Fraud detection in Finance

⭐ Fraud detection in Finance
I would like to highlight how important is it to create a fraud detection system and how easy it gets when you know the basics. I can guide you to chose the best model to create a fraud detection system integrated with KYC and how powerful it gets.

⭐ AI in Fintech-Fraud,Credit & Inclusion
AI in Fintech — Fraud, Credit & Inclusion

Can AI Turn Mobile Money Data into Better Credit and Safer Financial Services?

Artificial intelligence is rapidly changing financial services. Across Africa, AI is creating opportunities to improve how financial institutions assess risk, detect fraud, serve customers and extend financial services to people who have historically remained outside traditional banking systems.

But there is an important question behind this opportunity: How can we use AI to expand financial inclusion and strengthen financial security without allowing technology to reproduce or amplify the problems we are trying to solve?

My interest in this question comes from experience at the intersection of microfinance, financial services and audit. Before joining Deloitte Cameroon as an auditor, I worked as a Technical Field Assistant with CamCCUL, supporting audit work within microfinance institutions. This experience exposed me to communities where people were economically active and had genuine opportunities to generate income, yet many struggled to access appropriate financial services. I encountered farmers, small business owners and entrepreneurs whose economic activity was not necessarily reflected in the traditional information financial institutions use to assess creditworthiness.

This experience shaped my interest in the potential of digital financial data and AI.

What if AI could help financial institutions better understand economically active people who are invisible to traditional credit systems, while simultaneously making digital finance safer?

1. The financial inclusion opportunity: looking beyond traditional credit data

Traditional credit assessment often depends on information such as formal employment, bank statements, previous loans, collateral or established credit histories.

For many microentrepreneurs, farmers and informal businesses, these signals may be limited or completely absent. A person may have a viable business and regular income but still be considered “thin-file” or even “no-file” because their financial activity does not appear in traditional credit systems.

This is where alternative data becomes important.

Mobile money transactions can create a digital financial footprint. With appropriate customer consent, legal safeguards and data-protection compliance, legitimate transaction data can potentially provide additional indicators such as the consistency of cash flows, recurring business payments, transaction frequency, average transaction values, repayment behaviour, seasonal income patterns and unusual changes in normal activity.

The objective should not be to say, “This person uses mobile money, therefore they deserve a loan.” A single transaction cannot establish creditworthiness.

Rather, the opportunity is to combine multiple legitimate signals to create a richer picture of a customer’s financial behaviour.

This distinction is important because alternative data does not replace responsible credit assessment. It supplements the information available to financial institutions.

The IFC’s recent work on alternative data and AI highlights the potential of mobile-money transactions, digital payments and business or platform records to improve credit assessment for underserved and thin-file borrowers, including small businesses and entrepreneurs.

2. Why Cameroon is particularly interesting

The opportunity becomes even more relevant as digital financial activity expands.

According to the World Bank’s Global Findex 2025, 60.9% of adults in Cameroon had an account in 2024. Importantly, the Global Findex definition of account ownership includes not only accounts at banks and financial institutions, but also individuals who personally used a mobile-money service.

However, access is not evenly distributed. The same World Bank data shows that 52.9% of women in Cameroon had an account in 2024.

These figures illustrate both progress and opportunity.

AI cannot create financial inclusion from nothing. It needs data.

And Africa is increasingly generating digital financial data through mobile money, digital payments and other platforms.

Cameroon is also becoming part of a broader regional movement to strengthen the infrastructure around credit information. In January 2026, the Bank of Central African States (BEAC), with support from the IFC, launched the CEMAC Credit Information Bureau, BIC-CEMAC, operated by CreditInfo Central Africa. Its objective is to reduce information asymmetry, strengthen credit-risk management and improve access to finance, particularly for MSMEs and underserved segments. The IFC has also identified the adoption of alternative data for credit decision-making as a component of its CEMAC Alternative Data project.

This creates an important foundation.

Credit infrastructure can make financial information more accessible. Alternative data can make the financial footprint richer. AI can make it possible to analyse that information at scale.

3. From mobile money data to AI-assisted credit decisions

Consider a small agricultural entrepreneur.

She may not have a formal salary, substantial collateral or a long history of bank borrowing. However, she may receive regular payments, pay suppliers, make business-related mobile-money transactions and demonstrate recurring financial activity over several seasons.

Individually, these transactions may tell us very little.

But thousands or millions of transactions can reveal patterns.

AI and machine-learning systems can analyse large volumes of historical data much faster than humans. They can identify relationships between financial behaviour and outcomes such as repayment or default and use these patterns to support risk assessment.

The process could look like this:

Mobile Money Activity → Alternative Data → AI Analysis → Risk Assessment → Human Decision → Credit

The crucial word is support.

AI should not automatically become the final decision-maker simply because it can process more information. A responsible financial institution should understand what data the model is using, whether that data is reliable, what the model is actually predicting and whether its outcomes are appropriate.

The objective is not simply to approve more loans.

It is to enable financial institutions to make better-informed, more proportionate and potentially more inclusive decisions.

4. The same technology can strengthen fraud detection

The opportunity does not stop at credit.

As financial transactions become increasingly digital, financial crime can also operate at greater speed and scale.

The GSMA estimated that mobile-money fraud in Africa exceeded $1 billion in 2023, demonstrating the scale of the challenge facing the digital financial ecosystem.

AI can support fraud detection by identifying patterns that may be difficult to detect manually.

For example:

Transaction → Pattern Analysis → Risk Score → Alert → Investigation → Action

A customer’s behaviour may normally follow a relatively stable pattern. Suddenly, there may be unusual transaction frequency, unexpected geographic behaviour, abnormal amounts or interactions with accounts associated with suspicious activity.

