AI

5 Use cases, Benefits & Challenges in 2023

5 Mins read

One remarkable development in artificial intelligence (AI) and machine learning has been ChatGPT, a language model capable of generating human-like text responses.

We’ve covered +40 ChatGPT’s use cases before, including:

  • Textual (i.e., content creation, translation)
  • Coding
  • Educational
  • Marketing & SEO
  • Customer service
  • Human resource

Today, we are looking at the top use cases of ChatGPT in audit, such as

  • Automating audit responses
  • Identifying anomalies 
  • Training and helping auditors

In addition, we will also explore the challenges of implementing ChatGPT in auditing practices and lay out a framework for establishing an effective collaboration between the human auditor and ChatGPT. 

5 Use Cases of ChatGPT in Audit

Reporting automation

Manual processes, such as reporting, could be time consuming and error-prone. ChatGPT can help auditors automate repetitive tasks, such as paperwork and reports. Specifically, it can: 

  1. Produce standardized reports (Figure 1) that offer consistency in how findings are presented
Figure 1: Introduction of ChatGPT generated sample audit report
Figure 1: Introduction of ChatGPT generated sample audit report

Data analysis

Audit programs involve the frequent analysis of large swaths of financial and operational data. 

ChatGPT can automate some of these data analysis duties, such as in: 

  • Performing computations 
  • Aggregations 
  • Dataset comparisons 

For example, auditors can engage with ChatGPT to input data and questions, and the model can then analyze the data to produce insights, trends, and patterns (Figure 2). This can save the time and effort required for manual data analysis, freeing up auditors to concentrate on analyzing the findings and coming to insightful conclusions.

Figure 2: The Introduction of a sample ChatGPT conducting data analysis in an audit report. Source: OpenAI
Figure 2: The Introduction of a sample ChatGPT conducting data analysis in an audit report. Source: OpenAI

Real-time risk monitoring

ChatGPT can also be helpful in real-time risk monitoring. Auditors can interact with the model to discuss the organization’s activities, control systems, and business environment. ChatGPT can assist auditors assess risk levels (Figure 3), identify priority areas for more investigation, and get insights into potential hazards. The accuracy and depth of risk assessment can be improved by auditors by utilizing the model’s knowledge and skills. Figure 3 displays 4 possible uses of ChatGPT in real-time risk monitoring:

Figure 3: Four possible uses of ChatGPT in real-time risk monitoring according to model itself.
Figure 3: Four uses of ChatGPT in real-time risk monitoring according to model itself.

Pattern recognition and anomaly detection:

    Large datasets may reveal trends, patterns, or outliers indicating dangers or abnormalities. 

    ChatGPT can help auditors to spot and flag these abnormalities for further examination. When incorporated with human evaluation correctly, ChatGPT can be useful in identifying potential fraud and enhancing internal audit functions.1

    Auditors can use the model’s natural language processing capabilities to reveal potential risks that might be difficult to identify manually by feeding it relevant data and asking it to look for odd or unexpected patterns. As Figure 4 displays, ChatGPT can generate python codes to detect anomalies. The code trains an anomaly detection model using the Isolation Forest algorithm from the Scikit-Learn library. Anomalies in the dataset are then predicted, and based on the model’s output, it determines which instances are considered as anomalies. 

    Figure 4: Python code generated by ChatGPT to detect anomalies.
    Figure 4: Python code generated by ChatGPT to detect anomalies.

    Further case studies reveal that GPT was tested to depict issues such as reentrancy attacks. 2

    Training auditors

    ChatGPT is widely used in education to help students, teachers, and researchers.

    In audit, it can train auditors by offering them expertise, explanations, and examples that are relevant to their jobs.

    1. Conceptual knowledge: Auditors can ask ChatGPT about auditing concepts, rules, and best practices. It can give inspectors explanations and examples to help them understand complicated topics and the basic ideas behind them.
    2. Case studies: With relevant data provided by the auditor, ChatGPT can model real-life situations and give inspectors case studies to look at.Using the provided information, ChatGPT then creates a hypothetical situation that includes pertinent financial data, documents, and any other elements that could be required. These scenarios can be examined and discussed by auditors to help them develop their auditing methods or techniques.
    3. Review and comments: Auditors can ask ChatGPT for comments on their audit work or strategies they want to try. They can talk about their results, methods, or ideas for answers and get advice or criticism from ChatGPT. Here are 2 ways auditors can benefit ChatGPT in reviewing:
    Figure 5: 2 ways auditors can benefit ChatGPT
    Figure 5: Two ways auditors can benefit ChatGPT

    Challenges to consider on ChatGPT’s use in auditing

    When using ChatGPT’s features in the audit process, human oversight and teamwork are the most important factors. 

