Rolling out digital banking without addressing algorithmic bias risks excluding the very populations it promises to uplift. Meghna Chotaliya uncovers a financial system that is technologically advanced and socially regressive at the same time and explains why the stakes are particularly large for emerging economies that are stressing rapid digital transformation while still dealing with weak institutions, large informal sectors and poor connectivity.
Artificial intelligence and digital banking are quietly rewriting the rules of who gets access to money.
Across the world, governments and financial institutions have embraced digital banking as a solution to one of the oldest problems in development economics: how to bring the unbanked into the formal financial system. Mobile apps, AI-driven credit scoring and automated lending platforms have all been held up as the answer. The results in some areas have been real. Millions of people in emerging economies now have access to digital payments and savings tools that simply didn’t exist for their parents’ generation.
During COVID-19, digital transfer systems kept money moving when physical infrastructure couldn’t. Kenya’s M-Pesa, the mobile money service launched by Safaricom in 2007, remains the emblematic case. Research published in Science estimates that access to the service lifted around two percent of Kenyan households out of extreme poverty. More recently, Brazil’s central bank built Pix, an instant payment system that reached roughly 80 per cent of Brazilian adults within three years of its 2020 launch and is now studied as a model across Latin America. Yet the same World Bank data show that 1.3 billion adults worldwide still have no account.
But there is a tension sitting underneath all of this that doesn’t get nearly enough attention.
These systems don’t run on goodwill but on data. AI credit assessments, risk profiles and lending decisions all depend on having a legible digital footprint such as regular employment patterns, predictable transaction activity and established digital financial records. If you have those things, algorithmic banking largely works in your favour. If you don’t, if you work informally, live somewhere with patchy internet, or have simply never had reason to build up a formal financial record, you don’t get treated unfairly by these systems so much as rendered invisible to them. Which, from a practical standpoint, is often worse.
This matters most in places that can least afford it. Rural populations, elderly people, informal workers and women with interrupted financial histories are exactly the groups whose real-world circumstances produce the kind of thin, irregular data that algorithmic systems don’t know what to do with. The old barrier to financial access was geographical: no bank branch nearby. The new one is informational: no usable data profile. It’s a quieter form of exclusion, and in some ways harder to fight.
There is also a deeper problem with how we talk about algorithms. The term itself suggests neutrality: a system built on data appears more objective than decisions shaped by individual judgement. But algorithms are only as neutral as the data they’re trained on, and financial data is not neutral. It reflects decades, sometimes centuries, of structural inequality in who got credit, who got hired and who built wealth. A system that learns from that history will tend to reproduce it, even without any intent. This isn’t a theoretical concern. International evidence increasingly shows that AI lending tools have produced discriminatory outcomes in multiple contexts. A study of America’s mortgage market found that algorithmic lenders charged black and Latino borrowers more for credit than otherwise identical white borrowers. And when Apple launched its credit card in 2019, complaints that women were being offered far smaller credit limits than their husbands prompted a formal investigation by New York’s financial regulator into the algorithm behind it. If this can happen in one of the world’s most heavily supervised financial markets, it is worth asking what is happening in markets with thinner supervision.
The technology is not the only culprit, governance is
Countries with strong regulatory institutions, clear accountability frameworks and populations with high digital literacy are much better positioned to catch and correct these problems. Where regulation is weak or vague, algorithmic systems make consequential decisions about loans, credit limits and financial eligibility but with limited transparency and almost no meaningful recourse for the people affected. The fintech industry moves fast. But almost everywhere oversight structures are still catching up.
Kenya is instructive here, too. But this time as a cautionary tale. The same market that produced M-Pesa later saw a proliferation of largely unregulated digital lending apps, some charging annualised interest rates running into the hundreds of per cent and mining borrowers’ phone contacts to shame defaulters. It took until December 2021 for Kenya’s parliament to pass legislation bringing digital lenders under the supervision of the Central Bank of Kenya. India followed a similar arc: after its own wave of predatory loan apps, the Reserve Bank of India issued digital lending guidelines in 2022 centred on transparency, data protection and grievance redress. The European Union, by contrast, has moved pre-emptively: its AI Act classifies credit-scoring algorithms as “high-risk” systems subject to data governance, transparency and human oversight requirements. That is a level of regulatory capacity – and bargaining power over global technology firms – that many emerging economies cannot yet match.
For policymakers, the implication is uncomfortable. Digital finance has been sold as a development tool, so it is primarily treated as a technology and innovation agenda. But it is also, and maybe more fundamentally, a governance challenge. Rolling out digital banking infrastructure without addressing algorithmic transparency, digital exclusion risks and the specific needs of marginalised groups isn’t financial inclusion. It is a form of financial modernisation that continues to exclude the same populations, albeit through more technologically sophisticated mechanisms.
The stakes are particularly large for emerging economies that are simultaneously stressing rapid digital transformation and still dealing with weak institutions, large informal sectors, and a lack of good connectivity. The risk isn’t hypothetical: you can end up with a financial system that is genuinely technologically advanced and genuinely socially regressive at the same time.
The promise of digital finance was never really about the technology itself. It was about what the technology could do for people. Whether that promise gets kept depends far less on the sophistication of the algorithms than on the institutional commitments, regulatory choices and political will that surround them.
This article gives the views of the author, not the position of LSE Business Review or the London School of Economics. You are agreeing with our comment policy when you leave a comment.
Image credit: crocoman provided by Shutterstock.