{"id":"risk-finding-rating-math","audience":"compliance","category":"legal-compliance","track":"legal","technique":"constraints","level":"advanced","title":"Apply a rating matrix to audit findings and show the math","free":false,"task":"Rate four findings using the matrix. Write a prompt for a markdown table Finding | Math | Proposed rating, rows A to D, under 130 words, no bold or headings. The Math cell shows the calculation like \"7 / 50 = 14%\" and any payment check. Ratings are only High, Medium or Low. The last line must read exactly: Ratings are proposals for the audit lead to confirm.","materialLabel":"Your material (attached to your prompt automatically)","material":"RATING MATRIX (Internal Audit methodology v2)\n- High: exceptions in more than 20% of items sampled, OR any exception on a payment above $50,000.\n- Medium: exceptions in 10% to 20% of items sampled (both ends included) and no exception on a payment above $50,000.\n- Low: exceptions in under 10% of items sampled and no payment above $50,000.\n\nFINDINGS\nA. Missing approvals: 7 exceptions in 50 items. Largest payment affected: $12,000.\nB. Late reconciliations: 3 exceptions in 30 items. Largest payment affected: none.\nC. Unsigned contracts: 6 exceptions in 25 items. Largest payment affected: $30,000.\nD. Duplicate vendor records: 1 exception in 40 items. That one payment was $62,000.","rules":[{"type":"plainText"},{"type":"maxWords","max":130}],"promptHints":[{"label":"Table columns","patterns":["finding","math","proposed rating"]},{"label":"Shows math","patterns":["show","calculation","\\d+\\s*/\\s*\\d+"]},{"label":"Boundary rule","patterns":["10%","payment","over \\$?50,?000|above"]},{"label":"Closing line","patterns":["proposals","last line"]},{"label":"Allowed ratings","patterns":["high","medium","low"]}],"mistakes":[{"title":"Boundary case missed","why":"AI rates B as Low because 10% sounds low.","fix":"Say: \"10% counts as Medium (both ends included).\"","checks":["r0"]},{"title":"Payment override missed","why":"AI rates D Low because only 1 in 40 had an exception.","fix":"Say: \"Any exception on a payment above $50,000 makes it High.\"","checks":["r1"]},{"title":"Presenting as final","why":"AI states ratings as decided.","fix":"Say: \"Last line exactly: Ratings are proposals for the audit lead to confirm.\"","checks":["r2"]},{"title":"Trusting the AI to count words","why":"AI is bad at counting and usually runs long, so a vague \"keep it short\" lands over the limit.","fix":"Cap each part instead of the total, for example \"each line under 15 words\".","checks":["maxWords"]},{"title":"Getting markdown back","why":"AI adds bold ratings and headings by default.","fix":"Say \"no bold, no headings\".","checks":["plainText"]}],"hint":"Show the math, state the boundary rule and give the exact closing line.","mission":"Rate four findings with the math shown.","musts":["Math in each row","Boundary rule followed","Ratings are proposals"],"rubricLabels":["The table has columns Finding, Math and Proposed rating with rows A, B, C and D, and every rating is exactly High, Medium or Low","The ratings are A Medium, B Medium, C High and D High, and the Math cells show the percentage calculations (14%, 10%, 24%, 2.5%) and the payment check for D","The last line reads exactly: Ratings are proposals for the audit lead to confirm."]}