CHAPTER TWO

LITERATURE REVIEW

2.1 Conceptual Issues

2.1.1 Concept of Fraud and Fraud Prevention

Fraud has been extensively conceptualized in academic and professional literature. The

Association of Certified Fraud Examiners (ACFE, 2022) defines fraud as "the deliberate use of

misrepresentation or concealment to secure unlawful financial or personal gain." Similarly, under

the Fraud Act 2006 (UK), fraud constitutes "a criminal offense involving deception, false

representation, or abuse of position for pecuniary or proprietary advantage.

From an organizational perspective, Okoye and Akamobi (2009) characterize fraud as "a

premeditated manipulation of truth or factual records by internal or external actors to derive

undue benefits. Generally, fraud is deliberate deception, often involving mispresentation or

concealment of information, with the intent to gain an unfair or unlawfull advantages or benefits

at the expense of another.

Collectively, these definitions highlights the following four (4) critical elements:

Intentionality: fraud is not accidental or unintentional but rather it requires conscious decision

to deceive.

Deception: fraud involves misleading or tricking others.

Unfair advantage: the goal of fraud is to gain something unlawfully or unfairly.

Harm: fraud can either cause financial loss, emotional distress, or damage of reputation.

9
Fraud is typically categorized into three major typologies: financial statement fraud, asset

misappropriation, and corruption.

Financial Statement Fraud: Involves deliberate falsification or manipulation of financial

records to mislead stakeholders. This may include overstating revenues, understating liabilities,

or manipulating earnings.

Asset Misappropriation: This is the most common type of fraud, often involving theft or

misuse of an organization’s assets, such as cash theft, cheque forgery, and unauthorized use of

company resources.

Corruption: Involves abuse of power for personal gain, including bribery, kickbacks, and

conflicts of interest.

Emerging forms such as cyber fraud, which includes identity theft, phishing, malware attacks,

and ATM skimming, have become increasingly prevalent with the digitization of banking

operations (Aliyu & Yusuf, 2023; Musa & Raji, 2022).

Fraud prevention represents an integrated organizational strategy designed to proactively deter,

detect, and mitigate fraudulent activities through multilayered controls. It involves a systematic

approach that combines policies, procedures, and technologies aimed at minimizing the

opportunities for fraud to occur within an organization. According to Obara and Nangih (2017),

fraud prevention goes beyond simply responding to fraudulent acts after they occur; instead, it

emphasizes proactive measures such as ethics training for employees, effective internal control

systems, the use of modern technological tools, and periodic audit procedures. Building on Obara

and Nangih`s perception we will understand the following:

10
 a. Employee ethics training forms a foundational element of fraud prevention, as it

cultivates a culture of integrity and ethical responsibility among staff members. When

employees are well-informed about the ethical expectations and consequences of

fraudulent actions, the likelihood of collusion or misconduct diminishes significantly.

Moreover, robust internal control systems, including checks and balances, segregation of

duties, authorization protocols, and regular reconciliations, serve as both preventive and

detective mechanisms.

b. Technological intervention has also become increasingly vital in modern fraud prevention

strategies. The integration of Artificial Intelligence (AI), data analytics, forensic

accounting tools, and real-time monitoring systems enhances an organization’s ability to

flag anomalies, detect irregular patterns, and respond swiftly to potential threats.

Technology thus enables early warning systems that are critical in deterring fraud.

c. Audit practices, both internal and external, provide an additional layer of assurance by

evaluating the adequacy and effectiveness of existing controls, identifying weaknesses,

and recommending improvements. These audits also reinforce accountability and

transparency across all levels of the organization.

In essence, fraud prevention is not a one-time measure but an ongoing commitment that requires

collaboration across departments, continuous evaluation, and adaptation to emerging fraud risks.

The goal is to create a high-integrity environment where the cost and risk of committing fraud far

outweigh the perceived benefits, thereby reducing its occurrence.

2.1.2 Historical Background of Fraud in Banking.

Fraud in banking is not a recent phenomenon, it has existed since the inception of trade and

financial transactions. However, its methods and scale have evolved alongside advancements in

11
banking systems. In Nigeria, bank fraud traces back to the post-independence era (1960s-1970s),

when manual banking processes were dominant. Fraudulent activities during this period were

largely internal, involving forgery, cheque fraud, and account manipulation by bank employees

or collaborators (Adeniran & Adeyemo, 2017).

The introduction of automated banking systems in the 1980s and 1990s marked a shift in fraud

patterns. While digitalization improved efficiency, it also opened new avenues for fraudsters.

The Nigerian banking sector witnessed high-profile fraud cases, including insider abuse, loan

scams, and money laundering. The lack of stringent regulations and weak enforcement

mechanisms further worsen the problem (Central Bank of Nigeria [CHN], 2014).

With the rapid adoption of electronic banking (e-banking) in the 2000s, fraud migrated to digital

platforms. The CBN's cashless policy (introduced in 2012) accelerated online transactions but

also increased cyber fraud risks. Fraudsters exploited vulnerabilities in online banking, mobile

banking, and card transactions, leading to phishing scams, identity theft, and unauthorized fund

transfers (EFCC, 2020).

2.1.3 Traditional Fraud Prevention and Detection Methods

Traditional methods employed by commercial banks for fraud detection and prevention include:

Internal Controls: These are systems, procedures, and protocols designed to ensure accurate

financial reporting, safeguard assets, and detect irregularities. Examples include segregation of

duties, authorization limits, and reconciliations.

Auditing Procedures: External and internal audits are conducted periodically to identify

misstatements or signs of fraud. Audits play a preventive role by increasing the perception of

detection.

12
Whistleblower Hotlines: Anonymous reporting channels encourage employees or stakeholders

to report unethical behavior without fear of retaliation.

Manual Reconciliation: Routine checking and matching of records, transactions, and balances

to detect discrepancies.

Though valuable, these traditional systems have limitations in keeping pace with sophisticated

and high-volume fraud attempts in today’s digital age (Nwokike et al., 2024).

[Link] Limitations and Challenges of Traditional Methods in the Digital Age

While traditional mechanisms have served as foundational pillars of fraud control, their

effectiveness is waning due to modern-day complexities which include:

Manual Processes: Time-consuming and prone to human error, limiting real-time

responsiveness to fraud.

Reactive Nature: Traditional tools often detect fraud after the damage has occurred rather than

preventing it proactively.

Scalability Issues: These systems struggle to cope with the vast data volumes and transactional

complexities in today’s banking ecosystem.

Lack of Predictive Capability: Traditional methods cannot efficiently predict or anticipate

novel fraud patterns (Ibrahim & Okeke, 2022).

This inefficiency has accelerated the shift towards more adaptive, technology-driven tools like

Artificial Intelligence (AI) integrated into Accounting Information Systems (AIS).

