Blog

Digital agriculture is advancing. So why is it not scaling?

Posted on by harshita

Plant now or wait? For Makiwa, choosing the wrong answer could affect an entire season. 

A 49-year-old farmer with a diploma in business administration, Makiwa manages her farm with a smartphone. She uses several agricultural apps for weather updates, planting advice, pest alerts, input recommendations, and market information. Yet, the advice does not always align. One app tells her to plant, while another warns her to wait. When the stakes are high, she turns to her extension officer. Her experience highlights a growing reality in digital agriculture. More information does not always lead to better decisions. Farmers still need support that is trusted, coordinated, and actionable. The question is what a system built around that principle would look like 

Makiwa’s experience is not unusual. Across Africa, digital agriculture expands through farmer registries, e-wallets, subsidy platforms, weather advisories, market information systems, and digital credit solutions. While the contexts differ, examples from elsewhere show how digital systems can better coordinate services and support farmer decision-making. Bihar Krishi in India offers one such example. It brings multiple agricultural services together through shared digital infrastructure and common digital foundations. By connecting services that are often delivered through separate platforms, it aims to provide farmers with more coordinated and consistent support. 

A landmark study by the Technical Centre for Agricultural and Rural Cooperation (CTA) identified nearly 400 active digital agriculture solutions and more than 33 million registered farmers and pastoralists across the continent by 2019. Kenya Integrated Agricultural Management Information System (KIAMIS) has registered more than 7 million farmers. This growing digital infrastructure provides a foundation for agricultural transformation. Yet, the real question is whether these systems help farmers make better, faster, and more confident decisions. 

The digital reach does not always translate into sustained use. The Global System for Mobile Communications Association (GSMA) estimates that nearly one billion Africans live within mobile broadband coverage but do not use mobile internet. The same gap affects digital agriculture, where solutions continue to grow. Yet, many smallholder farmers, agribusinesses, and public agencies do not use them for daily decision-making. 

The challenge often lies less in the technology than in the failure of the delivery model. Pilots succeed because projects absorb the costs of onboarding, training, field support, devices, data, and incentives. When funding ends, support declines, ownership becomes unclear, and platforms struggle to sustain use. Technology may work, but the model’s long-term viability remains uncertain. 

Successful pilots do not automatically translate into sustainable scale. Many digital agriculture initiatives show that a solution can work under controlled conditions. Fewer continue to deliver value when donor funding, project support, and intensive onboarding decline. DigiFarm in Kenya highlights the need to move beyond standalone farmer platforms toward ecosystem models. These models connect farmers with financial institutions, input providers, markets, and service providers. 

Similarly, MSC’s work on Bihar Krishi in India shows how reusable digital foundations, interoperable systems, and AI-enabled services can support integrated farmer solutions at scale. These experiences show that the challenge goes beyond digital tools. It also involves sustainable models in which different actors can participate to create value beyond the initial project cycle. 

In digital agriculture, onboarding does not show the impact. Registrations, downloads, and messages sent may look impressive, but they do not show whether farmers make better decisions, improve productivity, increase incomes, or reduce risks. MSC’s experience in digital financial services reinforces this lesson. Meaningful inclusion depends on sustained usage, trust, and customer value, rather than account opening. 

Farmers continue to use digital services when they solve real problems. These services help them access the right advice, inputs, finance, markets, or risk-management solutions when they need them. Sustainable scale, therefore, requires a shift from isolated applications toward interoperable ecosystems. Shared digital infrastructure can connect multiple services into a single and coherent farmer journey. 

Figure 1: Fragmented apps create duplication, fatigue, and weak trust 

The solution requires more than a single government-run platform. It encompasses a well-governed, interoperable ecosystem where different actors connect through shared digital infrastructure. The World Bank’s Digital Agriculture Roadmap Playbook highlights the importance of reusable, modular building blocks rather than siloed systems. 

