Audit workflow

Design a workflow for audit, leading to better efficiency

Comment

Project Story: AI Agent for Audit Workflow and Excel Automation

1. User Story

  • As an Internal Auditor / Financial Compliance Analyst,
  • I want to leverage an AI Agent to automate repetitive audit procedures (experiments) and handle related Excel data processing,
  • So that I can reduce the time spent on tedious manual data reconciliation and focus on investigating high-risk anomalies.

2. Background & Challenges

In traditional auditing, verifying compliance and identifying risks requires performing numerous "audit experiments" (e.g., three-way matching, ledger-to-subledger reconciliation, and transactional consistency checks).

However, this process is often hindered by several operational challenges:

  • Tedious and Time-Consuming Tasks: Audit experiments require significant manual effort to clean data, perform VLOOKUPs, merge multiple Excel workbooks, and apply formulas.
  • Highly Repetitive Nature: Many audit tests are structurally identical but must be repeated across different quarters, business units, or entities, leading to redundant manual labor.
  • Limited Audit Scope: Due to the constraints of manual processing, audit teams are often forced to rely on limited sample sizes, potentially missing anomalies in untested data.

3. Proposed Solution (The Agentic Workflow)

To address these challenges, this project introduces an AI Agent-based workflow designed to assist auditors with data preparation and standard test execution.

The workflow is structured as follows:

[Audit Request] -> [Agent Task Planning] -> [Automated Excel Processing] -> [Iterative Test Execution] -> [Discrepancy Reporting (Excel)] -> [Auditor Review]
  • Task Parsing: The Agent interprets natural language instructions (e.g., "Compare the Q3 sales ledger with ERP records and identify discrepancies greater than $100").
  • Automated Excel Processing: The Agent utilizes data processing tools (such as Python/Pandas or Excel APIs) to automatically read, clean, and merge Excel spreadsheets, eliminating the need for manual sheet manipulation.
  • Iterative Audit Experiments: The Agent runs defined test logics across multiple datasets or periods, documenting the results systematically.
  • Discrepancy Flagging: The Agent outputs a structured Excel working paper where discrepancies and potential anomalies are highlighted, preparing the data for final review by the audit team.

4. Expected Outcomes

By integrating the AI Agent into the audit workflow, the project aims to deliver the following improvements:

  • Reduced Manual Effort: Minimizes the time auditors spend on routine data gathering, cleaning, and basic spreadsheet operations.
  • Expanded Test Coverage: Enables the automated testing of larger datasets and more frequent intervals, helping to identify risks that might be missed in smaller samples.
  • Standardized Procedures: Establishes a consistent execution method for standard audit tests, reducing minor errors due to manual fatigue while keeping the final qualitative evaluation in the hands of experienced auditors.

Built With

Share this project:

Updates