A bank account provides access to the financial system. It does not automatically provide access to credit. A lender may still lack the complete and current information needed to assess a customer fairly.

At Snappi, our mission is to close that information gap: to make creditworthy customers more visible without compromising affordability or sound risk management. Everyone deserves a fair opportunity to demonstrate creditworthiness. But creditworthiness also carries responsibility: credit must be affordable, and a borrower must be willing to repay.

1. The challenge: financial lives have changed faster than credit files

World Bank data show that 88.6% of Greek adults had an account in 2024. That is important progress, but account ownership is not the same as access to credit. An account connects someone to the financial system; it does not necessarily tell a lender whether income is stable, whether a proposed instalment is affordable or how likely the customer is to repay.

Responsible access to affordable credit matters. It helps households manage temporary shocks and invest in their future. It gives viable entrepreneurs the capital to grow, increase productivity and create jobs.

The problem is that society has changed faster than many credit models.

Traditional underwriting still tends to imagine a familiar borrower: one country, one employer, one monthly salary and one primary bank. That borrower remains important, but no longer represents everyone.

• Income may now come from employment, freelance work, a small business, property rental activity or a digital platform—and it may arrive irregularly or through several accounts. In 2025, self-employed people represented 24.8% of employment in Greece, compared with 14.2% across the EU. This does not make them inherently riskier; it makes salary-centred assumptions less suitable for an unusually large part of the Greek workforce.
• Families and careers have become more international. People study in one country, work in another, marry across nationalities and later return home. They may bring income, skills and assets, yet have little domestic borrowing history — often referred to as “thin-file” customers.
• Financial relationships have become more fragmented. Customers use several bank accounts, payment cards and digital wallets, sometimes across providers and jurisdictions. A single lender may therefore see only one fragment of the customer’s financial life.

A thin-file applicant is not necessarily unbankable, low-income or high-risk. It simply means that the lender has too little familiar information to reach a confident decision. The problem may not be high risk. It may be low visibility.

2. Why does this information gap matter: incomplete information can become exclusion or mispricing

When credit models are built mainly around conventional employment and established borrowing histories, unfamiliar or incomplete information may be interpreted as higher risk. The customer can face two consequences: rejection, or an offer with a higher interest rate, lower limit or more restrictive terms than the actual risk justifies.

This is not only a problem for low-income borrowers. I am a case in point. I have an established international career in finance and a PhD, yet when I moved to Greece to take up my role, I found I had little ability to borrow — I still have no domestic borrowing history recorded in Tiresias. Many people who studied or worked abroad and later returned to Greece face the same gap.

Models that reward a long repayment record naturally favour people who have already had access to credit. Those without such a record may be declined or charged more; yet without access, they cannot build the history the model requires. Structural exclusion can therefore arise without discriminatory intent. It is a classic Catch-22.

The scale of this information gap is difficult to quantify. The Bank of Greece’s Central Credit Register is an important step towards broader and more transparent credit information, including borrower access and correction rights. But no authoritative public statistic shows how many Greek applicants have the depth of positive credit history required by conventional lending models—or how many are declined specifically because their file is thin.

The necessary caveat: inclusion cannot mean inappropriate lending

Financial exclusion is not one problem with one solution. Some applicants cannot safely afford additional debt. Others have sufficient repayment capacity but are difficult to assess because conventional data reveal only part of their financial position. Treating these groups in the same way creates two errors: lending where the debt is unaffordable, and rejecting viable borrowers simply because their capacity is not sufficiently visible.

This distinction matters particularly in Greece. In 2025, 27.5% of the population was at risk of poverty or social exclusion, while 14.9% experienced material and social deprivation. In 2025, more than half of Greeks could not afford an unexpected expense (such as unexpected doctor fees), compared with 29.2% across the EU. Access to responsible, affordable credit matters. Wider access must never mean placing more debt into financially fragile households. Affordability must remain an independent and non-negotiable gate, however sophisticated the model becomes.

Greece also carries the memory of mispriced credit risk. After extreme hardship, following the crisis, the banking sector’s non-performing-loan ratio fell to 3.3% at the end of 2025—its lowest level since Greece joined the euro area—although substantial private debt remains outside bank balance sheets. That hard-won progress must be protected.

The answer is not greater risk appetite. It is better risk resolution: distinguishing among affordability, expected repayment performance, fraud and insufficient information. Responsible inclusion means making more accurate ‘yes’, ‘no’ and ‘not yet’ decisions—rejecting fewer good borrowers for the wrong reasons, while approving fewer unaffordable loans in the name of access.

3. How we are solving it at Snappi: better information, better judgement

At Snappi, we have taken on this challenge: expanding responsible access to credit for customers whom traditional models struggle to assess.