Farmers should be able to register once through a trusted digital identity or farmer registry and access advisory services, payments, insurance, credit, and markets without repeated onboarding. Shared digital foundations also allow service providers to focus on innovation and customer value rather than rebuilding parallel systems. 

However, interoperability alone will not drive adoption. Digital services must also be trusted, affordable, and aligned with farmers’ realities. Tools may appear affordable during pilots because projects absorb the costs of onboarding, training, devices, and support. Farmers will pay only when the value is immediate and reliable. Seasonal payments, pay-as-you-use models, and embedded service fees may be a better fit than annual subscriptions. 

Digital services also require trust, as farmers share sensitive data on land, production, and credit behavior. Adoption will remain weak unless farmers understand who controls their data, how providers use it, and what they receive in return. Trust is not optional. It is infrastructure for scale. 

Governments should shape the ecosystem rather than build every application. Their role is to establish foundational digital infrastructure, set priorities, protect farmer data, enable interoperability, and create an environment for private innovation. MSC’s experience with India’s AgriStack and Bihar’s Digital Farmer Services platform demonstrates the value of trusted farmer registries, which support data systems, consent frameworks, and common standards. These foundations can create shared digital rails for multiple public and private services. 

For African countries, the lesson is to invest in adaptable digital building blocks rather than replicate a specific model. These building blocks can enable diverse actors to deliver services that support food security, climate resilience, and market access. 

Figure 2: Bihar Krishi brings 16 integrated features into a single farmer-facing digital platform, spanning advisory, schemes, markets, grievance redressal, finance, and agricultural services. 

On 19th May 2025, 14 months after its launch, the Bihar Krishi platform had 1.61 million registered farmers. Around 0.35 million monthly active users represent roughly 22% of the registered base. This data shows that engagement extends beyond initial onboarding. 

Farmer use varies each quarter based on crop cycles and needs. The 22% figure, therefore, understates the platform’s broader use over time. Bihar Krishi demonstrates how government investment in shared digital foundations can unlock an open ecosystem in which public and private actors build services on top of these foundations. These services range from advisory and market services to finance and climate resilience. 

 

Figure 3: Bihar Krishi internal platform analytics and operational dashboard for the reported period. The monthly active share and the 94.4% smallholder figure are based on the reported numbers. 

So, what needs to change? Digital agriculture must move from fragmented pilots to reusable ecosystems. Governments should invest in shared digital foundations and use open digital public goods (DPGs). They can adapt proven solutions to local priorities rather than build parallel systems from scratch. Donors should measure success through sustained usage, farmer outcomes, and long-term sustainability beyond project funding. 

Technology can support agricultural transformation, but people remain central to its success. The future is human-enabled rather than digital-only. Extension officers, cooperatives, agro-dealers, and agents of financial service providers remain critical. They help farmers interpret, trust, and use digital services. 

The strongest models equip trusted intermediaries with better tools to serve farmers more effectively. As a result, open, interoperable, and trusted ecosystems that turn digital access into better decisions and stronger livelihoods. 

For farmers like Makiwa and farmers in Bihar, scale means more than access to digital tools. It means connected, effective, and reliable advice when decisions matter most. Interoperable systems, built on trusted digital foundations and strengthened by human intermediaries, can turn fragmented information into timely, confident decisions. Farmers need information that is relevant, timely, and trustworthy, which enables them to make informed decisions when it matters most. When digital services work together, farmers like Makiwa no longer need to navigate competing advice on their own. They gain the confidence to act, invest, and plan for the future.  

Why wait for a complaint? How automation and AI can detect missed social protection entitlements

Posted on by harshita

When a tranche of social protection payment does not arrive, the government’s digital systems know about it before the beneficiary does.

Usually, in such cases, one system records that the person is enrolled. Another shows that payment is due. A payment platform records whether an instruction was issued, while the receiving bank returns a failure code. Yet, these signals frequently remain scattered across institutions. The burden of joining them together falls on the person who has not received the benefit.