2.1.4 Accounting Information Systems (AIS)

An Accounting information System (AIS) is a structured framework that integrates people,

procedures, software, and data to collect, process, store, and report financial transactions

13
(Romney & Steinbart, 2018), In commercial banking, AIS constitute the technological and

procedural infrastructure that enables them to maintain operational integrity and strategic

decision making.

In a nutshell, Accounting information system (AIS) is a system that businesses uses to collect,

store, manage, process, retrieve and reports its financial data to produce an information that can

be used for decision making. It tracks all accounting and business activity for a company and

which consists of various interdependent component.

In Nigerian's commercial banking sector, AIS performs three critical institutional functions:

a. Financial stewardship: maintains records of transactional data (customer accounts, loan

portfolios, interbank settlement).

b. Regulatory assurance: generates automated audit trails complaint.

c. Decision-making: provides predictive analytics for credit risk modeling and

liquidity management.

[Link] Core Components of Accounting Information System (AIS)

Generally, AIS consist of six (6) primary component. These components are:

1. Human Being (people) who manufacture and use the system and give instructions to AIS.

2. Procedures and Instrument used to collect, process and store data.

3. Data about the organization`s business transaction and all that has to do with the

business.

4. Software that are used to processed the data inputed

5. Infortech Infrastructure, including the computer and other pharipharal devices.

14
 6. Internal cintrol and security measures that guided AIS data and protect the application in

the system.

[Link] Vulnerabilities of Modern AIS to Fraudulent Activities

Despite their advantages, AIS are susceptible to several fraud-related vulnerabilities:

Data Manipulation: Insiders with access may alter or delete critical financial data.

Unauthorized Access: Weak authentication protocols can allow external breaches.

Transactional Anomalies: Fraudsters may exploit system loopholes to conduct irregular

transactions.

These vulnerabilities are compounded in systems lacking up-to-date cybersecurity measures

(Umar & Bello, 2023).

2.1.5 Artificial Intelligence (AI) Fundamentals

[Link] Definition and Key Branches of AI

Artificial Intelligence (AI) is a prominent branch of computer science dedicated to creating

machines and systems that can perform tasks which traditionally require human intelligence.

These tasks include learning, reasoning, problem-solving, perception, and understanding natural

language (Russell & Norvig, 2016). AI enables computers to simulate intelligent behavior and

make decisions based on data, patterns, and programmed algorithms.

In the field of accounting, AI has revolutionized how financial information is processed and

utilized. It plays a vital role in enhancing the efficiency and effectiveness of Accounting

Information Systems (AIS), particularly in the areas of fraud detection, risk management,

auditing, and financial reporting. AI systems can analyze vast amounts of transactional data at

15
high speeds, uncover hidden anomalies, detect suspicious patterns, and automate decision-

making processes with greater accuracy and consistency than traditional manual methods.

Key Branches of Artificial Intelligence

1. Machine Learning (ML):

This branch of AI focuses on developing algorithms that allow computers to learn from data and

improve their performance over time without being explicitly programmed. In accounting, ML

can be used to train models that detect fraudulent activities by learning from historical data and

identifying patterns that deviate from normal behavior.

2. Deep Learning:

A specialized subset of ML, deep learning uses artificial neural networks with multiple layers

(deep architectures) to model complex relationships and recognize intricate patterns in data. It is

especially useful for analyzing unstructured data such as images, audio, and text, and is

increasingly applied in the detection of complex financial frauds that involve layered or hidden

patterns.

3. Natural Language Processing (NLP):

NLP allows machines to understand, interpret, and respond to human language. In the accounting

context, NLP can be employed to process and analyze textual financial documents, emails, audit

logs, and reports to identify potential indicators of fraud, misconduct, or compliance breaches.

Expert Systems:

These are AI systems designed to replicate the decision-making abilities of human experts. They

use a knowledge base of facts and rules to simulate the reasoning process of professionals. In

fraud prevention, expert systems can be used to flag suspicious transactions based on predefined

accounting rules or regulatory compliance standards.

16
The integration of these AI branches into accounting systems is transforming the fraud detection

landscape in commercial banks. AI-powered fraud prevention tools are capable of real-time

monitoring of financial activities, automatic anomaly detection, and predictive modeling to

anticipate potential threats. According to Alrama and Alras (2023), the use of AI technologies in

banking institutions significantly improves the detection of fraudulent behaviors and enhances

internal control mechanisms.

[Link] Core Concepts in Machine Learning

ML is the most widely applied AI subset in fraud prevention and detection and includes:

Supervised Learning: Algorithms are trained on labeled data (e.g., flagged fraudulent

transactions).

Unsupervised Learning: Models detect patterns in data without predefined labels—ideal for

identifying unknown fraud schemes.

Reinforcement Learning: Models learn through trial-and-error, often used in dynamic

environments.

Commercial banks use these methods to train models capable of identifying both known and

emerging fraud patterns (Ahmed et al., 2022).

[Link] Data Requirements and Preprocessing for AI Applications

AI relies heavily on data quality and volume. Effective AI-driven fraud detection requires:

a. Large datasets, including transaction logs, user behavior, and system access logs.

b. Data preprocessing steps like cleaning, normalization, transformation, and feature

selection to prepare inputs for machine learning models.

17
Inconsistent or biased data can reduce model accuracy and increase false positives/negatives,

undermining fraud detection efforts (Yakubu & Musa, 2023).

2.1.6 Ethical Considerations and Biases in AI Development and Deployment

As Artificial Intelligence (AI) becomes increasingly embedded in financial systems, including

Accounting Information Systems (AIS), the ethical implications of its development and

deployment have come under serious scrutiny. While AI presents vast potential for fraud

detection and operational efficiency, it also introduces significant ethical and legal challenges

that must be addressed to ensure fairness, accountability, transparency, and compliance with data

protection standards.

1. Algorithmic Bias and Fairness

One of the critical ethical concerns is algorithmic bias. AI systems, particularly those based on

machine learning, are trained on historical data. If the training data reflects existing societal or

institutional biases, the AI model may learn and replicate those biases. This could lead to unfair

treatment or discriminatory outcomes, especially in fraud detection models where certain

customer profiles may be disproportionately flagged due to biased data inputs. For example, if

past fraud cases were more frequently recorded in particular regions or among specific

demographic groups, the system might unfairly target those groups in future decisions. Hence,

developers must ensure that AI systems are trained on balanced and representative data to

promote algorithmic fairness (Umar, 2023).

