概述
It matters because internal audit teams are expected to cover more risk with limited staff. The IIA's Global Internal Audit Standards still hold auditors responsible for evidence, judgment and confidentiality.
深入探討
Internal audit gives a board and senior management independent assurance that risks are managed and controls work. The Institute of Internal Auditors (IIA) sets the profession's standards. Its Global Internal Audit Standards took effect in January 2025, replacing the previous framework. They require a risk-based audit plan, sufficient and reliable evidence, and protection of confidential information, and all of this applies when AI is used. Continuous auditing is internal audit's use of automated, recurring tests on system data. Examples include daily checks for duplicate payments, for conflicts where one person holds duties that should be separated, or for changes to vendor bank details followed by payment. It differs from continuous monitoring, which is management's own ongoing oversight. Under the IIA's Three Lines Model, management owns controls and monitoring, while internal audit provides independent assurance. If internal audit builds a monitoring tool that management then relies on, it should hand over ownership, or it risks auditing its own work. Risk-based planning is the second major use. Instead of building the annual plan mainly from interviews, AI can combine key risk indicators, incident logs, prior findings, control test results and outside signals. It ranks areas to audit and flags when a risk changes during the year. The chief audit executive still decides the plan and must be able to explain it. Generative AI helps draft audit programs, summarize policies and turn workpaper notes into draft findings structured around criteria, condition, cause and effect. The risks are drafts that include statements the evidence does not support, and confidential data pasted into public tools. Internal audit is also increasingly asked to audit AI itself, including model governance, data quality and bias. The IIA has published an AI auditing framework to support this work. A common misconception is that continuous auditing replaces the audit plan. It is one input to it.
現實世界的實施
A nightly script checks the ERP for vendor bank detail changes followed by a payment within seven days, and sends each hit to an internal auditor to follow up with accounts payable.
The audit team feeds key risk indicators, incident logs and prior findings into a risk ranking. Midyear, it moves a planned facilities audit back and brings forward an audit of a fast-growing third-party payments program.
After fieldwork, an auditor uses an approved enterprise AI tool to turn workpaper notes into draft findings structured as criteria, condition, cause and effect. The manager then checks each statement against the evidence.
內部審計審查公司如何治理信用評分模型,檢查資料品質、監控偏差以及誰批准模型變更。
The Future of AI in Internal Audit
內部審計團隊可能會更依賴持續測試和資料驅動規劃,並花更多時間審計其組織部署的 AI 系統。較小的部門可能會從生成式工具在文件編寫方面獲得最大收益,但它們也最缺乏檢查這些工具的能力。技能組合正轉向資料分析、技術風險與溝通。AI 能夠負責多少報告與實地工作仍在調整中。專業標準對證據、獨立性與機密性的要求將繼續設定界限。
付諸實踐
為什麼這很重要
- 背景與規則. 產業背景決定了人工智慧創意能否與現實接觸。
- 品質管控. 領域約束會影響可接受的錯誤率和監督模型。
- 配裝選擇. 成功的部署使技術能力與第一線工作流程保持一致。
留意
- 監理要求可能會使原本強大的原型失效。
- 歷史資料可能會編碼損害特定社區的偏見。
- 遺留系統可能會造成整合瓶頸和隱性成本。
後續步驟
- 讓領域專家參與從問題框架到評估的整個過程。
- 在啟動前設計審計追蹤和文件。
- 儘早驗證合規性和安全義務。
- 分階段推出,並有明確的停止和回滾標準。
不斷探索
常見問題
什麼是內部審計中的 AI?
內部審計中的 AI 指使用分析、機器學習與生成式 AI 進行持續控制測試,將審計計畫聚焦於最高風險,並加速測試、文件編寫與報告草擬。這很重要,因為內部審計團隊預期在有限人力下覆蓋更多風險。IIA 的全球內部審計標準仍要求審計師負責證據、判斷與機密性。
在指南中,持續審計與持續監控有何區別?
根據三線模型,管理層負責監控,內部審計則透過持續審計等活動提供獨立保證。
內部審計建立監控儀表板,管理層開始將其作為控制依賴。指南建議什麼?
若內部審計持續執行管理層依賴的控制,將冒著審計自身工作的風險。將其移交可保護客觀性。
哪個 IIA 模型說明管理層擁有控制,內部審計提供獨立保證?
IIA 的三線模型描述了這些角色,指南利用它將監控與審計分離。
IIA 的全球內部審計標準何時生效?
全球內部審計標準於 2025 年 1 月生效,取代先前的框架。
指南中的哪個持續審計測試針對常見的支付詐騙模式?
對供應商銀行資訊的變更後迅速進行付款,可能表示付款被重新導向。指南以此作為腳本化測試的例子。