The recipient must notice that an installment is missing and check whether others have received it. They must then check the online status and interpret any unfamiliar error message. Finally, they must identify the right institution to approach, such as the bank, the local administration, the program helpline, or the identity service provider. The complaint may then move between departments because no official has a complete view of the delivery chain.

As social protection systems become increasingly digital, governments should ask a simple question: If the state can detect that an expected entitlement has not been delivered, why should it wait for the beneficiary to report the failure?

This blog uses India as its main example because the country’s Direct Benefit Transfer (DBT) programs combine large beneficiary registries with digital payment systems. The underlying proposition is more broadly relevant, particularly to countries with transaction-level program and payment data.

From reactive grievances to proactive analysis

Governments can solve this issue by building a reconciliation and case-detection layer on top of existing social registries, program management systems, and payment infrastructure. This layer would reconcile what a person was expected to receive with what the delivery system confirms they received.

The process should begin with a rule engine rather than an artificial intelligence (AI) model at this stage. For recurring payments, a rule may identify the following discrepancy:

Most coded failures could be diagnosed and routed through predefined rules. AI could address the remaining cases where the response is generic, records are conflicting, or the available code does not reveal the underlying cause. In such cases, AI could assemble the evidence, compare the case with verified history, and suggest a possible cause for official review. It could also convert fragmented records into a short operational report, as illustrated:

The case-detection layer would assign the case, start a resolution clock, and notify the program recipient. If the recipient is needed to act, the system would send a message that explains the required action and suggests the next steps. If the failure is internal, the administration should retain responsibility for correction.

How could the approach work?

 

These layers in the approach are distinct. Rules identify and route known discrepancies, while AI assists with unresolved diagnoses, summarization, and pattern recognition. Meanwhile, humans verify the assessment and authorize consequent action.

AI should never independently terminate a benefit, alter eligibility, modify a bank record, or close a case. Where records conflict, the model should flag the inconsistency rather than choose which database to believe. Only an authorized official should make any decision that could reduce or stop assistance, with an intelligible explanation and an opportunity for the beneficiary to respond.

Privacy by design, instead of unlimited data access

Proactive problem resolution must not justify the creation of a centralized profile for every beneficiary. An AI model does not need unrestricted access to Aadhaar numbers, bank accounts, balances, biometrics, or complete household histories.

A secure gateway could provide only the minimum necessary information. This gateway could include whether a payment was expected, whether it was initiated, the failure category, and whether relevant conditions were complete. Direct identifiers could be replaced with case references. The underlying information would remain with the institution responsible for it. Authorized officials would access identifiable records only when needed to resolve matters. Purpose limitation is equally important. Data accessed to restore an entitlement should not later be used for any other purpose.

How much time and effort could this save?

The potential savings begin before the formal grievance clock starts. In India, central government grievances resolved through the Centralized Public Grievance Redress and Monitoring System (CPGRAMS) took an average of 14 days in 2026. This figure, however, measures the period after a complaint is registered. It excludes the time a beneficiary spends when they notice a missing payment, find its probable cause, and locate the responsible institution. MSC’s research on PM KISAN, an income support program for farmers in India, found that recipients had to visit offices multiple times to resolve their grievances.

India’s standard operating procedure for DBT payments already provides daily status updates and detailed success or failure responses, including failure codes, to flow back to the implementing ministry. Once daily payment data are reconciled, the system could flag a possible missing payment and send it for review within one working day. It would, therefore, replace an uncertain period of discovery and institutional navigation with near-immediate administrative awareness.

It would not make every correction instantaneous. Notably, a bank account problem, a disputed land record, or an eligibility review may still require human action. The realistic benefit would be faster detection, more accurate routing, and fewer visits. The following table presents an illustrative India-based scenario that shows the potential effect of the intervention:

*Note: These scenarios are illustrative. The one-working-day period is a proposed service standard after the failure becomes visible in the reconciled payment date. The seven- and 14-day complaint periods are assumptions. The 14-day period is the FY25–FY26 CPGRAMS average disposal time for central government grievances.

