How AI Is Transforming Credit Scoring and Lending

The Algorithm Will See You Now

When Mashreq Bank announced in February 2026 that its AI-driven credit engine had reduced personal loan default rates by 34 percent across its UAE and Egyptian portfolios, it confirmed what many in the industry had suspected: the traditional credit scoring model — built on rigid bureau data and backward-looking metrics — is being dismantled, piece by piece, by machine learning systems that see risk where humans cannot, and opportunity where legacy models see only blanks.

The shift is not incremental. A McKinsey Global Institute report published in January 2026 found that AI-augmented lending platforms now process $1.2 trillion in annual credit decisions globally, up from $340 billion in 2023. In the Gulf Cooperation Council states and across emerging markets from Nigeria to Indonesia, the transformation is arriving faster than in mature economies — precisely because the infrastructure gap that once held these regions back has become an advantage. Without legacy systems to defend, lenders are building natively intelligent credit architectures from the ground up.

Beyond the Bureau: Alternative Data and Financial Inclusion

The fundamental limitation of conventional credit scoring has always been its dependence on formal financial histories. In markets where 40 to 70 percent of the adult population remains unbanked or underbanked — Saudi Arabia's migrant workforce, Egypt's informal economy, sub-Saharan Africa's vast smallholder farming communities — FICO-style models simply produce no signal at all. These populations are not uncreditworthy. They are merely invisible to the existing system.

AI changes this calculus entirely. Tarabut Gateway, the Bahrain-headquartered open banking platform, now aggregates mobile payment histories, utility records, telecom data, and even e-commerce transaction patterns to generate credit profiles for individuals who have never held a bank account. The company reported in March 2026 that its alternative scoring model, deployed across banking partners in the UAE, Saudi Arabia, and Jordan, had enabled $780 million in new consumer lending to previously unscored individuals, with non-performing loan ratios holding at 2.1 percent — below the regional average for traditionally scored portfolios. That is a significant number, and it undercuts one of the oldest objections to alternative credit data.

In sub-Saharan Africa, Kenyan fintech M-Shwari — a product of the Commercial Bank of Africa and Safaricom partnership — now uses an AI model trained on M-Pesa transaction data to issue micro-loans within 90 seconds. The platform crossed 45 million cumulative loan disbursements in early 2026. Nigeria's FairMoney, backed by Tiger Global, processes over 200,000 AI-scored loans monthly, pulling in device metadata, behavioural signals, and psychometric inputs alongside traditional data points.

Private Wealth and Family Offices: The Quiet Adoption

While much attention focuses on consumer and SME lending, a less visible but equally significant transformation is underway in the private credit strategies of Gulf family offices and sovereign-adjacent investment vehicles. Few outside the region have noticed. Several prominent single-family offices in Riyadh and Abu Dhabi have begun deploying AI credit assessment tools to evaluate direct lending opportunities — bypassing traditional bank intermediation altogether.

Wamda Capital, the Dubai-based venture firm with deep family office ties across the GCC, disclosed in its Q1 2026 investor letter that it now uses proprietary machine learning models to assess creditworthiness in its venture debt portfolio, analysing over 400 variables per borrower including real-time revenue data, customer churn patterns, and supply chain concentration risk. The result: a 28 percent improvement in risk-adjusted returns compared to its 2023 vintage deals scored through conventional due diligence.

Lombard Odier, which manages substantial wealth for Gulf-based families, launched an AI-enhanced private credit fund in January 2026 specifically targeting emerging market mid-cap lending. The fund uses natural language processing to scan regulatory filings, news sentiment, and litigation records across 14 jurisdictions in real time, supplementing quantitative credit models with qualitative intelligence that would take human analysts weeks to compile.

Regulatory Frameworks Struggle to Keep Pace

The speed of adoption has opened an uncomfortable gap between innovation and oversight. The Central Bank of the UAE issued updated guidance in April 2026 requiring all licensed lenders using AI in credit decisions to maintain "explainability protocols" — essentially demanding that institutions be able to articulate, in human-readable terms, why a loan was approved or denied. The Saudi Arabian Monetary Authority followed with similar draft regulations in May, proposing mandatory algorithmic audits for any AI system handling more than 10,000 credit decisions per quarter.

These measures reflect genuine concerns. ProPublica-style investigations into algorithmic bias have found that certain AI models trained on historically skewed data can perpetuate discriminatory lending patterns, particularly against migrant workers and women in informal employment. Tarabut Gateway's chief risk officer, Nadia Hasan, acknowledged the challenge at the Arab Monetary Fund's April conference in Abu Dhabi: "The technology can be profoundly inclusive or profoundly exclusionary. The difference lies entirely in the training data and the governance framework around it."

The European Union's AI Act, which entered full enforcement in February 2026, classifies credit scoring as a "high-risk" application, mandating transparency requirements that several Gulf regulators are now studying as potential templates for their own frameworks.

The Competitive Reconfiguration Ahead

The implications for established financial institutions are stark. Banks that fail to integrate AI credit assessment face a pincer movement: fintechs capturing the unbanked and underbanked from below, and sophisticated family offices and alternative lenders cherry-picking the best credits from above. The middle ground — the traditional bank lending book — is being compressed from both directions.

Emirates NBD has responded aggressively. The bank invested $150 million in its AI and data infrastructure through 2026 and hired over 200 machine learning engineers, many recruited from Silicon Valley and Bangalore. Group chief executive Shayne Nelson told the Dubai Fintech Summit in March that AI-driven lending now accounts for 41 percent of the bank's new consumer credit origination, up from 12 percent in 2024. That is a significant shift.

First Abu Dhabi Bank and Saudi National Bank have made comparable commitments. But smaller regional banks — particularly in Oman, Bahrain, and Kuwait — risk being stranded, unable to afford the talent and infrastructure required to compete, yet increasingly unable to price risk accurately without it.

What is emerging here is not a technological upgrade to an existing process. It is a fundamental reconception of what creditworthiness means, who gets to define it, and who profits from the definition. Across the Gulf and the developing world, that reconception is happening now — faster, and with higher stakes, than most incumbents seem to appreciate.