Between phone bills and installment purchase records, credit algorithms reshape lending in Egypt and the Gulf

Financial institutions in the Middle East and North Africa are turning to artificial intelligence algorithms to decide on loan approvals for individuals and small businesses, driven by pressure on operating costs. Manually reviewing small financing requests drains staff time and consumes most of the lender’s earnings, according to World Bank estimates.McKinsey reports that automating these decisions with software sharply reduces operating costs.which gives banks and non-bank financiers an economic incentive to expand in the micro-finance and startup loan sectors.

In a financial environment where more than half of adults in Egypt lack bank accounts or digital wallets, according to World Bank data, traditional creditworthiness criteria appear unable to assess millions of new borrowers. Here, artificial intelligence steps in by analyzing non-bank alternative data. Ahmed Abaza, co-founder and CEO of Synapse Analytics, explains that the models review mobile phone data and examine the repayment behavior of telecommunications and utility bills to build a precise credit assessment. In the United Arab Emirates, the experiment goes further: since July, credit rating agencies have begun incorporating micro-loan repayment records from buy-now-pay-later platforms such as Tabi and Tamara into credit reports, with plans to have them examined by artificial-intelligence agents.

However, credit automation carries a silent algorithmic dilemma, as models train on historical records saturated with existing market biases. If women, youth and informal workers were absent from borrowers’ ledgers in previous contracts, algorithmic models tend to automatically exclude them from new credit lines. Hossam Haiba, managing partner at Orbis Capital and former CEO of Egypt’s General Investment Authority, notes that integrating these tools should enhance human efficiency rather than replace it, stressing that algorithmic outputs require continuous human review for potentially impactful assessment errors.

This push toward automation stops at the limits of trade financing and large corporations. Bassem Faiq, managing director and partner at Boston Consulting Group, confirms that banks already use artificial intelligence to assess companies’ debt-servicing capacity, yet leaving the final decision to the machine is a critical risk for institutional reputation.Credit responsibility is not delegated to an algorithm when a bank’s reputation and depositor funds are at stake.Todd Wilcox, CEO and vice chairman of HSBC Egypt, summarized the stance at the Enterprise Forum, stating that no banking leadership can justify institutional shortcomings to depositors and shareholders by relying on a ruling from an artificial-intelligence model.

Regulatory bodies in the region are seeking to keep pace with these shifts through legislation that frames the use of alternative data. In Egypt, the General Authority for Financial Supervision requires consumer finance companies, starting in April 2027, to refer to an index issued by the country's sole licensed private credit information company, which is based on behavioral analysis of non-traditional data to feed risk algorithms. In terms of infrastructure, Emirates NBD is preparing to become the first bank in the region to run ChatGPT on local servers within the UAE, and it is the bank that ranked first among 25 regional banks in the Evident AI adoption index.

This shift requires risk and credit officers in Egypt and the Gulf to recalibrate financing strategies and data governance.The operational challenge goes beyond merely consuming alternative data; it also involves building auditing mechanisms that dismantle model bias against financially excluded groups.Before automated decisions become systematic exclusion that harms the market. Running models within local data centers, as leading Gulf banks do, also establishes a new mandatory standard for protecting depositor confidentiality and complying with digital sovereignty requirements.