Why should this matter to the government?

Early detection would also make program administration more efficient. An unresolved payment can generate repeated work across program teams, grievance officers, local offices, banks, and auditors. Identifying the problem at its source would reduce the need for complaint handling, case transfers, manual reconciliation, payment reprocessing, and follow-up. It would also allow officials to correct a common failure that affects a bank, payment batch, or district once, rather than address it separately for every recipient. This would free staff capacity for more complex cases and other delivery priorities. Early detection would also strengthen public trust by showing that the program can act on failures without waiting for people to pursue multiple institutions.

How can governments test this approach?

In many programs, particularly in low- and middle-income countries (LMICs), the first challenge will be to connect fragmented data and establish who can act on the resulting alert. A pilot should therefore test the underlying infrastructure and the proposed intervention.

1. Assess readiness and select a viable program

The pilot should map the journey from eligibility approval to final delivery. It should establish:

  • Whether the program has a reliable record of who was due to receive what and when;
  • Transaction-level payment outcomes;
  • Identifiers that can link the records;
  • Sufficiently frequent data updates.

The assessment should also identify the legal basis for data exchange and the institution authorized to resolve each type of failure. The first phase in the pilot should focus on building these foundations where they are absent.

2. Connect the data and test the reconciliation rules

The next step should be to create a controlled link between program and payment records, using secure data extracts or APIs where available. This step would begin with clear rules that compare expected and confirmed payments. AI-assisted diagnosis should be introduced only where standardized failure codes or sufficiently verified case histories are available. Some pilots may initially remain entirely rule-based until the data are sufficient to support AI analysis.

3. Run a human-led case-management pilot

The AI-based system should first run in shadow mode. During this phase, officials should verify flagged cases, assess the suggested cause, and route each case to the responsible institution. They should record every action, including whether the alert was correct, what information was required, and whether the responsible institution could resolve the case. Officials should also verify AI recommendations, investigate conflicting records, and authorize any action that affects eligibility or payment. A case should close only after delivery is confirmed or officials communicate a clear decision to the beneficiary.

4. Evaluate feasibility before scaling

The pilot should measure data completeness, the proportion of records successfully matched, alert accuracy, false positives, detection and resolution times, entitlements restored, and the effort required from officials and recipients. The evaluation should also estimate the cost to address infrastructure and coordination gaps. Scaling would make sense only where alerts are reliable, and the institutions that receive them can act.

AI is already finding its way into public programs, often through complaint analysis and monitoring. A less explored opportunity is to use AI for earlier case detection. Instead of waiting for a beneficiary to discover a missing benefit and navigate the grievance system, the administration could identify the failure first. This could flag a possible exclusion, rejected payment, or missing benefit before the affected person must navigate the grievance system.

Social protection programs should build on these early uses and test applications that improve day-to-day delivery, including early identification of missed or failed payments. This would allow administrative action to start sooner. The result will save time and resources for both governments and program recipients.

From banking agents to community service providers: The path to BC Sakhi sustainability

Posted on by Rashi Bhardwaj

The report explores how the BC Sakhi model can evolve from basic banking agents into sustainable, multi-service community providers. With most BC Sakhis earning below INR 5,000 monthly, it highlights gaps in product depth, revenue, and program management. It recommends expanding citizen, financial, digital, and livelihood services; deepening credit and financial intermediation; and building partnerships with public and private platforms. Stronger training, technology, monitoring, awareness, and institutional partnerships are identified as critical enablers for sustainable growth.

Playbook for Fintech-FSP Partnerships

Posted on by Rashi Bhardwaj

The playbook offers a practical roadmap for FinTech founders who seek partnerships with banks, NBFCs, MFIs, and other FSPs in India. It covers ecosystem mapping, partnership models, regulatory readiness, founder self-assessment, value proposition development, stakeholder engagement, trust, pilots, and scale. The central message is that successful partnerships require more than strong technology. Founders must demonstrate credibility, compliance, measurable business value, adaptability, and execution discipline. Small, focused pilots and evidence-based results help build trust and support progress toward long-term partnerships.

