{"id":"search-mgr-compare-two-ai-answers-pilot","audience":"managers","category":"ai-search","track":"research","technique":"grounding","title":"Compare two AI answers about a pilot against the real report","free":false,"task":"Two AI tools summarized your company's flexible-Friday pilot report differently, and you need to know which one to forward to your VP. Write exactly 4 numbered lines labeled A error:, B error:, Both missed: and Verdict:, judged only against the report. The Verdict line names A or B and must come last. Plain text, under 100 words.","materialLabel":"The two AI answers and the pilot report (attached to your prompt automatically)","material":"ANSWER A (Tool 1)\nThe flexible-Friday pilot ran for 6 months across 4 teams. Productivity, measured as tickets closed, rose 12%. Employee satisfaction went up sharply. Sick days fell by half.\n\nANSWER B (Tool 2)\nThe pilot ran for 3 months across 4 teams. Tickets closed rose 12%. Satisfaction scores rose from 6.8 to 7.9. Customer response times got faster.\n\nPILOT REPORT (People Ops, Sept 2026)\n- Length: 3 months (June-August 2026), 4 teams, 46 people.\n- Tickets closed per person: up 12% vs. the same months last year.\n- Satisfaction survey: 6.8 to 7.9 out of 10.\n- Sick days: down 9%.\n- Customer first-response time: 4% slower on Fridays; unchanged other days.\n- Limits: summer months are usually slower; no control group.\n- Recommendation: extend to 6 months, add a control group.","rules":[{"type":"numberedItems","count":4},{"type":"maxWords","max":100},{"type":"plainText"}],"mistakes":[{"title":"Not naming the judge","why":"The AI compares the two answers to each other or to what sounds right, instead of to the report.","fix":"Add: \"Judge both only against the pilot report.\"","checks":["r0"]},{"title":"Skipping what both missed","why":"The AI only lists differences, so the shared gap (no control group) never comes up.","fix":"Add a line: \"Both missed: what neither answer mentions.\"","checks":["r1"]},{"title":"No clear verdict","why":"The AI says \"both have strengths\" and leaves you to decide.","fix":"Add: \"Last line: Verdict: A or B.\"","checks":["r2"]},{"title":"No fixed shape","why":"The AI writes a long side-by-side table you can't forward.","fix":"Say: \"Exactly 4 numbered lines labeled A error:, B error:, Both missed:, Verdict:.\"","checks":["numberedItems","maxWords"]},{"title":"Getting markdown back","why":"AI defaults to bold labels and headings.","fix":"Ask for \"plain text, no bold.\"","checks":["plainText"]}],"promptHints":[{"label":"Uses the report as the judge","patterns":["against the report","only (the|from the) report","pilot report"]},{"label":"Names the 4 labels","patterns":["a error","b error","both missed","verdict"]},{"label":"Asks for a verdict last","patterns":["verdict","which one","last line"]},{"label":"Sets the length","patterns":["\\b(4|four)\\b.{0,20}lines","under \\d+ words","exactly"]},{"label":"Asks for plain text","patterns":["plain","no bold","no markdown"]}],"hint":"Two confident answers can both be wrong. Make the report the only judge and ask for a verdict you can act on.","rubricLabels":["The A error line says A got the length wrong (6 months instead of 3) and/or sick days (half instead of 9%), and the B error line says B claimed faster customer response times when Friday responses were 4% slower","The Both missed line mentions the report's limits (summer months or no control group)","The last line is Verdict: and names B as the better one to forward (A has more errors), with no other text after it"]}