18
2. Transparency and Explainability

Another pressing ethical issue is the lack of transparency in AI decision-making, particularly in

deep learning models which function as “black boxes.” These models can produce highly

accurate results but often do so without providing understandable explanations of how the

decisions were reached. This lack of explainability undermines accountability and trust in AI-

based fraud detection systems, especially when financial decisions or accusations of fraud are

made without clear reasoning. Regulatory frameworks and ethical AI standards now emphasize

the need for interpretable AI systems, where users and auditors can trace and understand the

logic behind key decisions.

3. Data Privacy and Protection

AI systems require vast amounts of data for training and decision-making. In banking

environments, this includes sensitive customer information, financial histories, and transaction

details. The collection, storage, and processing of such data must be done in strict compliance

with data protection laws such as the Nigeria Data Protection Regulation (NDPR) and the

General Data Protection Regulation (GDPR) in the European Union. Any misuse, breach, or

unauthorized access to such data could not only violate customer privacy but also expose the

financial institution to legal liabilities and reputational damage.

4. Ethical AI Governance and Regulation

To ensure responsible AI deployment, banks and financial institutions must adopt ethical AI

governance frameworks. This includes setting up ethical review boards, ensuring human

oversight in AI operations, conducting regular bias audits, and maintaining transparency in AI

system design and usage. Ethical AI practices are necessary not only to foster public trust, but

19
also to ensure alignment with regulatory expectations and the broader goals of social

responsibility and justice in financial systems.

2.1.7 AI Techniques for Fraud Prevention and Detection

Artificial Intelligence (AI) has revolutionized fraud detection by offering powerful tools and

techniques capable of analyzing vast amounts of data with precision and speed. Commercial

banks increasingly integrate AI into their Accounting Information Systems (AIS) to proactively

detect and prevent fraudulent activities. The following are key AI techniques that enhance fraud

prevention and detection in financial environments:

[Link] Supervised Learning Models for Classification

Supervised learning models are among the most widely used techniques in fraud detection. These

models learn from historical, labeled datasets where each transaction is marked as either

fraudulent or legitimate. Through this learning process, the model builds a classification

framework that can predict the likelihood of new, unseen transactions being fraudulent.

Key supervised learning techniques include:

a. Decision Trees: These models break down data into simple decision rules, making them

intuitive and easy to interpret. They are particularly useful in rule-based fraud detection

systems where clarity is essential.

b. Random Forests: An ensemble of multiple decision trees, Random Forests help reduce

overfitting and improve classification accuracy. They provide more robust predictions

than a single decision tree.

20
 c. Support Vector Machines (SVM): SVMs are particularly effective in binary classification

tasks. They work well in high-dimensional spaces and can detect complex fraud patterns

that linear classifiers might miss.

d. Neural Networks: These models mimic the human brain's neural architecture and are

capable of detecting highly non-linear and intricate patterns in data. They are particularly

effective when dealing with large, complex datasets in banking transactions.

Commercial banks employ these models to classify transactions as either fraudulent or non-

fraudulent, improving the accuracy and speed of fraud detection (Ajayi, 2023).

[Link] Unsupervised Learning Models for Anomaly Detection

In many real-world situations, labeled data may not be readily available, especially when

detecting new or evolving fraud types. Unsupervised learning models address this challenge by

identifying patterns and deviations from normal behavior without requiring predefined labels.

Popular unsupervised techniques include:

a. K-Means Clustering: This method groups similar transactions into clusters. Transactions

that do not fit well into any cluster are flagged as anomalies, which could be indicative of

fraud.

b. Isolation Forests: Rather than profiling normal transactions, this model isolates

anomalous ones by creating random partitions. Transactions that are isolated quickly are

likely to be fraudulent.

c. Autoencoders: These are neural networks trained to reconstruct input data. When a

transaction yields a high reconstruction error, it suggests that the data is anomalous and

potentially fraudulent.

21
These models are crucial for identifying fraud patterns that are previously unseen, thus

strengthening fraud detection efforts in dynamic financial environments (Abubakar, 2023).

[Link] Natural Language Processing (NLP) for Analyzing Unstructured Data

Fraud signals are not always numeric; they can also be hidden in textual data. Natural Language

Processing (NLP) enables AI systems to analyze and interpret unstructured text for signs of

fraudulent behavior.

Applications of NLP in fraud detection include the analysis of:

a. Customer Complaints: NLP helps identify patterns or repeated complaints that may

indicate unethical banking practices or internal fraud.

b. Whistleblower Reports: Text mining techniques help detect key risk indicators or

recurring issues raised by whistleblowers.

c. Internal Emails: NLP algorithms can flag suspicious communication patterns that may

suggest collusion or fraud schemes.

By supplementing traditional data models, NLP provides a broader view of fraud risk by

including unstructured text analysis (Musa, 2022).

2.1.8 Integration of Artificial Intelligence within Accounting Information Systems (AIS) for

Fraud Prevention.

The integration of Artificial Intelligence (AI) into Accounting Information Systems (AIS) marks

a pivotal evolution in the field of digital accounting, particularly in enhancing fraud prevention

mechanisms in commercial banking. As financial institutions face increasingly sophisticated

fraudulent schemes, embedding AI into AIS allows for more agile, intelligent, and data-driven

22
responses to fraud threats. The fusion of AI capabilities with the transactional, reporting, and

audit functionalities of AIS facilitates both preventive and detective fraud controls, ensuring that

modern banking systems are more resilient and responsive to anomalies (Musa et al., 2023).

[Link] Architectural Considerations for Embedding AI Modules in AIS

To effectively implement AI in AIS, the system’s architecture must accommodate intelligent

processing capabilities without disrupting core accounting operations. There are three main

architectural configurations that institutions may consider:

1. Cloud-based AI Platforms: These platforms offer scalability, flexibility, and reduced

costs. Cloud infrastructure allows banks to leverage advanced AI models and big data

processing with minimal internal hardware requirements. Additionally, cloud solutions

enable frequent updates and maintenance of AI tools to stay ahead of emerging fraud

techniques.

2. On-Premise AI Solutions: Some financial institutions, particularly those with

heightened data privacy concerns or regulatory restrictions, opt for on-premise solutions.

This setup allows for greater control over data, better compliance with internal policies,

and reduced dependence on third-party vendors.

3. API and Middleware Integration: Modern AIS often need to interface with legacy

systems. Application Programming Interfaces (APIs) and middleware serve as connectors

between AI modules and existing accounting infrastructures, facilitating seamless data

flow and system interoperability.

23
Most commercial banks adopt a hybrid architecture that blends the flexibility of cloud services

with the control of on-premise deployments. This dual approach ensures operational security

while enabling advanced AI functionalities (Musa et al., 2023).