Who is responsible? Closing the accountability gap in financial fraud prevention

Posted on by harshita

Financial fraud is a systemic problem being addressed by individual solutions. Across most markets, the response to a scam victim is to report it to the bank, file a police complaint, submit evidence to a regulator, and hope. Meanwhile, the fraud operation that targeted that victim is already processing the next call, the next transfer, and the next cash-out.  

Fraud ecosystems have been deliberately engineered to move faster than the institutions designed to stop them. Banks and regulators need to move beyond better awareness campaigns or faster forms of grievance management to close this gap. It requires a fundamental redesign of accountability and understanding of who is responsible for what and when in the transaction chain.  

Fraud today is professionally organized, behaviorally engineered, technically enabled, and insufficiently counter-checked in most jurisdictions. In fast-payment ecosystems, safety must be built into design, supervision, and grievance resolution systems.  

The blind spots across the ecosystem  

The accountability gap is not located in a single institution. Rather, it runs across the entire ecosystem, where each actor’s blind spot enables the next fraud.  

Banks are positioned closest to the transaction, which should make them the most effective line of defense. Yet most fraud detection systems are calibrated to identify unauthorized transactions, such as credential theft, account takeover, and card skimming. Authorized push payment (APP) fraud is another type of fraud in a completely different category, where the victim is manipulated into initiating a payment themselves. APP bypasses most conventional controls. MSC’s Mind the Gap report found that more than 60% of fraud victims across India, Bangladesh, and Kenya did not know what grievance mechanisms existed. 48% of victims who attempted to report were dismissed for their inability to furnish evidence. 

Existing grievance resolution systems are largely designed around what the institution needs and not what the victim can provide. MSC’s consumer protection research in India has consistently documented this gap across the financial services lifecycle. We have traced the cycle from transparency of product terms at onboarding, to the accessibility of recourse channels post-harm. Our customer protection in the Indian digital financial services series mapped specific failure points related to recourse and transparency that leave customers without an effective remedy when things go wrong.  

In India specifically, the Prevention of Money Laundering Act (PMLA) creates a structural paralysis, as banks cannot freeze suspected mule accounts without authorization from the court or law enforcement. This creates a legally mandated delay that fraudsters systematically exploit. The IBA Working Group has proposed that banks be granted enhanced authority to place temporary holds on suspected mule accounts before formal orders arrive, which is a necessary regulatory design reform that remains pending.  

Fraud is also rampant in the telecom sector. Caller ID spoofing, SIM swaps, and the leasing of backend numbers to route fraudulent calls are all examples of vulnerabilities at the telecom layer. The UK’s Ofcom interventions show that these are solvable. Regulators can impose mandatory blocking of international calls that fake domestic numbers, block invalid caller IDs, and ban leasing backend numbers used to hijack calls. Although these are technical controls with measurable impact, most jurisdictions have not implemented them.  

Digital platforms, such as social media, messaging apps, and digital marketplaces, are another source of most scams. Yet, platform accountability for fraud from their infrastructure remains largely voluntary. Recorded Future’s 2024 Payment Fraud Report identified nearly 1,200 scam domains linked to fraudulent merchant accounts and nearly 11,000 e-commerce domains infected by Magecart skimmers. This is a threefold increase from 2023. Takedown times for fraudulent pages are measured in days, while scam operations are measured in hours.  

MSC’s Building trust through design report adds another perspective. Deceptive interface design, which comprises manipulative consent flows, hidden fees, and guilt-tripping prompts, erodes user agency and creates conditions for external fraud to thrive. Regulatory accountability must extend to the design layer, not only to obviously illegal content.  

