Introduction

Artificial intelligence (AI) is revolutionizing auditing by enhancing accuracy, efficiency, and compliance in financial analysis. Quadratic AI and ChatGPT emerge as transformative tools for bank audits, addressing challenges like data confidentiality, audit sampling, and regulatory compliance. This paper demonstrates their application in auditing loan portfolios, leveraging AI for masking sensitive data, stratified sampling, and IRAC compliance analysis.

____________________________________________________________________________________________________

Problem Statement

A bank provided an Excel spreadsheet containing loan balance data for audit. Key challenges include:

- Data confidentiality: Masking sensitive fields (Account Number, Customer ID, Name).

- Efficient analysis: Filtering and sorting data per IRAC norms.

- Audit sampling: Implementing stratified random sampling across risk categories.

- Compliance: Identifying discrepancies and provisioning requirements.

____________________________________________________________________________________________________

Methodology

1.  Data Masking with ChatGPT

Objective: Mask sensitive fields while preserving data utility.

Implementation:

- Account Numbers: Use Excel formulas like =REPLACE(A2,5,4,"XXXX") to hide digits.

- Customer Names: Apply =LEFT(B2,1)&REPT("*",LEN(B2)-1) to retain initials.

- ChatGPT Integration: The model generates and troubleshoots formulas, reducing manual effort.

1.  Data Analysis with Quadratic AI

IRAC Filtering: Isolate FY2018–19 data using Python scripts in Quadratic for dynamic filtering.

1.  Stratified Random Sampling

Strata Design: Categorize loans into high, medium, and low-value categories and draw representative samples.

Quadratic Execution: 60% of the samples are high-value loans, 30% medium, and 10% low. This stratification helps focus on high-risk areas.

1. Compliance Analysis

IRAC Provisioning: Ensure compliance with IRAC norms by identifying non-compliant accounts and calculating provisioning.

Quadratic Role: Identify non-compliance patterns and generate actionable insights.

Results

1. Large number of SMA-4 and SMA-5 accounts indicate delayed classification
2. Need to verify if 90-day NPA norms are being strictly followed
3. Review required for accounts showing EMI dues but not classified as NPA

Benefits of AI-Driven Audit

- Accuracy: Reduced human error in data masking and sampling.

- Efficiency: Automated workflows cut analysis time by 40%.

- Scalability: Solutions adaptable to larger datasets and diverse regulations.

Conclusion

Quadratic AI and ChatGPT streamline bank audits through automated data handling, risk-based sampling, and compliance checks. Future work should explore AI ethics and real-time monitoring. This approach offers auditors a scalable, objective framework for financial analysis in the digital age. This approach is scalable and can be applied to various audit scenarios.