[Link] Data Integration and Interoperability Challenges

A critical requirement for successful AI deployment in AIS is the effective integration of data

across various banking subsystems. However, several challenges hinder smooth implementation:

a. Lack of Standardized Protocols: Different departments within banks often use disparate

systems that do not communicate using a common protocol, making integration complex

and time-consuming.

b. Inconsistent Data Formats: Variations in data storage formats (structured vs.

unstructured) can result in loss of fidelity or misinterpretation when feeding data into AI

systems.

c. Data Silos Across Departments: When operational units function independently without

centralized data management, important transactional and behavioral data become

inaccessible to AI systems.

Overcoming these challenges requires robust ETL (Extract, Transform, Load) processes that

prepare data for AI consumption. Additionally, implementing enterprise-wide data governance

frameworks ensures uniform standards for data collection, access, and security. These structures

are vital for the AI-enhanced AIS to deliver reliable fraud detection and prevention capabilities

(Ahmed, 2023).

Integrating AI into AIS represents a strategic advancement in modern banking fraud control.

However, success relies heavily on architectural foresight and data infrastructure readiness. By

24
addressing the technical, organizational, and security aspects of integration, banks in Maiduguri

and beyond can harness the full potential of AI to safeguard financial integrity and maintain

stakeholder trust.

Fraudulent activities often involve networks of colluding individuals or entities. Graph analytics

allows AI systems to analyze relationships and connections between entities such as accounts,

devices, and [Link] technique can uncover:

a. Collusion Among Employees and Clients

b. Money Laundering Rings

c. Insider Fraud Schemes

Graph-based AI models visualize relationships in the form of nodes and edges, making it easier

to detect unusual network structures and transactional loops. Traditional detection models may

miss such hidden patterns, but graph analytics can expose these connections through pattern

recognition and link analysis (Yakubu & Okon, 2022).

2.2 Review of Existing Literature

This section provides a critical review of scholarly works and existing studies directly aligned

with the specific objectives of this research. The review is structured according to each objective

to trace what has been established in prior studies and to establish a foundation for the

contribution of this study.

2.2.1 Extent of AI Adoption in Accounting Information Systems (AIS) for Fraud Detection

Several studies have examined the adoption of artificial intelligence (AI) technologies within

accounting information systems (AIS) and their role in fraud detection. According to Al-Hiyari

et al. (2022), commercial banks have gradually begun integrating AI tools such as machine

25
learning and predictive analytics into their AIS to enhance fraud prevention. Similarly, Khalid et

al. (2023) noted that in the Tabuk region, banks reported growing confidence in using AI for

early anomaly detection, but highlighted that adoption was still at a moderate stage due to

infrastructural challenges.

The adoption rate is influenced by internal organizational readiness and external regulatory

encouragement. Aladejebi (2022) emphasized that Nigerian banks face unique constraints like

inadequate power supply, low digital literacy among employees, and cybersecurity risks that

hinder AI implementation in AIS. Furthermore, Asiri (2023) highlighted that many institutions

view AI as expensive and complex, which limits its full adoption in emerging economies.

2.2.2 Challenges Associated with Implementing AI-Driven Fraud Prevention Mechanisms

Implementing AI in fraud prevention is fraught with numerous obstacles. Key among these is the

availability of clean and high-quality data for training machine learning models (Owolabi &

Makinde, 2022). Data inconsistencies and legacy systems in Nigerian banks make it difficult to

integrate AI modules into AIS.

Moreover, regulatory compliance poses a serious challenge. As noted by Raji and Thompson

(2023), most financial institutions lack a clear legal framework to support AI usage, leading to

concerns over accountability, data privacy, and the ethical use of AI tools. Organizational

resistance to change is another factor that significantly slows AI integration, particularly among

older employees who lack digital skills.

In a study conducted by Bello and Musa (2021), over 60% of surveyed commercial banks cited

inadequate funding, cybersecurity threats, and limited expertise as core challenges inhibiting AI-

driven fraud prevention in Borno State.

26
2.2.3 Effectiveness of AI in Detecting and Mitigating Fraudulent Activities

AI has shown considerable potential in improving fraud detection. Supervised and unsupervised

learning algorithms have outperformed traditional rule-based systems in identifying fraudulent

transactions. As reported by Yakubu (2023), AI systems such as neural networks and support

vector machines can accurately classify suspicious behaviors and reduce false positives.

Similarly, studies like that of Abdulraheem et al. (2022) emphasized that AI-enabled continuous

auditing tools help auditors detect unusual activities in real-time, which leads to quicker

mitigation. Furthermore, Banks using AI-powered anomaly detection recorded a significant

decline in fraud losses between 2020 and 2022, particularly in digital channels.

However, the effectiveness of AI is dependent on system training, volume of quality data, and

periodic algorithm updates. Without this, AI models become stale and ineffective, as pointed out

by Jibril and Haruna (2022).

2.2.4 Regulatory and Operational Factors Influencing AI Integration in Fraud Prevention

There is a growing awareness of the need for regulatory alignment in the deployment of AI for

fraud prevention. CBN’s 2023 Fraud Risk Management Guidelines encouraged banks to adopt

more advanced technologies while emphasizing customer data protection. Nevertheless,

enforcement remains weak.

From an operational standpoint, integration complexity between legacy AIS platforms and AI

software remains a major bottleneck. As found by Omeiza and Salisu (2023), AIS platforms in

Nigerian banks were not originally built to accommodate intelligent systems, making integration

both costly and time-consuming.

27
Leadership commitment and organizational culture are also critical. Institutions with a high level

of digital maturity and leadership support are more likely to succeed in AI adoption, as

demonstrated by comparative studies between banks in Lagos and Borno State.

2.3 Empirical Framework

The empirical framework serves as the foundation upon which the theoretical assertions of this

study are tested and validated using real-world data and observations from past research. This

section highlights previous empirical findings relating to the role of Artificial Intelligence (AI) in

enhancing Accounting Information Systems (AIS) for fraud prevention in commercial banks.

The framework presents studies aligned with the specific objectives of this research, offering

critical insight into prevailing methodologies, patterns, and conclusions.

2.3.1 Empirical Studies on AI Adoption in AIS for Fraud Detection

Several studies have explored the extent of AI adoption in accounting systems for fraud

detection. Alsharari (2021), in a study conducted in Saudi Arabia’s Tabuk region, observed that

while commercial banks were gradually integrating AI tools such as machine learning and expert

systems into their AIS, adoption remained at the early stages due to infrastructural and regulatory

limitations. Similarly, research by Nwaiwu and Eze (2023) in Nigeria revealed that less than

40% of commercial banks had fully adopted AI-enabled fraud detection systems, despite

acknowledging the potential of such technologies to significantly reduce fraud incidences.

In a study involving 150 respondents across five Nigerian commercial banks, Ogunleye and

Abiola (2022) found that most banks relied more on rule-based detection and less on machine

learning or natural language processing tools. This partial adoption was attributed to high costs

28
and lack of skilled personnel. These findings show that while the theoretical benefits of AI in

AIS are acknowledged, practical implementation lags behind due to contextual limitations.