Even where consumers know how to report, the system often fails them. MSC’s TRUST framework identifies the precise dimensions in which most grievance systems fail. Transparency suffers when consumers do not understand the terms they agreed to. Consumers lose recourse when complaint channels close, require multiple steps, or operate in a language they do not speak. They face information asymmetry at every point, which undermines their ability to understand what happened. Security investments fall behind. And once fraud occurs, timeliness matters most, yet procedural compliance systematically sacrifices it. 

What effective accountability looks like  

Three jurisdictions have moved furthest towards a systemic accountability model. Their approaches converge on the common principle that fraud prevention is a shared responsibility across the transaction chain, not a consumer obligation.  

Since October 2024, payment service providers in the UK must reimburse APP fraud victims more than USD 100,000, with costs split between sending and receiving banks. The one-year assessment showed that reimbursement alone is insufficient, as it simply compensates victims but fails to break the criminal business model. The more impactful interventions have been at the telecom layer, such as caller ID blocks, filters for invalid numbers, and requirements for operators to verify business customers. 

In Australia, the Scams Prevention Framework was passed in early 2025. It requires banks, telcos, and digital platforms to implement defined controls or bear liability for losses, with penalties of more than USD 35 million or 30% of turnover during the breach period. It includes digital platforms as designated entities and becomes the first framework to formally extend accountability to the channels where scams originate.  

Singapore’s Shared Responsibility Framework, effective December 2024, allocates liability in a defined sequence. In it, liability falls first on the financial institution, then on the telecom operator, and finally on the consumer. This can occur only if both institutions have fulfilled their obligations. This waterfall model establishes clear, predictable accountability for each actor and upholds the principle that consumers should not bear losses when institutions have failed in their duties.  

What behaviorally informed prevention actually requires  

Accountability frameworks set the incentive structure. But institutions and regulators must make different design decisions to build the actual prevention architecture. 

  • Contextual friction, not generic warnings: Transaction-level interventions, such as cooling-off periods for high-risk payments, purpose prompts, and real-time behavioral flags, are demonstrably more effective than warning messages delivered at onboarding. India’s NPCI removed P2P UPI collect requests, a concrete example of friction designed into the system.  
  • Cross-institutional intelligence sharing: Mule account registries, cross-bank fraud signals, and real-time data sharing between banks, telcos, and platforms turn individual detection into network-level prevention. India’s MuleHunter.AI at the RBI Innovation Hub is an early example of this direction.  
  • SupTech and RegTech investment: MSC’s SupTech data maturity framework finds that one-third of regulators still rely on manual submissions, and fewer than 10% of smaller jurisdictions have full data standardization. The limiting factor is data quality. Many authorities validate data manually. Regulators cannot supervise what they cannot measure. Fraud cannot be detected if reporting systems lag transactions by weeks.  
  • Capability-building as a regulatory obligation: Financial literacy must shift from a CSR activity to a mandatory, measurable output. To address this, MSC’s PTE Framework offers a practical model for how phygital capability delivery driven by teachable moments can be designed, contextualized, and evaluated across diverse user segments. 
  • Victim-centered grievance design: Regulators can use the MSC TRUST framework to redesign grievance resolution systems to build solutions from the victim’s perspective. These should be accessible in local languages, multi-channel, credible, and be able to initiate protective holds without a court order.  

The way forward 

Fraud today is institutionally tolerated due to design fragmentation: Banks are responsible for transactions, telcos are responsible for calls, platforms are responsible for content, and regulators are responsible for their own sectors. In the space between those silos, fraud operations run freely.  

The evidence from the UK, Australia, and Singapore, and from MSC’s field research across Asia, Africa, and the Pacific, proves that institutions must design protection into the transaction rather than just bolt it on. It also requires a grievance management system built for the person who has just been deceived, not for the institution that manages its liability. 

The fraud supply chain is end-to-end, and the protection system that counters it must reflect this reality. 

