Study reveals how Kenyan banks lose billions to fraud

Artificial intelligence (AI)-driven transaction analysis can explain a striking 82.7 per cent of Kenyan commercial banks’ financial performance, a new study has found, underscoring the powerful link between advanced monitoring and bank profitability. 

Beyond simple detection, advanced transaction analysis, particularly with the aid of AI, is emerging as the primary bulwark against increasingly cunning financial criminals, shifting the focus from reaction to proactive prevention.

The revelations come as the Central Bank of Kenya (CBK) faces public scrutiny for failing to prevent bank failures linked to ill health and governance weaknesses. 

Past incidents such as the collapse of Imperial Bank have often been attributed to deep-seated fraud that went undetected for years, highlighting critical vulnerabilities in the banking system’s defences and questioning the effectiveness of past regulatory and auditing practices. 

This context underscores the pressing need for robust oversight and early detection mechanisms within the financial sector.

The new study, “Transaction Analysis and Financial Performance of Commercial Banks in Kenya,” published in the July 2025 issue of IRE Journals, was authored by Kenyan researchers Robert Ojiambo Sande, Rashid Simiyu Fwamba, and Joshua Olang’o Abuya. 

Iconic Research and Engineering Journals (IRE Journals) is a globally recognised platform for the dissemination of academic and scientific research.

Sande and Fwamba are affiliated with the Department of Economics, Finance and Accounting at Kibabii University, while Abuya is from the Department of Business Administration and Management at the same institution. 

They conducted their research by employing a descriptive design, targeting 111 professionals comprising forensic accountants, finance officers, and auditors from 37 commercial banks headquartered in Nairobi.

The study found an “extraordinary correlation” between a bank’s robust transaction analysis and its financial health.

This statistical measure indicates that over four-fifths of the variations in a bank’s financial performance can be attributed to its transaction analysis practices, which signifies that investment in monitoring technologies is not merely a compliance burden but a strategic imperative for profitability and resilience. 

The findings “demonstrate that effective transaction monitoring is a critical success determinant,” the report asserts.

Innocuous deposit

Consider a large, illicit transfer attempting to slip through a bank’s defences, perhaps disguised as a legitimate inter-bank transaction, or a complex web of small, seemingly innocuous deposits designed to avoid standard thresholds – classic hallmarks of money laundering or sophisticated fraudulent schemes. 

Forensic investigators, whose insights were crucial to this study, often recount how fraudsters exploit blind spots: shell companies used to obscure ultimate beneficiaries, rapid movement of funds across multiple accounts to break the audit trail, or even the creation of ghost loans that exist only on paper. 

As the researchers note, the “increasing sophistication of financial fraud” demands tools capable of identifying “subtle, non-obvious patterns.” 

AI, with its capacity to analyse vast datasets and learn from past behaviours, is precisely poised to detect these anomalies with unprecedented accuracy, moving beyond the limitations of traditional rule-based systems.

AI and machine learning systems can serve as the next generation of “transaction analysis practices” recommended in the report. 

By moving beyond traditional, rule-based systems that can be easily outsmarted by cunning criminals, AI can process vast quantities of data in real-time to identify anomalies and suspicious patterns. 

The report’s finding that “transaction analysis demonstrates a strong positive relationship with financial performance” is directly supported by AI’s ability to improve the accuracy and speed of this analysis. 

This allows banks to not only detect fraud but to do so more efficiently, directly contributing to the “82.7 per cent of variance in financial performance” explained by these practices.

A key function of AI in this context is its ability to perform advanced anomaly detection and pattern recognition. Unlike human analysts who can only process a limited number of variables, AI models can analyse thousands of data points for each transaction. 

For instance, an AI system can identify a network of seemingly unrelated small transactions across multiple accounts, which, when viewed collectively, indicate a larger money laundering scheme. 

This capability is vital for combating “increasing sophistication of financial fraud” as mentioned in the report, allowing banks to uncover complex schemes that might have been missed by manual reviews or simpler software.

Furthermore, AI enables a shift from reactive to proactive fraud prevention. 

By analysing historical data on past fraudulent activities, AI models can predict and flag potential high-risk transactions before they are even completed.  This allows a bank to put a hold on a suspicious transaction and conduct a thorough review, potentially preventing financial loss before it occurs. 

This predictive power helps banks to not only stop crime but also to build more resilient and secure financial systems, directly aligning with the report’s conclusion that banks “should prioritise investment in advanced transaction analysis systems.”

Finally, AI significantly enhances operational efficiency and resource allocation.  By automating the screening of the majority of routine transactions, AI systems free up the time of highly skilled professionals like forensic accountants, finance officers, and auditors who were interviewed for the study. 

Operational costs

These experts can then dedicate their specialised knowledge to investigating the most complex and high-value alerts generated by the AI. 

This improved workflow not only makes the fraud detection process more effective but also reduces operational costs and improves overall productivity, reinforcing the study’s central argument that robust transaction analysis practices are a fundamental driver of a bank’s financial health. 

The researchers specifically highlighted that “timely fraud detection” and “uncovering irregularities” were among the highest-rated benefits of transaction analysis practices by the surveyed professionals.

The ghost of past bank collapses in Kenya looms large, serving as a stark reminder of the consequences of undetected fraud and weak oversight. 

Cases like that of Imperial Bank, which was placed under receivership in 2015 due to what the CBK later described as “widespread fraud” involving billions of shillings, illustrate the catastrophic impact of unchecked financial malpractices.  In such instances, forensic audits later revealed a systematic scheme of off-book transactions and fictitious loan accounts that bypassed internal controls for years. 

This raises pointed questions about how auditors, both internal and external, along with regulators, missed significant signs of ill-health and potential fraud that were festering within the banks. 

“Banks with robust transaction monitoring systems show significantly better financial outcomes,” the report states, directly addressing the inverse: banks without such systems are at higher risk. While official figures are often elusive, Kenyan banks are understood to collectively lose billions of shillings annually to increasingly complex financial fraud schemes. 

These significant financial hits are not always publicly disclosed, as financial institutions often opt to absorb the losses or manage them internally to safeguard their reputations. 

The fear of triggering customer panic or undermining investor confidence often outweighs the transparency that might otherwise come with publicising the full extent of fraudulent activities, leaving a significant portion of these economic crimes shrouded in secrecy and contributing to a persistent challenge for the banking sector’s integrity. 

These past failures underscore the urgent need for a more proactive, technology-driven approach to transaction monitoring that can alert authorities to irregularities long before they reach systemic proportions, the study suggests.

In an era where “increasing financial crime sophistication” is rapidly evolving in complexity, leveraging advanced analytical tools is becoming non-negotiable, the researchers say.

The report recommends that commercial banks “prioritise investment in advanced transaction analysis technologies, including AI and machine learning systems.” 

This strategic shift towards AI-powered systems is vital for “enhanced fraud detection, improved efficiency, and stronger stakeholder confidence,” according to the findings.

While commercial banks are urged to upgrade their systems, the study recommends that “the CBK should incorporate transaction analysis effectiveness into formal supervision processes,” including setting “minimum standards for monitoring capabilities” and providing “best practice guidance”. 

This recommendation suggests that the current supervisory framework may not adequately evaluate the effectiveness of these critical systems, creating a de facto “loophole” or an area ripe for strengthened regulatory scrutiny. 

By formalising this oversight, the CBK could ensure that all 37 licensed commercial banks in Kenya adopt and effectively utilise these technologies, mitigating systemic risks and fostering greater financial stability, concludes the study.