2.3.2 Empirical Studies on Challenges of AI-Based Fraud Prevention Mechanisms

The implementation of AI systems in AIS is often met with operational and structural challenges.

According to Ugwoke et al. (2022), key challenges include data quality issues, lack of advanced

computational infrastructure, regulatory uncertainty, and staff resistance. Their survey of bank IT

managers in Lagos and Abuja identified model explainability and fear of job displacement as

major barriers to AI acceptance.

Another study by Okafor and Ijeoma (2021) analyzed the challenges of AI deployment across

public and private banks in Nigeria. The study emphasized that inadequate data privacy

frameworks, insufficient training, and high upfront costs were recurring impediments. These

results suggest that beyond technology, sociocultural and economic factors play a critical role in

AI implementation.

2.3.3 Empirical Evidence on the Effectiveness of AI in Fraud Detection and Mitigation

Several empirical studies have established the effectiveness of AI in detecting and mitigating

fraud in banking operations. The work of Ahmed and Al-Yahyaei (2021) demonstrated that AI

systems utilizing supervised learning algorithms like decision trees and random forests achieved

fraud detection accuracy of up to 92% when trained on historical transactional data. Similarly, a

cross-sectional study by Bala and Gimba (2023) in Borno State showed that banks employing

AI-based anomaly detection systems reported a 35% decline in fraud cases over three years.

29
Moreover, comparative results from the Central Bank of Nigeria’s 2023 fraud report illustrated

that banks using real-time AI alert systems (such as GTBank and Zenith) had lower fraud loss-

to-income ratios compared to those still reliant on manual internal control systems. These

findings empirically support the proposition that AI significantly enhances the fraud detection

capabilities of AIS.

2.3.4 Empirical Studies on Regulatory and Operational Factors Influencing AI Integration

Research by Egbunike and Umeoduagu (2020) identified several regulatory barriers affecting AI

deployment in financial systems. Their study emphasized the lack of a specific regulatory

framework for AI-based accounting applications in Nigeria. Furthermore, operational challenges

such as outdated legacy systems and low digital literacy among banking staff hinder seamless AI

integration.

In another empirical survey conducted by Adebayo et al. (2022), respondents from the

compliance departments of Nigerian banks emphasized the role of the Central Bank’s Know-

Your-Customer (KYC) and Anti-Money Laundering (AML) policies in shaping AI deployment

strategies. The study highlighted that AI adoption is more successful when aligned with

regulatory compliance tools, improving both effectiveness and institutional trust.

Conclusion of the Empirical Framework

The empirical studies reviewed reveal a clear pattern: AI has significant potential to transform

fraud prevention in commercial banks through its integration with AIS. However, challenges

such as low adoption rates, infrastructural deficits, regulatory ambiguity, and user resistance

continue to slow progress. The studies also offer practical strategies—such as phased integration,

30
capacity building, and alignment with compliance regulations—that future research and practice

can adopt. This body of empirical work provides a strong benchmark for the present study, upon

which new insights specific to Maiduguri, Borno State, will be drawn.

2.4 Theoretical Framework

A theoretical framework is the foundation upon which a research study is built. It serves to

explain the key concepts and the relationships between them using established theories. This

study focuses on how Artificial Intelligence (AI) can enhance Accounting Information Systems

(AIS) for fraud prevention in commercial banks in Maiduguri, Borno State. Therefore, this

section will explain and connect relevant theories that help understand technology adoption,

fraud behavior, and system effectiveness.

2.4.1 Technology Acceptance Model (TAM)

Origin and Background

The Technology Acceptance Model (TAM) was proposed by Fred Davis in 1986, building upon

the Theory of Reasoned Action (TRA) developed by Fishbein and Ajzen. TAM is a widely

accepted theoretical framework used to explain and predict the acceptance and use of technology

by individuals within an organization or system.

Davis (1989) introduced TAM specifically to understand computer usage behavior by examining

the psychological variables that influence users’ decisions about adopting new technologies. The

model identifies two key beliefs that drive technology acceptance:

a. Perceived Usefulness (PU): This refers to the degree to which a person believes that

using a particular system or technology would enhance their job performance. If users see

31
 the system as beneficial in improving their efficiency, accuracy, or decision-making, they

are more likely to adopt it.

b. Perceived Ease of Use (PEOU): This represents the extent to which a person believes

that using the system will be free from effort. In other words, if the system is user-

friendly, intuitive, and does not require excessive technical skills or training, users are

more inclined to use it.

TAM posits that both PU and PEOU directly influence the attitude towards using a system,

which in turn affects the behavioral intention to use, ultimately leading to actual system usage.

Application to This Study

The Technology Acceptance Model is particularly relevant to this study on the role of Artificial

Intelligence (AI) in enhancing Accounting Information Systems (AIS) for fraud prevention in

commercial banks in Maiduguri, Borno State.

In this context, TAM provides insights into the attitudes and behavioral responses of bank

employees (such as accountants, auditors, IT personnel, and managers) towards the adoption of

AI-powered fraud detection systems. For such AI technologies to be successfully integrated into

existing AIS frameworks, users must perceive them as both useful and easy to use.

If bank staff believe that AI tools can significantly improve the detection and prevention of

fraudulent activities, reduce financial losses, and enhance the overall integrity of financial

reporting, their Perceived Usefulness will be high.

Additionally, if these AI applications are designed to be user-friendly, require minimal training,

and integrate seamlessly with current AIS interfaces, their Perceived Ease of Use will also be

high.

32
By understanding and applying the TAM model, developers and bank management can better

design, deploy, and manage AI-driven systems that are aligned with user expectations and

capabilities. This understanding is crucial to overcoming resistance to technological change and

ensuring a smoother transition to AI-enhanced fraud prevention tools.

Moreover, TAM helps this study establish a theoretical link between human factors (attitudes,

beliefs, and behavior) and technological innovation (AI in AIS). It offers a practical basis for

examining how perceptions influence technology acceptance, which is key to the successful

implementation of AI systems in the banking sector.

Conclusion

In summary, TAM explains the cognitive and psychological drivers behind the acceptance of AI-

based technologies in fraud prevention. It aligns with the study's objectives by providing a

framework for understanding how end-user attitudes influence the success of integrating AI into

AIS in commercial banks. Recognizing and addressing these acceptance factors can enhance

system adoption, improve fraud detection capabilities, and contribute to more secure and

transparent financial operations.

2.4.2 Fraud Triangle Theory (FTT)

Origin and Background

The Fraud Triangle Theory was developed by Donald R. Cressey in 1953, a renowned

criminologist who studied the behavioral patterns of incarcerated white-collar criminals.