AI can analyse these signals continuously and help financial institutions identify potentially fraudulent behaviour earlier.

The objective should not be to block every unusual transaction.

It should be to develop intelligent risk detection that protects customers while minimising unnecessary disruption to legitimate financial activity.

There is also evidence of the growing importance of advanced data infrastructure within the mobile ecosystem. LigaData describes its work with MTN across multiple markets to establish a governed data foundation that unifies operator and subscriber information and supports analytics, AI and applications including mobile financial services.

For me, the broader lesson is particularly relevant to Cameroon:

Before AI can make better financial decisions, organisations need the ability to collect, connect, govern and analyse data.

5. Considerations: AI does not automatically make fintech fair

The opportunity is significant, but responsible deployment requires us to consider what can go wrong.

Algorithmic discrimination

AI learns from data.

If historical lending data contains inequalities or biased patterns, an algorithm may reproduce those patterns. It can also identify new correlations that unintentionally disadvantage particular groups.

Therefore, AI does not automatically remove human bias from fintech.

It can reproduce existing biases, create new forms of discrimination, and allow both legitimate financial institutions and criminals to operate at greater scale.

Financial institutions should therefore ask:

* Are approval and rejection rates materially different across relevant customer groups?
* Are proxy variables unintentionally reproducing sensitive characteristics?
* Is the model disadvantaging customers because of characteristics unrelated to their ability to repay?
* Can the institution explain why a customer received a particular risk assessment?

Fairness cannot be treated as a one-time model-development exercise. It requires continuous monitoring.

Deepfakes and AI-enabled financial crime

AI also gives criminals more sophisticated tools.

Deepfake images, cloned voices, synthetic videos and AI-generated documents can be used to impersonate customers, executives or other trusted individuals. These techniques can strengthen social engineering and potentially undermine traditional identity-verification processes.

This means financial institutions must evolve their controls.

A voice, photograph or AI-generated document should not automatically be treated as proof of identity or authenticity. Stronger authentication, transaction monitoring, behavioural analysis and layered verification become increasingly important.

AI hallucinations and incorrect outputs

Generative AI introduces another consideration: systems can produce information that sounds convincing but is factually incorrect.

In financial services, this can have serious consequences.

Imagine an AI system incorrectly stating that a customer previously defaulted on a loan, or generating an inaccurate explanation for a credit decision.

This is why high-impact financial decisions should rely on verified data and controlled systems, with appropriate human oversight.

AI can analyse, identify patterns and recommend.

But where the consequences are significant, institutions must be able to verify the output and identify who remains accountable for the decision.

Data quality and governance

Finally, responsible AI begins before the algorithm.

Garbage in, garbage out.

If mobile-money or financial data is incomplete, inaccurate, duplicated, poorly linked or incorrectly classified, an advanced algorithm can simply produce sophisticated decisions from unreliable information.

A responsible AI lifecycle therefore needs to look beyond model performance:

Data Collection → Data Quality → Data Governance → Model Development → Testing → Deployment → Continuous Monitoring

This is where the role of governance, risk management and audit becomes increasingly important.

6. Where audit and risk management fit into the AI-fintech equation

As AI becomes embedded in financial services, the question should not only be, “Does the model work?”

We should also ask:

Can we trust the data?

Can we explain the decision?

Can we detect systematic disadvantage?

Can we identify manipulated or AI-generated information?

Can we reconstruct how a significant decision was made?

Who is accountable when the system is wrong?

These are not purely technical questions. They are governance, risk and control questions.

This creates an important opportunity for auditors, risk professionals and financial institutions to work alongside technology teams.

The future of responsible AI in fintech should not be built by technology teams alone.

It should be a multidisciplinary effort involving data specialists, financial institutions, regulators, risk professionals, auditors and business leaders.

7. The bigger opportunity for Cameroon

Financial inclusion should ultimately mean more than simply owning an account.

It should mean being able to participate meaningfully in the economy: to save, transact, manage risk, obtain appropriate financing, grow a business and protect oneself from financial crime.

Imagine a woman running an agricultural business who has never received a traditional bank loan. Her lack of formal credit history should not automatically make her economically invisible if legitimate digital financial activity provides additional evidence of her financial behaviour.

Mobile money can create the footprint.

Alternative data can make that footprint useful.

AI can identify patterns within it.

Credit infrastructure can connect information across the financial ecosystem.

And governance, risk management and audit can help ensure that the resulting system remains fair, reliable, secure and accountable.

That is the opportunity I see for Cameroon and the wider African fintech ecosystem.

Conclusion

My experience in microfinance and my current work in audit have shown me two sides of the same problem.

I have seen people with genuine economic potential struggle to access appropriate financial services. I have also seen why financial institutions need reliable information, strong processes and effective controls to make responsible decisions.

AI gives us the possibility of connecting these two realities.

But we should not think of AI as a magic solution.

Scale works both ways.

It can scale financial inclusion, and it can scale fraud.

It can scale better decisions, and it can scale existing bias.

The answer is not to reject AI.

The answer is to build the conditions under which AI can be trusted.

That means moving from:

DATA → AI → DECISION → CONTROL

We need better data.

We need responsible AI.

We need informed human decisions.

And we need strong governance and continuous controls.

The question I would like to explore at CITS26 is therefore not simply whether AI can transform fintech.

It is:

How do we use the data and technology we already have to build a financial system that sees more people, serves more people and protects more people?

That, to me, is the real opportunity for AI in African fintech.