This is Blog 3 of a three-part MSC series on fraud supply chains. Blog 1 examined why ordinary people fall for fraud. Blog 2 examined how fraud operations are industrialized and monetized.  

 

Climate-resilient affordable housing finance: From affordable at origination to affordable over time

Posted on by Rashi Bhardwaj

Today’s changing climate alters what affordability means for borrowers, homes, and housing finance portfolios. Affordability can no longer be assessed solely by whether a borrower can afford the equated monthly installment (EMI) during origination. It must also account for the household’s ability to continue to occupy, maintain, repair, and repay the home over the life of the loan.

India’s affordable housing challenge is often assessed by the scale of housing demand addressed through homes sanctioned, constructed, or financed. While these metrics remain important, they do not fully capture whether a home will remain safe, livable, and affordable over the loan period. A home that appears affordable at origination may become increasingly expensive if the household faces higher cooling and water costs, climate-related health expenses, or repeated repairs after floods, cyclones, or extreme heat. Affordability should therefore cover the life of the home and the loan.

The urgency of this issue is evident. More than 80% of India’s population lives in districts at risk of climate-induced disasters. Rising temperatures, changing rainfall patterns, groundwater depletion, intense cyclones, and sea-level rise can affect homes and livelihoods. This creates a dual risk for affordable housing borrowers. A climate event can damage the financed property and disrupt the household’s income. This interaction between asset damage and livelihood interruption is vital for lenders that serve households with limited financial buffers.

The risks are particularly significant for economically weaker sections (EWS), lower-income groups (LIGs), and households with informal or variable incomes. Government housing programs provide important financial support, but they may not always fully cover construction costs. This can potentially leave beneficiaries with a residual financing requirement. Pradhan Mantri Awaas Yojana-Urban (PMAY-U), Pradhan Mantri Awaas Yojana-Gramin (PMAY-G), and various state-level housing schemes provide subsidies and other forms of assistance. These schemes create a potential role for complementary lending, including top-up and stage-based construction finance. However, secondary evidence suggests that access to such formal credit remains relatively limited for many beneficiaries. For instance, as noted in the Standing Committee on Housing and Urban Affairs’ analysis of PMAY-U, the average cost of an EWS house was estimated at approximately INR 0.65 million (USD 6,850) as of 2022. The center, the states, the urban local bodies (ULBs), and the beneficiary were expected to share the cost. While the central contribution is fixed on a per-unit basis, the contributions from the states and ULBs help keep the overall cost affordable for the beneficiary. The Committee observed that variations in the extent of state-level contributions could, in some cases, result in beneficiaries being required to make higher contributions. The average beneficiary contribution was estimated at around 60%, highlighting the potential need for additional financing. In this context, financial institutions can help bridge the residual financing gap through appropriately structured top-up and stage-based construction finance, particularly for households with limited access to conventional housing finance.

The International Finance Corporation (IFC) and Aavas Financiers’ market research on green affordable housing finance also found that 62% of loans for new housing from affordable housing finance companies fund This profile of small, self-built, incrementally expanded housing has important implications for climate resilience. Such homes often fall outside the reach of green-building certification and developer-led green mortgages, yet they are among the most exposed to extreme heat, flooding, and structural vulnerability. Resilience cannot arrive in this segment as a ready-made product. Households must build it feature by feature through stronger roofs, raised plinths, and better drainage during construction or improvement. Climate-responsive finance must therefore reach the household that adds a room, replaces an unsafe roof, or builds gradually on an owned plot.

The research suggests that the intersection of climate and affordable housing finance should extend beyond green mortgages for developer-built projects. Much of the demand involves self-construction, renovation, reconstruction, vertical expansion, and completion of houses stalled by funding gaps.