Through his research, Cressey discovered that individuals who committed occupational fraud

typically did so under the convergence of three critical conditions. These three components form

what is now widely known as the fraud triangle:

33
1. Pressure (Incentive):

This refers to the motivation or external/internal force that pushes an individual toward

committing fraud. Common pressures include financial difficulties, unrealistic performance

targets, medical expenses, addiction, or even social status maintenance. The individual may feel

a sense of urgency to resolve their problems through illegitimate means.

2. Opportunity:

This element represents the situation or environment that allows fraud to be perpetrated. It often

arises from weaknesses or loopholes in internal controls, lack of supervision, poor segregation of

duties, or ineffective audit mechanisms. Without a clear opportunity, even motivated individuals

may not act on their intentions.

3. Rationalization:

This is the cognitive process where the fraudster justifies their dishonest actions to themselves.

They may convince themselves that they are only "borrowing" the money, or that their employer

"owes" them, or that no one will be harmed. Rationalization allows individuals to live with their

unethical actions without feeling criminal or guilty.

Together, these three elements form a comprehensive psychological and situational model that

explains the “why” behind fraudulent behavior in organizations.

Application to This Study

The Fraud Triangle Theory is directly applicable to this research, which explores the use of

Artificial Intelligence (AI) in strengthening Accounting Information Systems (AIS) to prevent

fraud in commercial banks in Maiduguri, Borno State.

34
In this context:

Pressure: Employees in banking institutions may experience financial or personal pressures that

could tempt them to commit fraud. While this element is often internal and hidden, certain

behavioral patterns (such as sudden lifestyle changes or high debt levels) can be flagged for risk

assessment.

Opportunity: This is the most actionable component for system-based prevention. Weak or

outdated AIS frameworks may create exploitable gaps, such as unrestricted access to financial

records, inadequate approval hierarchies, or ineffective audit trails. By identifying these

opportunity zones, banks can deploy AI algorithms to tighten controls and automatically monitor

for suspicious behaviors.

Rationalization: Although difficult to measure directly, rationalization can be detected

indirectly through behavior analysis. For instance, if an employee consistently overrides system

checks or works outside of normal hours without justification, AI can flag these as potential

indicators of internal threats.

AI-enhanced AIS can be programmed to monitor all three sides of the triangle. For example:

Machine learning models can detect abnormal transaction patterns that indicate pressure-induced

behavior.

Rule-based systems can block or alert on unauthorized access attempts, reducing opportunity.

Natural Language Processing (NLP) can even analyze written communication (emails, reports,

etc.) to detect language indicative of rationalization.

By applying the fraud triangle theory, this study underscores how AI tools should not just detect

fraud after it occurs but proactively analyze the behavioral and system-related precursors of

35
fraud. This significantly strengthens internal controls and serves as a deterrent against unethical

financial practices.

Conclusion

The Fraud Triangle Theory provides a behavioral and systemic foundation for understanding and

combating fraud. When applied to AIS, it helps institutions identify critical risk zones and align

AI-driven fraud prevention systems accordingly. Integrating this theory into system design

promotes ethical banking environments, encourages transparency, and reduces the likelihood of

fraud in commercial banks. It also supports the ethical imperative of safeguarding public trust

and institutional accountability in the Nigerian financial sector.

2.4.3 Unified Theory of Acceptance and Use of Technology (UTAUT)

Origin and Background

The Unified Theory of Acceptance and Use of Technology (UTAUT) was proposed by

Venkatesh et al. (2003) as a comprehensive model that integrates and extends eight prior models

of technology acceptance, including the Technology Acceptance Model (TAM), the Theory of

Reasoned Action (TRA), the Theory of Planned Behavior (TPB), and the Motivational Model,

among others.

UTAUT was developed to provide a unified framework that better explains user intentions and

subsequent usage behavior with regard to new technologies in organizational settings. The model

identifies four core determinants of intention and usage:

1. Performance Expectancy (PE): This is the degree to which an individual believes that using

the system will help them achieve gains in job performance. It is similar to Perceived Usefulness

in TAM and is often the strongest predictor of intention to use a system.

36
2. Effort Expectancy (EE): This refers to the degree of ease associated with the use of the

system. If a technology, such as an AI-enabled fraud detection tool, is perceived as simple and

user-friendly, employees are more likely to adopt it.

3. Social Influence (SI): This involves the extent to which an individual perceives that important

others (e.g., supervisors, colleagues, or industry leaders) believe they should use the new system.

Organizational culture and peer pressure can significantly affect technology adoption in a formal

setting like banking.

4. Facilitating Conditions (FC): These are the organizational and technical infrastructures that

support the use of the system. This includes training programs, access to skilled IT support,

equipment, reliable internet, and power supply. Facilitating conditions are particularly important

in developing regions.

Additionally, UTAUT includes four moderating variables — gender, age, experience, and

voluntariness of use — which influence the strength of the above relationships.

Application to This Study

UTAUT is particularly relevant in the context of this study, which investigates how Artificial

Intelligence (AI) can enhance Accounting Information Systems (AIS) to prevent fraud in

commercial banks in Maiduguri, Borno State. In this setting, environmental, social, and

organizational factors play a significant role in whether AI technologies are accepted and used

effectively by banking staff.

Performance Expectancy: For employees in Maiduguri’s commercial banks to accept AI tools

integrated into AIS, they must believe that such technologies will significantly improve their

ability to detect and prevent fraud, leading to better decision-making, increased transparency,

and operational efficiency.

37
Effort Expectancy: Since many users may not have advanced technical backgrounds, AI

systems must be intuitive and easy to use. If bank staff perceive the tools as too complex or

requiring extensive training, they may resist adoption.

Social Influence: In Maiduguri’s hierarchical work environments, the attitudes of managers, IT

heads, and top-level executives greatly influence adoption decisions. If these leaders actively

promote AI adoption, others are more likely to follow suit. Similarly, peer influence among

colleagues can affect perceptions of usefulness and ease of use.

Facilitating Conditions: This factor is critically important in Maiduguri, where infrastructure

challenges (e.g., inconsistent power supply, limited internet connectivity, or lack of training)

may impede the successful deployment of AI-driven AIS. The model highlights the need for

adequate institutional support, training programs, and enabling environments for AI tools to

function effectively and be accepted by users.

Through the UTAUT lens, this study emphasizes that AI adoption is not only a technological

issue but also a social and organizational process. Addressing the barriers identified by UTAUT

(such as lack of infrastructure or leadership support) can significantly enhance the acceptance

and sustained use of AI technologies in commercial banking.

