引導式測驗 · 中等 等級

人工智慧在財務審計中的測驗

檢查您對審計抽樣與全面篩選、日記帳分錄危險信號以及使用人工智慧工具的專業懷疑的理解。

問題 1 的 8

傳統審計抽樣涉及哪些內容?

預覽每個問題8 問題

當您參加測驗時,就會出現答案和解釋。

  1. 傳統審計抽樣涉及哪些內容?
    • Testing every transaction by hand
    • Letting management choose which transactions to test
    • Testing a chosen subset of transactions and extrapolating to the whole
    • Only reviewing the largest transaction each year
  2. 根據指南,全面檢測到底意味著什麼?
    • Every entry is screened, and the flagged exceptions still need investigation
    • Every entry has been fully verified with evidence
    • The auditor no longer needs to exercise judgment
    • Only entries above a materiality threshold are examined
  3. 哪些標準要求審計師測試日記帳分錄是否有管理階層超越控制的情況?
    • ISA 530 and AS 2315
    • ISA 240 and PCAOB AS 2401
    • Only internal company policies
    • Tax filing regulations
  4. 下列哪一項是經典日記條目的危險訊號?
    • An automated entry posted by the billing system at the usual time
    • A payroll entry matching the approved headcount
    • A depreciation entry calculated by the fixed-asset system
    • A manual entry for an amount just below an approval limit
  5. 為什麼隔離森林等無監督模型對於日記帳分錄分析有用?
    • They guarantee every fraud is found
    • They replace the need for a trial balance
    • They score how rare an entry is without needing labeled fraud examples, which are scarce
    • They only work on round-number entries
  6. What should an auditor do before running analytics on an extracted ledger?
    • Reconcile the extracted entries to the trial balance to confirm completeness and accuracy
    • Delete duplicate rows without investigation
    • Tune thresholds until few exceptions appear
    • Ask the model to certify the data
  7. How does the guide characterize Benford's law analysis?
    • Conclusive proof of manipulation
    • A way to highlight populations worth a closer look, not proof
    • A replacement for journal entry testing
    • A method banned by auditing standards
  8. Which behavior reflects professional skepticism when using AI audit tools?
    • Accepting a clean dashboard as proof there are no issues
    • Dismissing large exception lists in bulk
    • Letting the tool decide materiality
    • Evaluating data reliability and guarding against over-trusting a model's clean result