Banks Turn to AI and Alternative Data as Non-Performing Loans Rise to Sh695.4 Billion
NAIROBI, Kenya, Aug 10 – Kenya’s banking sector held Sh695.4 billion in non-performing loans at the end of March 2026, up Sh21 billion from Sh674.4 billion in December and lifting the default ratio to 15.6% from 15.4%, according to Central Bank of Kenya data.
The increase came despite lower interest rates, and CBK Governor Kamau Thugge attributed it to deterioration in the personal and household, trade, agriculture and manufacturing sectors.
Lenders are responding by extending artificial intelligence beyond the loan approval stage, using machine learning, natural language processing and large language models to monitor borrowers continuously, flag financial distress early and intervene before an exposure sours.
Caritas Microfinance Bank chief executive David Mukaru identified the application of AI and data analytics to non-performing loans as one of the industry’s central challenges.
The models classify borrowers into risk categories by drawing on alternative data including mobile money transactions, utility bill payments and merchant activity alongside conventional payslips and bank statements, a shift that matters most for small businesses and informal workers whose digital footprints reveal more than thin credit files.
Absa Bank chief operating and digital officer Julius Kamau said non-financial and alternative data are proving better at evaluating creditworthiness and reducing risk, adding that customer demand for hyper-personalized products depends on banks knowing clients in far greater depth.
AI systems also function as early warning tools, alerting lenders to declining income, delayed payments or sector-wide stress so repayment schedules can be restructured before default.
Final lending decisions still rest with human judgment, a Thomson Reuters report notes, oversight that has grown in importance as consumer advocates warn that models trained on flawed or incomplete data could penalise informal traders, freelance workers and borrowers from particular neighbourhoods.
