Addy Waichigo

Director Systems Assurance and Data Science SAI Kenya

Abstract

Artificial Intelligence (AI) and Machine Learning (ML) are revamping public sector auditing by the introduction of data-driven insights, and enhanced anomaly detection. In the context of the Supreme Audit Institution (SAI) Kenya, AI and ML have the potential to transform the audit process, enabling efficient and thorough auditing while addressing complex high-risk audit components. This article explores the opportunities and challenges of adopting these technologies in public sector auditing, emphasising practical applications, legislative frameworks, and strategies for implementation including data protection.

Introduction

Artificial Intelligence is defined as the ability of machines to mimic human capabilities. Machine learning (ML) on the other hand is a subset of AI, where systems can learn from data and improve performance based on data provided. In this context, the advent of AI and ML offers opportunities to modernize processes, enhancing accuracy, efficiency and insightful analysis.

As Supreme Audit Institutions (SAIs) play a critical role in ensuring accountability and transparency in the use of public resources, the adoption of AI and ML would transform the audit process.

1. Role of AI and ML in Public Sector Auditing

AI and ML have practical applications in the audit process that would enhance the following:

a) Risk assessment: this is a critical phase in audit planning, where auditors will prioritise high-risk audit components. The use of AI and ML would enhance this process by:

• Machine learning algorithms can analyse historical audit findings, operational data and external factors to predict certain audit areas that most likely exhibit high risks within a sector.
• AI can assign risk scores based on unusually large transactions, recurring vendor issues and one-off transactions. These scores would help auditors to prioritise the audits and ensure efficient allocation of resources.

b) Fraud detection: Detection of fraud using traditional audit methods is time-consuming and often reactive and most likely detected after losses have been incurred. Fraud detection would be enhanced through:

• AI can learn from transaction patterns and flag outliers that may indicate fraudulent activities.
• Through analysing historical data, AI can identify red flags and unusual patterns, which are commonly associated with fraud. This will enable auditors to make actionable recommendations on preventive measures.

c) Data Analytics: Public sector entities generate large amounts of data, both structured and unstructured, from diverse systems. Traditional data analysis methods would struggle to handle such volumes efficiently leading to errors and omissions. AI and ML addresses these limitations by:

• AI tools can process millions of transactions across multiple databases in real-time, enabling auditors to efficiently test 100% of the data instead of relying on sampling. This approach improves the accuracy of audits, reduces the likelihood of missing critical anomalies and significantly mitigates the audit risk.
• Use of AI in automation of repetitive tasks such as reconciliations or review of financial statements would allow auditors to focus on strategic analysis and decision-making.

2. Opportunities for AI and ML in Public Sector Auditing

The integration of AI and ML into public sector auditing offers transformative opportunities that would enable SAIs to enhance efficiency, accuracy and effectiveness in their oversight role. The key opportunities are:

a) Increased efficiency in the audit processes

The automation of repetitive tasks and real-time processing of transactional data reduces the time needed to detect issues and enables auditors to focus more on high-value activities including strategic planning.

For instance, SAI Kenya can automate the review of budget expenditure reports for compliance with the Public Finance Management Act. This would reduce manual workload and improve turnaround times.

b) Enhanced audit quality and accuracy

Data-driven insights reduce reliance on sampling and ensure all data is reviewed thus increasing the accuracy of audit findings. Also, the standardization of reviews ensures consistency in audit procedures and findings by applying uniform standards across all data sets.

For instance, SAI Kenya extracts and analyses all the transactional data from the main Integrated Financial Management Information System (IFMIS) and generates standardized entity reports on transactional data. These reports are shared with the audit teams for either substantive testing or audit findings. This process also provides insights into high-risk items.

c) Enhanced stakeholder engagement

AI tools can generate summaries of audit reports, translating technical findings into easy-to-understand language for stakeholders. In addition, the use of data visualization through AI-powered tools can create intuitive and interactive dashboards, which would make it easier for stakeholders to better understand the results.

For instance, SAIs can use generative AI to make their reports more accessible to stakeholders, where they are able to efficiently obtain summarized reports for several financial years.

3. Challenges of AI and ML in Public Sector Auditing

Although the integration of AI and ML offers vast potential to transform public sector auditing, there are challenges that need to be addressed.

a) Data quality and availability

There is an inherent risk in the public sector data, which is mostly incomplete, outdated and inaccurate. These factors make it difficult for AI tools and systems to generate reliable results. The limited access to data and data silos may also have an impact on the seamless integration and analysis of the data by AI systems. Also, the use of handwritten records would complicate the processing of documents by AI.

b) Skill gaps within the SAI audit teams

The SAI staff lack technical skills to design, develop, implement and interpret the AI models effectively and this may lead to over-reliance on external vendors. Therefore, there is a need for capacity building of auditors on the use of AI to reduce the skill gap.

c) Complexity of AI systems

AI and ML models are not transparent processes, and this can erode stakeholder trust due to difficulty in understanding how decisions are made. Also, some of the off-the-shelf AI solutions may not align with the specific requirements related to public sector auditing. This non-alignment may require extensive customizations of the AI systems that would increase the cost and implementation time.

4. Conclusion

AI and ML are solutions that present an opportunity for SAIs to modernise and enhance their audit processes. The challenges of adopting AI and ML in public sector auditing are multifaceted and require strategic commitment, robust training programs and collaboration with stakeholders. By addressing the challenges, SAIs are able to harness the transformative potential of AI, ensuring audits are efficient, accurate and transparent.