A new report by the International Finance Corporation, published in May 2026, examines how artificial intelligence and alternative data are transforming credit scoring for borrowers who lack formal financial histories, and finds a fast-growing sector whose benefits remain unevenly distributed.
The problem
The global financing gap for micro, small, and medium enterprises stands at $5.7tn, up $1.3tn since 2015. For businesses owned by women, it is $1.9tn and widening. Traditional credit scoring excludes anyone without repayment history, collateral, or formal income — which describes the majority of borrowers across Africa, South Asia, and other emerging markets:
- 3bn people worldwide lack adequate credit histories
- 1.3bn have no bank account at all
- Women are less likely to apply for credit, more likely to be rejected, and receive smaller loan amounts when approved
How the new models work
Rather than relying on formal financial records, alternative scoring systems draw on a much wider data pool:
- Mobile money and digital wallet activity
- Utility and bill payments
- Call and SMS metadata
- Geolocation and device usage patterns
- Psychometric assessments
AI is applied selectively, not end-to-end. Some firms automate scoring entirely via machine learning; others use AI only for transaction categorisation or fraud detection. Rule-based and hybrid approaches remain common, often due to regulatory constraints or the need to explain decisions to borrowers.
The market
The IFC mapped 448 alternative credit scoring firms globally, drawing on interviews with more than 30 experts and borrower-level data from Eshandi in Zambia and Vexi in Mexico. Key findings:
- Over 75% of mapped firms were founded in the past decade
- Personal loans and MSME credit account for 47% and 32% of offerings respectively
- North America and Europe dominate mature, well-funded segments; Africa has a disproportionate share of unfunded firms
- Buy-now-pay-later and agricultural lending are emerging as smaller but growing niches
The gender gap
Where alternative models have been deployed inclusively, women have consistently outperformed expectations:
- Eshandi in Zambia has disbursed nearly 1mn loans to women; female borrowers score higher and repay more consistently
- Yellow Factoring in Cameroon reports lower default rates and better loan terms among women
- Vexi in Mexico found that women use credit for business expenses more often than men and receive higher credit limits over time
Yet inclusion is largely accidental. Only 12% of mapped firms publicly reference women in their mission or model design. Gender is rarely used as a predictive variable, and biases embedded in training data (gaps in women’s digital footprints, mismeasurement of informal income) go undetected. Proxy variables such as education level, geography, or device type can quietly entrench discrimination even when gender is explicitly excluded.
What the IFC recommends
The report calls for coordinated action across regulators, lenders, and fintechs:
- Establish regulatory sandboxes to test AI models before market deployment
- Mandate fairness audits and bias checks across model lifecycles
- Require sex-disaggregated reporting on applications, approvals, loan sizes, and repayments
- Encourage banks to integrate alternative data responsibly alongside traditional frameworks
- Build open-finance frameworks combining interoperability with robust consent mechanisms
The underlying finding is both encouraging and cautionary. AI and alternative data have demonstrated real capacity to extend credit to borrowers that conventional systems have long ignored. Without deliberate design choices, however, that capacity will remain underused and the biases of the old system will be reproduced at greater speed and scale.