    Even though ChatGPT can be helpful and simplify some parts of auditing, it shouldn’t be seen as a replacement for human auditors. Instead, like any AI model, it should be seen as a tool that  improves human judgment and enhances their knowledge. 

    Here’s why human supervision and employee-robot collaboration are important in auditing: 

    Understanding and interpretation of context:

    ChatGPT, as an AI model, cannot completely comprehend (Figure 5) the subtleties, ramifications, and context of the audit tasks and objectives. 

    Human auditors bring their expert judgment, sector expertise, and comprehension of the particular conditions of the business to the table. They can give the essential context and, using their knowledge, interpret the results produced by ChatGPT. Human auditors can evaluate the usefulness, accuracy, and applicability of the data provided by ChatGPT and take appropriate action.

    Figure 6: Comparison of ChatGPT and Professional Human Auditor Source: Certik
    Figure 6: Comparison of ChatGPT and professional human auditor Source: Certik

    Communication:

    A key component of auditing is having effective communication, for instance, with stakeholders. 

    Communication of audit findings, suggestions, and the assurance given to stakeholders is crucial and is best done by human auditors. They can adjust the message to the audience, respond to inquiries, and offer more insights outside of what ChatGPT can.

    Forging a relationship of trust and preserving open lines of communication with stakeholders, human auditors are crucial.

    Accountability:

    ChatGPT can help auditors, but ultimately, the buck stops with the human worker. 

    The auditors are responsible for the: 

    • Choices they make
    • Quality of the audit work 
    • Maintenance of professional standards

    Human oversight ensures that inspectors are held accountable and do their jobs with the right amount of care.

    Establishing a right collaboration between human auditors and ChatGPT in 3 steps

    Related to challenges given above, auditors must master how to interface with and analyze AI technologies like ChatGPT due to their rapid development. AI technology training is a must for auditors because of these reasons:

    Awareness in AI fundamentals:

    In order to interact with AI technologies like ChatGPT, auditors need a basic awareness of AI concepts, principles, and procedures. This includes understanding the many categories of AI models, their constraints, and their uses in the auditing industry. 

    Auditors can more accurately evaluate the capabilities, potential hazards, and acceptable use cases for AI in auditing by having a basic understanding of AI.

    Evaluating AI systems:

    Auditors should be capable of assessing AI systems, especially ChatGPT-like models. 

    This entails evaluating the fairness, precision, and dependability of AI algorithms as well as comprehending their advantages and disadvantages. 

    The design of the model, the training data used, and the procedures for continuing monitoring and validation of the AI system should all be able to be critically assessed by auditors. With the use of this knowledge, auditors may decide wisely whether to use AI in the auditing process.

    Identifying potential biases in AI systems and evaluating data quality:

    Auditors must be skilled at evaluating data quality and seeing potential biases in AI systems. 

    This entails being aware of the implicit and explicit biases present in training data, as well as their possible effects on audit results. 

    The representativeness and integrity of the data used to train AI models should be evaluated by auditors, and they should be able to put policies in place to lessen biases that can affect the objectivity and fairness of audit procedures.

    To discover more about AI and supervision:

    Find the Right Vendors

    1. An Interview with ChatGPT: How Will AI Revolutionise Internal Audit – And Will It Actually Work?” Linkedin. June 6, 2023. Retrieved on June 6, 2023.
    2. Auditing with ChatGPT: Complementary But IncompleteCertik. June 6, 2023. Retrieved on June 6, 2023.

    Burak is an Industry Analyst in AIMultiple. He received his Masters’ degree in Political Science from Middle East Technical University. He has background in researching location-based platforms.


Source link

Related posts
AI

Google AI Releases Gemini 2.0 Flash: A New AI Model that is 2x Faster than Gemini 1.5 Pro

2 Mins read
Google AI Research introduces Gemini 2.0 Flash, the latest iteration of its Gemini AI model. This release focuses on performance improvements, notably…
AI

Microsoft Research Introduces AI-Powered Carbon Budgeting Method: A Real-Time Approach to Tracking Global Carbon Sinks and Emission

3 Mins read
Since the Industrial Revolution, burning fossil fuels and changes in land use, especially deforestation, have driven the rise in atmospheric carbon dioxide…
AI

Evaluating Gender Bias Transfer between Pre-trained and Prompt-Adapted Language Models

1 Mins read
*Equal Contributors Large language models (LLMs) are increasingly being adapted to achieve task-specificity for deployment in real-world decision systems. Several previous works…

 

 

Leave a Reply

Your email address will not be published. Required fields are marked *