The relevant interventions need not always involve complex technology. Depending on local conditions, they may include cool roofs, improved shading and ventilation, rainwater harvesting, water-efficient fixtures, raised plinths, stronger drainage, flood-safe electrical systems, and improved roof anchoring. However, affordability remains a challenge. A National Housing Bank (NHB)- study estimated that green features increase residential construction costs by approximately 3.6%. Although these measures may lower utility and repair costs over time, the additional upfront expense can strain lower-income households.

The IFC and Aavas Financiers’ primary research in Jaipur and Indore reinforces this finding. Households in the affordable segment showed broad willingness to adopt green features, but around 60% cited higher upfront costs as the main barrier. Lower-income respondents would, on average, need the next tier of green features at roughly 30% lower cost before they would opt in. Awareness of green home loan products stood at just 7%, and most respondents said they would take one only if it offered interest-rate concessions or fee waivers.

Addressing these risks requires ain how affordability is defined from affordability based on loan amount to affordability based on design. A house is not truly affordable merely because its EMI fits the borrower’s present income. It must also remain affordable to occupy, maintain, and repair. Construction rarely relied on one source of finance. Households relied on a mix of savings, chit funds, gold, personal loans, informal borrowing, and housing finance. A climate event that reduces income or creates an unexpected repair, water, electricity, or healthcare expense can disturb this already delicate balance.

For lenders, the risk extends beyond household cash flow. A flood can damage the mortgaged property, disrupt a small business, and destroy household assets all at once. A heatwave can reduce working hours for outdoor workers while increasing cooling and healthcare expenses. The resulting chain is clear: 

This distinction is relevant in a market where smaller loans are not necessarily simpler loans. India’s housing loan market stood at approximately USD 410 billion in portfolio outstanding as of FY25. It grew at around 13% annually from FY20 to FY25, with nonbanking financial companies (NBFCs) and housing finance companies (HFCs) holding roughly one-fifth of the market.

In this scenario, the affordable segment has a distinct risk profile. CRISIL estimates show that gross non-performing assets (NPAs) for affordable housing loans stand at approximately 2.6% in FY25, roughly twice the 1.2% recorded for overall housing loans. This occurred even as asset quality across retail lending improved.

This gap does not establish climate change as a cause of delinquency. It does, however, indicate that affordable-segment borrowers already operate with thinner margins for error. These borrowers are predominantly self-employed or informally salaried households that manage phased construction, mixed financing, and limited financial buffers. Climate-related shocks could intensify these underlying pressures.

The response must begin with an understanding of the local market. Affordable housing markets already differ across districts in terms of income patterns, construction practices, land documentation, housing aspirations, and repayment capacity. Climate risk introduces another layer of variation. A heat-prone district may require a different housing and finance package from a flood-exposed coastal area or a water-stressed inland market. A single national green home loan is unlikely to address these differences.

Product, underwriting, and servicing models must therefore evolve together. Construction-linked loans can include stage-based disbursement, modest top-up flexibility, and simple technical guidance. Property appraisal can consider flood history, drainage, heat exposure, water availability, and structural vulnerability alongside title, valuation, and loan-to-value ratios. The same site visits used to verify loan utilization can check whether essential resilience features are in place. Lenders can also establish predefined post-disaster protocols, including early customer communication, repair finance, and temporary repayment flexibility.

Public policy provides an entry point for this transition. Pradhan Mantri Awas Yojana-Urban (PMAY-U) 2.0 advises states and union territories to raise awareness of innovative construction technologies and materials that enhance thermal comfort, energy efficiency, disaster resilience, and cost-effectiveness. This initiative creates an opportunity to integrate simple climate-resilience checks into beneficiary-led construction, housing finance, and construction-stage monitoring.

Affordable housing finance must ultimately protect three things: the borrower’s repayment capacity, the home’s physical resilience, and the long-term quality of the lender’s portfolio. India still has an opportunity to embed climate considerations into millions of homes that are yet to be built or improved. Building climate resilience into affordable housing finance now will prove more affordable and equitable than financing homes without adequate resilience and paying for a retrofit later, after households and lenders have already absorbed the cost.