Conclusion

The UTAUT model provides a comprehensive understanding of the factors influencing

technology adoption in organizational settings. Its application to this study helps identify and

mitigate resistance to AI in AIS by recognizing the roles of performance expectations, ease of

use, social environment, and institutional readiness. In doing so, the study offers practical

recommendations for fostering a supportive ecosystem that facilitates the successful integration

of AI-driven fraud prevention tools in commercial banks within Maiduguri, Borno State.

38
2.4.3 Systems Theory

Origin and Background

Systems Theory was developed by Ludwig von Bertalanffy in 1945, initially as a conceptual

framework in biology. Over time, it evolved into a multidisciplinary approach applied across the

social sciences, management, and organizational studies. At its core, Systems Theory views any

organization or entity as a set of interconnected and interdependent components working

together to achieve a common goal. Changes in one part of the system inevitably influence other

parts, whether directly or indirectly.

The theory emphasizes the importance of wholeness, interaction, and interrelationship. Rather

than analyzing each unit or department in isolation, Systems Theory encourages a holistic view

— considering how subsystems (like accounting, IT, human resources, auditing, etc.) interact

and function as part of a larger, dynamic structure.

Key principles include:

1. Interdependence of Components: Every part of the system is connected. A change in one

unit affects the whole.

2. Open Systems Thinking: Organizations interact with and are influenced by their external

environment.

3. Feedback Loops: Systems rely on feedback to maintain equilibrium, adjust to changes,

and correct errors.

4. Goal Orientation: Each subsystem contributes toward achieving the overall objectives of

the organization.

39
Application to This Study

Systems Theory is particularly relevant to this study, which focuses on how Artificial

Intelligence (AI) can enhance Accounting Information Systems (AIS) for fraud prevention in

commercial banks within Maiduguri, Borno State.

In this context, AIS is not a standalone component. It is a subsystem embedded within a broader

organizational framework. As such, any integration of AI into the AIS is not merely a

technological upgrade — it represents a system-wide transformation that affects multiple

departments and processes across the bank.

Some specific applications include:

Data Collection and Flow: AI-powered AIS affects how financial data is collected, verified,

and stored. It may alter workflows in operations, customer service, and internal audit units.

Collaboration Among Units: Fraud prevention does not depend on AIS alone. It requires

effective cooperation among the accounting unit, internal audit, compliance officers, IT

departments, and even human resources. For instance, if the AI system flags suspicious behavior,

the compliance team must investigate, while IT must ensure system reliability and data security.

Organizational Impact: Introducing AI into AIS may require training programs, policy

adjustments, and new standard operating procedures, all of which influence other parts of the

system. For example, auditors may need to understand how AI algorithms detect anomalies,

while management may rely on AI-generated reports for decision-making.

Feedback Loops and System Learning: AI systems can benefit from continuous feedback,

learning from previous fraud cases to detect new patterns. This aligns with Systems Theory’s

emphasis on feedback mechanisms and adaptive learning, reinforcing the idea that effective

fraud prevention is an evolving, system-wide responsibility.

40
Systems Theory also cautions against departmental silos. In commercial banks, treating AIS and

AI implementation as an isolated IT or finance project can lead to suboptimal results. Instead,

adopting a systemic approach ensures that changes in technology are integrated thoughtfully

across all units, maximizing the effectiveness of fraud prevention efforts.

Conclusion

Systems Theory offers a holistic framework for understanding how technological interventions

— such as the integration of AI into AIS — impact not just financial processes, but the entire

organizational ecosystem. It highlights the interdependence of departments and underscores the

need for collaboration, feedback, and alignment of goals. In this study, Systems Theory

reinforces the importance of viewing fraud prevention as not just a technological solution, but a

multi-departmental, systemic strategy in Nigeria’s commercial banking sector.

2.4.4 Contingency Theory

Origin and Background

Contingency Theory was formulated by Paul R. Lawrence and Jay W. Lorsch in 1967, as a

response to the limitations of classical management theories that promoted universal best

practices. Unlike those theories that propose rigid, one-size-fits-all solutions, Contingency

Theory emphasizes that “there is no single best way to manage or structure an organization.” The

most effective strategies, structures, or technologies are contingent upon the specific internal and

external conditions an organization faces.

Key assumptions of the theory include:

1. Organizations operate in dynamic environments with varying levels of uncertainty and

complexity.

41
 2. Optimal effectiveness can only be achieved by adapting systems, structures, and

processes to fit the specific context in which they operate.

3. Variables such as organizational size, technological advancement, staff competency,

environmental stability, and regulatory constraints all influence what approach is most

suitable.

Contingency Theory is, therefore, a situational approach to problem-solving. It supports

flexibility, customization, and environmental responsiveness in managerial decision-making and

technological adoption.

Application to This Study

Contingency Theory is highly relevant to this study’s aim of evaluating the role of Artificial

Intelligence (AI) in enhancing Accounting Information Systems (AIS) for fraud prevention in

commercial banks in Maiduguri, Borno State.

In this setting, commercial banks differ significantly in several contextual factors:

Size and Operational Scale: Some banks (e.g., Zenith Bank or First Bank) operate on a large

scale with complex financial operations, multiple departments, and high transaction volumes.

Others (such as microfinance or regional banks) have smaller operations with limited staff and

simpler accounting systems.

Digital Maturity and Infrastructure: While a few banks in Maiduguri may have advanced IT

infrastructure and digitized processes, others may still rely on manual or semi-automated

systems. The level of technological readiness varies.

Human Resource Capacity: Staff qualifications, digital literacy, and experience with emerging

technologies also differ. A bank with experienced internal auditors and IT specialists may

42
comfortably implement sophisticated AI models, while others may struggle with basic system

automation.

Security Needs and Fraud Risk Levels: The nature and frequency of fraud threats can vary by

institution based on customer base, internal control systems, and exposure to digital platforms.

Contingency Theory therefore supports the customization of AI-driven fraud prevention

strategies based on these unique organizational variables.

For example: A larger bank with a mature digital ecosystem might implement machine learning

algorithms and predictive analytics to detect complex fraud patterns in real time. A smaller bank

with minimal digital capacity might instead use basic AI-powered alerts for unusual transactions,

user behavior analysis, or access control monitoring.

This approach avoids imposing a universal fraud detection model on all banks and instead

advocates for strategic alignment between AI tools and each bank’s contextual realities.

By applying Contingency Theory, this study encourages stakeholders to consider:

Organizational structure and digital readiness

Resource availability (human and technical)

Nature of fraud risks

Cultural and managerial attitudes toward innovation

Such considerations are crucial in ensuring the successful adoption, functionality, and

sustainability of AI-integrated AIS in different banking environments.

Conclusion

Contingency Theory reinforces the importance of contextualizing fraud prevention strategies. It

rejects the idea of uniform solutions and emphasizes that each commercial bank requires a

tailored approach to AI adoption, depending on its internal structure, resources, and

43
environmental conditions. By adopting a contingency mindset, banks in Maiduguri can optimize

the design and implementation of AI-enabled AIS, improving fraud detection outcomes while

ensuring organizational fit and efficiency.

