Sri Lanka should explore a community-owned microfinance system that places borrower protection at the centre of lending, with artificial intelligence used to support financial services rather than make decisions about borrowers, according to a proposal by an academic specialising in generative AI and machine learning.

The proposal, advanced by Dr Nalinda Somasiri, Associate Professor at York St John University in the United Kingdom, calls for a system in which communities have a meaningful role in designing and overseeing financial services, while trained professionals retain responsibility for lending decisions and borrower protection.

It comes following concerns about household debt, irregular incomes and the limitations of relying on small loans to address financial insecurity.

The proposed model would combine savings facilities, appropriately structured loans, emergency assistance, financial education and business support. It would also explore whether carefully selected AI tools could help identify financial hardship, improve access to information in Sinhala and Tamil, and strengthen institutional oversight.

Historical survey highlights gaps in financial inclusion

A nationwide household survey conducted by the Institute of Policy Studies of Sri Lanka (IPS) in 2004–05 found that 32% of surveyed households continued to use informal credit sources, including moneylenders, family members and traders, despite having access to formal microfinance institutions.

The survey covered 1,480 households across 50 Grama Niladhari divisions in 17 districts. Among microfinance participants, 34% also relied on informal finance.

The findings are historical and do not establish the current extent of multiple borrowing in Sri Lanka. However, they highlight a longstanding challenge for financial inclusion: access to formal credit does not necessarily eliminate reliance on informal lenders.
The same survey found that 54.5% of microfinance participants had no institutional savings account before joining a scheme. It also reported that 50.2% of households had experienced a fall in income during the preceding 12 months.

These findings raise questions about whether lending programmes adequately account for unstable household incomes and the need for financial protection during emergencies.

Repayment rates may not tell the whole story

The proposal argues that repayment performance alone is an inadequate measure of whether microfinance improves household welfare.

Families may continue paying instalments while cutting essential spending or borrowing from other sources to meet their obligations.

Fixed repayment schedules can become particularly difficult for households whose earnings depend on seasonal work, overtime or daily demand.

Rather than treating every low-income household as a potential borrower, the proposed approach would distinguish between families needing social assistance, households vulnerable to sudden financial shocks, existing small-business operators and people ready to invest in new economic activity.

For some households, grants or emergency support may be more appropriate than additional loans. Others may benefit from accessible savings, flexible repayment arrangements or business financing accompanied by training and market support.

The IPS survey found that 58.2% of formal loan amounts were directed towards income-generating activities, while 40.2% of informal loan amounts were used for consumption.

The figures illustrate the different purposes borrowing can serve, including maintaining basic living standards when income falls.

Community ownership proposed as alternative

The proposed model would draw on Sri Lanka's experience with cooperative institutions and Samurdhi community-based banking, alongside lessons from Bangladesh's Grameen Bank.

Under the suggested approach, selected communities would participate in a pilot involving member representatives and trained financial staff. Members would have a role in shaping services and holding providers accountable, supported by transparent charges, clear lending conditions and independent complaints procedures.

The proposal also calls for safeguards against coercive group pressure and public shaming. Borrowers' financial records would remain confidential, rather than being made available to other community members.

Women's participation would need to extend beyond attendance at meetings to meaningful influence over financial decisions.

The model remains a proposal requiring consultation and feasibility assessment. Its viability would depend on effective governance, professional management, appropriate regulation and independent evaluation.

AI to support decisions, not replace people

Artificial intelligence could assist the proposed system by helping staff identify possible financial hardship, analyse irregular income patterns and explain financial products in Sinhala and Tamil.

Digital monitoring tools could also help authorised supervisors identify unusual lending patterns or recurring complaints.

However, the proposal rules out allowing AI to independently approve or reject loans, impose penalties, initiate collections or determine hardship restructuring.

Borrowers would retain access to human officers and non-digital services, including in-person assistance and telephone support. The system would also need safeguards against algorithmic bias, privacy breaches and the exclusion of people with limited digital access.
The proposal emphasises that AI should be adopted only where it demonstrably improves outcomes compared with simpler alternatives.

Independent evaluation urged before expansion

A pilot programme would need to measure more than the number of loans issued and the proportion repaid on time.

Key indicators would include household debt levels, reliance on high-cost informal borrowing, access to savings, livelihood income, financial resilience and borrower wellbeing.

The evaluation would also need to assess whether women exercise meaningful decision-making power, whether vulnerable households receive appropriate support and whether the system can operate sustainably.

The proposal does not suggest that community ownership or technology alone can solve household poverty. Instead, it argues for combining appropriate financial products with social protection, accountable institutions and viable livelihood opportunities.
For Sri Lanka, the central question is whether microfinance can move beyond the delivery of credit towards a broader system that protects households against financial shocks.

As the proposal puts it, the measure of success should not simply be whether borrowers repay their loans, but whether families become more financially secure.

Dr Nalinda Somasiri is an Associate Professor in Generative AI and Machine Learning and leads the AI for Climate & Disaster Resilience Research Group at York St John University, UK. The views expressed are personal.