Pakistan’s banking industry has become very good at digitising what customers can see. Mobile apps are cleaner, payments are instantaneous, account opening takes minutes, and QR codes are increasingly replacing cash. Behind the screen, however, the experience can be rather different.
As some quip, many banks have built a digital front door, but once the customer walks through it, someone inside still reaches for a file, a stamp and a photocopier. This is because when an application begins online, it still has to pass through a series of manual verifications, disconnected databases, emails, spreadsheets and committees. The customer may see an app, but the bank is often still operating through processes designed for another era, and these manual processes often undermine efficiency and stall transactions. That is why a radical change is necessitated in how we design processes.
For most of its history, banking has been built around documents. Salary slips, financial statements, tax returns, property records, collateral documents and credit histories have helped banks establish whether a customer exists, earns an income, owns assets and can repay a loan. This model works well when economic activity is formal and comprehensively recorded. It is far less effective in markets where businesses are active, productive and profitable, but not extensively documented.
By leveraging AI and moving past paper trails, banks can reshape processes
Pakistan illustrates the problem clearly. Millions of farmers, traders, retailers, freelancers, transporters and small businesses generate income and employment. Yet much of their activity remains outside conventional financial records. Nearly 85 per cent of Pakistan’s micro, small and medium-sized enterprises operate informally. The traditional banking model often interprets this absence of documentation as an absence of creditworthiness.
That distinction is important. A small retailer may not have audited financial statements. Still, the business may have years of supplier relationships, recurring utility payments, consistent inventory purchases and predictable transaction flows. A farmer may not have a conventional salary, but land records, cropping patterns, input purchases, weather data and previous repayment behaviour may reveal a great deal about the farm’s economics. The customer may be document-poor but not necessarily data poor.
The real opportunity for banks is to convert these scattered signals into responsible credit decisions. Every economically active individual leaves behind information. Payments, purchases, mobile usage, utility bills, tax activity, land ownership, merchant relationships, government registrations and repayment patterns all provide partial evidence of financial behaviour. When combined carefully, these signals can help establish identity, estimate income, understand cash flows and predict repayment capacity.
This creates the possibility of moving from conventional underwriting, which relies on a static set of documents collected at one point in time, towards continuous underwriting. Credit limits can be reviewed as transaction volumes change. Repayment behaviour can influence future pricing. Early warning indicators can identify stress before an instalment is missed. A customer who initially qualifies for a small facility can gradually gain access to larger financing as a reliable financial record develops.
This is more useful than simply replacing a paper form with an electronic one. It turns lending from a one-time approval exercise into an ongoing relationship based on observable behaviour. Artificial intelligence can improve this process by identifying patterns across large and complex datasets. But AI should not be treated as a substitute for credit judgment. Its value lies in helping banks process information consistently, detect anomalies and improve the quality of decisions. The strongest model is likely to combine machine-based assessment with human review for exceptions, unusual cases and larger exposures.
Digital infrastructure can also change the structure of credit itself. Traditional lending often gives the borrower a sum of money and relies on subsequent monitoring to determine how it was used. Digital finance allows banks to design purpose-linked credit in which funds can be used only within an approved ecosystem. Agricultural financing, for example, can be disbursed through a card or digital instrument that is accepted by authorised seed, fertiliser, pesticide and equipment suppliers. The bank can see where the financing is being used, merchants receive payment directly, and the borrower retains flexibility within the approved purpose.
Purpose-linked disbursement can also reduce diversion of funds, improve traceability and create better information about the underlying economic activity. Over time, this information can improve future credit decisions and allow financing to be tailored more closely to the needs of each borrower. The same approach can support housing, education, healthcare, livestock, small-business inventory and clean-energy financing.
One reason digital transformation remains expensive is that banks often develop each product as a separate project. A new agricultural scheme receives one system. A small-business programme receives another. Housing finance is managed through a third. Each requires its own integrations, verification processes, dashboards, controls and reporting arrangements. This creates duplication and makes innovation slow. A more effective approach is to build shared institutional capabilities.
The principle is similar to modern software architecture. Banks should not rebuild the entire institution every time they launch a product. They should have modular capabilities that can be configured for different customer groups, sectors and risk profiles. Once these rails are in place, the cost and time required to develop a new product falls significantly. The bank can also apply consistent controls across programmes instead of creating a separate operational framework for each one.
The effort should increasingly be on the creation of this common infrastructure rather than a collection of isolated digital products. Integrations with Nadra, land-record systems, government platforms, payment networks and other databases allow information to move across the lending process without repeated manual intervention. This is particularly important in large public-sector and financial-inclusion programmes where a bank may need to process hundreds of thousands of applications within a relatively short period.
The next generation of lending may also take place less frequently inside branches and more often within the economic activities customers already perform. A farmer may access financing through an agricultural platform. A retailer may receive working capital through a merchant-payment system. A freelancer may be offered credit through an export or remittance platform. This is the broader promise of embedded finance: banking becomes part of the customer’s commercial journey rather than a separate activity that requires a visit to an institution.
The institutions that succeed will not necessarily be those with the most attractive applications or the largest number of digital transactions. They will be those that can combine fragmented information, institutional judgment and technology to understand customers who were previously difficult to see. For decades, banks asked customers to produce documents before they could be trusted. The emerging model allows trust to be built gradually through behaviour, transactions and performance. That shift has the potential to reshape banking.
The writer is the chief of staff & strategy at the Bank of Punjab
Published in Dawn, The Business and Finance Weekly, August 3rd, 2026