2.4.5 Information Systems Success Model (ISSM)

Origin and Background

The Information Systems Success Model (ISSM) was developed by DeLone and McLean in

1992, with a revised version in 2003. It remains one of the most influential frameworks in

Information Systems (IS) research, offering a multidimensional approach to assessing the

success and effectiveness of information systems within organizations.

DeLone and McLean proposed that system success is not determined by a single factor, but

rather by a set of interrelated components that reflect both technical and organizational

outcomes. These six core dimensions include:

1. System Quality: Refers to the technical performance of the system — including reliability,

speed, user interface design, functionality, and security. A high-quality system should be stable

and capable of performing complex tasks consistently.

2. Information Quality: Measures the accuracy, relevance, completeness, and timeliness of the

information the system produces. For accounting systems, it’s essential that reports and data

outputs reflect real-time and correct financial activity.

3. Service Quality: Encompasses the support, maintenance, training, and responsiveness

provided to users by the IT or technical support teams. A system with excellent support services

enhances user confidence and reduces resistance.

44
4. Use / Intention to Use: This dimension assesses how often and how effectively the system is

used. It can also include user willingness and motivation to use the system regularly and in

accordance with organizational goals.

5. User Satisfaction: Indicates the users’ attitudes, comfort, and trust in the system. If users feel

the system is helpful, accessible, and valuable to their job, satisfaction levels will be higher,

encouraging continued use.

6. Net Benefits: Captures the organizational gains or outcomes resulting from system use, such

as efficiency improvements, cost reductions, decision-making enhancements, and fraud

reduction.

These components are interconnected, meaning that improved system and information quality

will likely lead to greater use, higher user satisfaction, and ultimately, tangible organizational

benefits.

Application to This Study

The ISSM framework is crucial to this study, which investigates how Artificial Intelligence (AI)

can enhance Accounting Information Systems (AIS) for fraud prevention in commercial banks in

Maiduguri, Borno State. After the implementation of AI-driven systems, evaluating their success

must go beyond just deployment; it should assess how well the systems perform, serve users, and

produce results.

Here’s how each component applies:

System Quality: The AI-enhanced AIS must be technically sound — capable of processing

large volumes of transactions, identifying anomalies, and flagging potential fraud in real time

without crashing or lagging.

45
Information Quality: AI systems should generate precise, relevant, and timely data. Fraud

alerts and transaction analyses must be reliable to ensure that decisions are based on accurate

information.

Service Quality: Successful adoption depends heavily on training, user support, and

responsiveness. In the Maiduguri banking context, where digital familiarity may vary, staff must

receive ongoing assistance to build competence and confidence in using AI tools.

User Satisfaction: The system should be user-friendly and intuitive. Employees across

departments (auditors, accountants, compliance officers) should feel supported and empowered

by the system, not burdened.

Net Benefits: The ultimate goal of integrating AI into AIS is to reduce fraud incidents, improve

control accuracy, and increase organizational transparency. These benefits must be measurable

— for example, a decline in financial discrepancies, enhanced audit trail integrity, or improved

detection of irregular transactions.

By applying this model, banks can evaluate the real-world effectiveness of their AI investment. It

also serves as a feedback loop for continuous improvement, where shortcomings in system

quality or user satisfaction can be addressed to maximize the system’s fraud prevention

capabilities.

Conclusion

The Information Systems Success Model offers a comprehensive and structured framework for

evaluating the performance, impact, and sustainability of AI-integrated AIS in the banking

sector. Its relevance to this study lies in its ability to bridge technical performance with

organizational outcomes — especially in the area of fraud prevention. In the context of

Maiduguri’s commercial banks, the ISSM helps to assess not just whether AI tools have been

46
implemented, but whether they are effective, trusted, and beneficial across all levels of the

organization.

All the above theories combine to provide a strong foundation for this study:

TAM and UTAUT explain how AI tools can be adopted by bank staff.

Fraud Triangle Theory explains why fraud happens and where AI can be used to stop it.

Systems Theory and Contingency Theory help us consider how banks differ and how AI should

be tailored.

IS Success Model helps us measure if the AIS is effective after AI is added.

2.5 Conclusion

The review of literature presented in this chapter has revealed a rich and evolving body of

knowledge concerning the integration of Artificial Intelligence (AI) in Accounting Information

Systems (AIS) for the purpose of fraud prevention in commercial banks. Scholars and

practitioners alike have provided various perspectives on the definitions, typologies, causes, and

effects of fraud. Additionally, the traditional methods of fraud detection and prevention, such as

internal control systems, auditing procedures, and whistleblowing mechanisms, were found to be

reactive, limited in scope, and increasingly ineffective in the digital banking era.

The conceptualization of AIS has also been explored, particularly its core components, functions,

and evolution in line with technological advancements. The vulnerabilities of AIS to fraud,

especially in underdeveloped banking infrastructures, underscore the need for innovative

approaches such as AI. The review further discussed the fundamentals of AI, its branches, and its

practical applications in automating tasks, enhancing auditing procedures, and improving fraud

detection accuracy through machine learning models, natural language processing, and real-time

analytics.

47
A number of schools of thought were encountered in the literature. Some scholars argue that AI

integration in AIS can revolutionize fraud prevention through predictive capabilities, while

others caution against over-reliance on AI due to ethical concerns, high cost of implementation,

and data quality issues. Moreover, there is ongoing debate regarding the readiness of developing

economies, such as Nigeria, to adopt AI at scale within their financial institutions. This is

compounded by regulatory uncertainties and infrastructural gaps.

From the empirical studies reviewed, it is evident that although several countries and institutions

have adopted AI-driven fraud prevention systems, commercial banks in Maiduguri, Borno State,

are still at a nascent stage of implementation. Most of the reviewed works focus on either AI in

general accounting or fraud prevention separately, with few combining both within the context of

Nigerian banks, especially at the regional level.

This observation leads to the identification of a major gap in literature — there is limited

empirical evidence on the actual extent, effectiveness, and challenges of AI adoption in AIS

specifically for fraud prevention in commercial banks in Maiduguri, Borno State. Most studies

are either theoretical, lack local data, or fail to account for the unique environmental,

technological, and regulatory conditions present in the region.

This study, therefore, seeks to fill that gap by providing a focused investigation into how AI is

currently being adopted in AIS, evaluating its effectiveness in fraud detection and prevention,

identifying implementation challenges, and exploring the local realities within Maiduguri’s

banking sector. By doing so, the study aims to contribute to both academic knowledge and

practical frameworks that can inform policy and operational improvements.

48