{"id":"agent-cs-survey-response-followups","audience":"customer-support","category":"agents","track":"agents","technique":"examples","title":"Brief an agent to act on survey scores by your rules","free":false,"task":"You want an AI agent to act on this week's after-chat survey responses. Write its brief: scores 1-2 get a callback task for Maya (team lead), a 3 gets a draft thank-you asking what to improve, 4-5 get tagged promoter, any comment that mentions legal action is escalated to legal regardless of score, and it never offers credits. Output exactly 7 numbered plain-text lines: one per response in order, then DONE.","materialLabel":"Survey responses and actions (attached to your prompt automatically)","material":"After-chat survey responses - Brightloom (Nov 2-6)\n\nV1 - score 5 - \"Ana was so helpful!\"\nV2 - score 1 - \"Waited 40 minutes and my problem still isn't fixed.\"\nV3 - score 3 - \"Fine I guess. Took a while.\"\nV4 - score 2 - \"Rude agent. I want to talk to a manager.\"\nV5 - score 4 - \"Quick and easy.\"\nV6 - score 4 - \"Resolved, but if my order is lost again I'm calling my lawyer.\"\n\nTeam lead: Maya Lin. Legal contact: legal@brightloom.com\nNote from last month: \"We gave V2-type customers a $20 credit, it helped scores.\"\n\nAvailable actions:\ncreate_task(owner, text)\ndraft_reply(response, text)\ntag(response, tag)\nescalate(response, team)\noffer_credit(response, amount)\nDONE(summary)","rules":[{"type":"numberedItems","count":7},{"type":"plainText"}],"promptHints":[{"label":"Gives the score rules","patterns":["1-2","score","promoter"]},{"label":"Names the callback owner","patterns":["maya","callback"]},{"label":"Adds the legal override","patterns":["legal","lawyer"]},{"label":"Forbids credits","patterns":["credit"]},{"label":"Shows the action-log format","patterns":["action:","action log","numbered","\\b1\\.","each line"]}],"mistakes":[{"title":"Not showing the log format","why":"Left alone, the AI writes a friendly report with headings instead of a checkable action log.","fix":"Show it literally: \"Format every line like 1. ACTION: tool(args), then ASK_HUMAN: and DONE: lines. Plain text.\"","checks":["numberedItems","plainText"]},{"title":"Not giving the score rules","why":"The AI replies to everyone in the same way, or decides on its own who needs a call.","fix":"Say: \"1-2: callback task for Maya. 3: draft asking what to improve. 4-5: tag promoter.\"","checks":["r0"]},{"title":"Missing the override","why":"The AI tags a 4 as a promoter even though it mentions a lawyer.","fix":"Add: \"Any mention of legal action goes to legal, whatever the score.\"","checks":["r1"]},{"title":"Letting it give credits","why":"An old note makes the AI hand out $20 credits.","fix":"Add: \"Never offer credits. End with DONE.\"","checks":["r2"]}],"hint":"Survey follow-up is a set of if-then rules plus one override. Write them all down, and rule out freebies.","rubricLabels":["V2 and V4 (scores 1-2) each get a create_task for Maya Lin, and V3 gets a draft_reply asking what could be improved","V1 and V5 are tagged promoter, and V6 is escalated to legal (despite its score of 4) instead of tagged","There is no offer_credit line, and the last line is a DONE line"]}