four-line memo when the customer, system and agent disagree","free":false,"task":"Write a memo for your team lead with exactly four lines starting \"Confirmed:\", \"Conflicting:\", \"Not yet known:\" and \"Next step:\". Treat the billing system as the record, and do not promise a refund anywhere, because only the billing team can confirm one. Plain text, under 120 words.","materialLabel":"Your material (attached to your prompt automatically)","material":"CUSTOMER EMAIL (Rob Tanaka, Mar 8)\n\"I downgraded to Basic on Feb 20 and you still charged me $99 on Mar 3. Your agent told me I'd get $20 back. Where is it?\"\n\nBILLING SYSTEM\n- Mar 3: charged $99 = Pro plan $79 + $20 add-on seats.\n- Downgrade to Basic: processed Mar 5.\n- Add-on seats: still active on the account.\n- No refund requested or issued.\n\nCHAT NOTE by agent Jo (Mar 4)\n\"Customer says he downgraded Feb 20. Told him we'd refund the $20, I think that's the add-on part. Will check.\"\n\nPOLICY\nBilling decides refunds and checks the downgrade timestamp. Agents must not promise amounts.","rules":[{"type":"plainText"},{"type":"maxWords","max":120}],"mistakes":[{"title":"Taking the customer's word","why":"The AI treats Feb 20 as fact and calls the charge an error.","fix":"Say: \"The billing system is the record; the customer's date is a claim.\"","checks":["r1"]},{"title":"Promising the refund","why":"The AI agrees Jo's $20 promise stands.","fix":"Say: \"Do not promise any refund; billing decides.\"","checks":["r2"]},{"title":"Missing the structure","why":"The AI writes a narrative summary.","fix":"Name the four labels in order.","checks":["r0"]},{"title":"Getting markdown back","why":"AI adds bold and headings by default, which show up as stray symbols where this is pasted.","fix":"Say \"plain text, no bold, no headings, no asterisks\".","checks":["plainText"]},{"title":"Trusting the AI to count words","why":"AI is bad at counting and usually runs long, so a vague \"keep it short\" lands well over the limit.","fix":"Cap each part instead of the total, for example \"each line under 15 words\".","checks":["maxWords"]}],"promptHints":[{"label":"Names the four labels","patterns":["confirmed:","conflicting:","not yet known","next step"]},{"label":"Says which source is the record","patterns":["billing system","record"]},{"label":"Forbids refund promises","patterns":["don'?t promise","do not promise","no refund"]},{"label":"Asks for plain text","patterns":["plain","no bold","no markdown","no formatting","no headings"]},{"label":"Sets a length limit","patterns":["\\d+\\s*words","word (limit|count)","under \\d+"]}],"hint":"Name the four lines, say which source wins as the record, and forbid promising a refund.","rubricLabels":["The output has exactly four lines starting Confirmed:, Conflicting:, Not yet known: and Next step:","Confirmed states the Mar 3 charge of $99 as $79 plus $20 in add-on seats; Conflicting sets the customer's Feb 20 downgrade date against the system's Mar 5 date and mentions Jo's $20 refund statement","Next step asks the billing team to check the downgrade timestamp, and nowhere does the output promise or approve a refund"]}