Lecture 4: AI in Accounting and Auditing

What is Big Data?

• No unified definition:

  • Data exceeding the level of efficient manageability within traditional DB

  • Process of analyzing a large volume of diverse data, in any variety of form, using ground-breaking apparatus to identify opportunities to improve overall value

• Common trait: large population of data

• Components: volume, variety, velocity, veracity

• New to accounting and audit industry: no formal means to evaluate it, has not applied it in assessments

• Correlations vs. causation

• Examples of Big Data

  • GPS receiver in your cell phone

  • Cash registers when you make a purchase • Cameras in public places

  • Your car

  • Your digital photos

  • Your IoT devices

  • Sensors

Artificial Intelligence

• Artificial intelligence is the use of a computer to model intelligent behavior with minimal human intervention

  • Intelligence exhibited by machines

  • A machine mimics ‘cognitive’ functions that humans associate with other human minds

• Machines & computer programs are capable of problem solving and learning, like a human brain

• Definition: “Intelligence exhibited by machines. A flexible rational agent that perceives its environment and takes actions that maximize its chance of success at some goal. the term ‘artificial intelligence’ is applied when a machine mimics ‘cognitive’ functions that humans associate with other human minds, such as ‘learning’ and ‘problem solving’”

AI Attempts to Mirror Human Capabilities

The Turing Test

Evolution of AI

• Dreyfus (1964) classifies traditional Artificial Intelligence (AI) work into four main areas:

  • Game playing

  • Problem solving

  • Language translation

  • Pattern recognition

• Two decades later, in 1984, that original optimism hit a rough patch, leading to the collapse of a crop of A.I. start-up companies in Silicon Valley, a time known as “the A.I. winter

• Early 80s: focus of AI shifted from basic paradigms and pure logical development to the identification and formalization of human expertise

  • This led to the area of Expert Systems (ES)

• Areas of AI 1990’s

  • Natural Languages

  • Expert Systems

  • Cognition and Learning

  • Computer Vision

Expert Systems (ES)

• Expert Systems became the most popular area of AI and eventually the basis of many commercial, semi-commercial and prototype systems

• Vasarhelyi, M. A. “Expert Systems in Accounting and Auditing,” in Artificial Intelligence in Accounting and Auditing, Vols. 1 to 6 , Markus Wiener Publishing Inc., New York. (1989 to 2002)

AI Enablers

• Faster technology

• Larger yet cheaper storage

• Computerization

• High level of investments by industry (Google, Baidu, Microsoft, etc.)

  • Deepmind developed AlphaGo

  • IBM Watson uses in healthcare

  • Deloitte and Kira systems in contract analysis

Machine Learning

Unsupervised Learning

• No expert input, the computer identifies pattern in data and looks for outliers

• Particularly useful where the human expert does not know what to look for

• Does not require labeled data

Supervised Learning

• Human expert feeds the computer with training data

  • From that data the computer should learn the pattern

• Requires labeled data

Reinforced Learning

• Reinforced learning algorithm continuously learns from the environment in an iterative fashion

• Uses rewards and punishments

How Does AI work?

Examples of What Machine Learning Can Do

ML Applications in Accounting and Auditing

• Expert systems and DSS (duplicate detection)

• Predictive Analytics (profitability, bankruptcy)

• Outlier detection (fraud, CSA, refunds)

• User Segmentation (clustering)

• Inventory analysis (drones, images)

• Contract analysis (engagement letters)

• Sentiment analysis (FS, ESG)

• Automation (RPA and IPA)

• Recommender systems (App recommender, BotVisor)

Risks and Concerns of AI

• Accuracy

• Bias

• Explainability (Blackb)ox

• Security

• Privacy

• Information spillover

• De-skilling

• Job loss

Select Considerations when Establishing AI Development

• Several factors need to be considered when determining if an application meets the required standards or ethics and safety before implementation

(1) Trust and Accuracy

• Accuracy of data inputs and output results

• Over reliance on given information without due diligence on sources

• Explainability and traceability of outputs

(2) Privacy and Security

• Data collection with unclear use; will your sensitive data be made open to the public?

• What surveillance applications of GPT will society deem ethical?

• Cybersecurity concerns

(3) Fairness and Bias

• Bias toward certain subgroups due to public training data

• Bias in model can drive unfair outcomes in some business applications

• Toxicity in responses requires ongoing management

(4) Legal and Regulatory

• Potential copyright and IP infringement considerations

• Liability of use

• Compliance with regulations (existing and in development)

Where Are We Heading?

Innovation Mindset: Bryan’s Amazing Animals [Drones for Inventory Count]

• Auditors perform periodic inventory counts and focus on the innovative practice of using drones and automated counting software to complete the inventory process

• Using drone technology and automation to innovate inventory management and for auditing of inventory

• EY has launched a global proof of concept to expand the use of drones in inventory observations as part of its digital audit capabilities

  • In order to enhance audit quality, this extensive pilot project will use pioneering industry technology to improve the accuracy and frequency of the inventory count data collection

• While capturing images of the inventory can improve the evidence for the inventory counts, this alone does not necessarily improve the efficiency of the inventory process

  • Thus, an important complement to capturing images is to automate the identification and counting of the items in the image

  • To automate the identification of items, several companies have developed automated counting software