Over the past years, my leadership team and I have repeatedly returned to the same questions. “What information is genuinely relevant? How can customers share it lawfully and securely? What benefit should they receive in return? How do we assess affordability, predict repayment and prevent fraud without creating unnecessary barriers for legitimate applicants?”

Over the past decade, new legislation and advances in technology have enabled banks to answer these questions with far greater accuracy. Together, they allow lenders to understand customers better, make more informed decisions and serve them more effectively.

Why can we have a different approach to customers? What has changed?

A. The smartphone has become a financial control centre

Processing power matters, but the more important change is what our devices now reveal about financial life. Salary payments, electricity bills, supermarket purchases, subscriptions and transfers create an evolving cash-flow picture. With informed consent, these data can show income regularity, essential expenditure, recurring obligations, available buffers and recovery after a difficult month. For an irregular earner, twelve months of actual cash flow may be more informative than one payslip.

This power requires restraint. Data must be collected for a defined purpose, protected carefully and never treated as an automatic verdict. Used responsibly, the device in our hand can help replace assumptions with evidence—and make previously invisible customers easier to understand.

B. Open banking lets customers bring their evidence with them

PSD2 is the European framework that made open banking possible. In plain language, it allows a customer to instruct one bank to share selected payment-account information securely with another regulated provider. Sharing is not automatic: it requires specific permission, access is limited to the agreed purpose and the customer does not hand over bank passwords.

PSD3 and the accompanying Payment Services Regulation are intended to make this framework work more consistently across Europe, while strengthening fraud prevention, data security and consumer protection.

For lending, customer-authorised transaction data can complement Tiresias and a bank’s own records. It can reveal the pattern and stability of income, essential expenditure, recurring commitments, cash buffers and periods of financial stress. It does not replace affordability checks or credit judgement. It makes them more current and complete.

C. AI can reveal patterns—but it cannot remove responsibility

AI can bring together thousands of transactions and convert them into understandable indicators. It can help distinguish seasonal volatility from sustained financial weakness, measure cash buffers and identify unusual patterns that may indicate fraud. The purpose is not to remove human judgement, but to give decision-makers a clearer picture.

In 2025, the European Banking Authority reported that 92% of EU banks were deploying AI, while the remaining 8% were piloting or discussing applications. Creditworthiness assessment was among the use cases observed. This shows how widespread AI has become across banking; it does not mean that every bank is using AI in credit scoring.

International evidence is promising, but it cannot be transferred mechanically to Greece. A 2026 peer-reviewed study of a major US fintech platform found that alternative-data underwriting approved 15% to 30% of low-credit-score applicants whom traditional models would have rejected, often at lower rates. The greatest gains were among ‘invisible primes’: customers whose conventional files looked weak or revealed too little, despite low underlying default risk.

For Snappi, this is not a forecast. It is a hypothesis to test. Greek income patterns, customer behaviour, data coverage and economic conditions are different. Our models must be developed and validated using relevant local evidence, continuously monitored and kept behind independent affordability and fraud controls.

The EU AI Act generally classifies AI used to assess the creditworthiness or credit score of individuals as high-risk, subject to its conditions and exceptions. Under the July 2026 Digital Omnibus on AI, these high-risk obligations for standalone systems apply from 2 December 2027. At Snappi, accountable AI means data lineage, feature governance, independent validation, drift and outcome monitoring, reason codes and meaningful human oversight. The EBA Guidelines on loan origination and monitoring remain the prudential and consumer-protection foundation. AI should improve judgement, not conceal it.

D. Fraud no longer looks like one suspicious transaction

With our partners, Snappi is building fraud-detection capabilities that go well beyond traditional rules. Fixed thresholds and failed-login alerts remain useful, but fraudsters increasingly change devices, credentials, locations and identities to avoid them.

Modern systems can analyse hundreds of signals in near real time: whether a device is associated with suspicious activity, whether one identity is linked to several accounts, whether a location is inconsistent with normal behaviour, or whether a session appears to involve a bot, emulator or concealed network. Behavioural intelligence can also examine how someone types, handles a phone or navigates an application—helping to identify when legitimate credentials may be controlled by someone else.

LexisNexis reports that its Digital Identity Network processes approximately three billion monthly transactions and recognises around 1.4 billion digital identities. This can expose connections that an individual bank could not see alone: synthetic identities, networks of money-mule accounts, returning fraudsters using changed credentials, account takeover and cases where a genuine customer may be manipulated into authorising a scam payment.

Our objective is not simply to block more activity. It is to identify risk more precisely, intervene when necessary and reduce unnecessary friction for legitimate customers.

Our approach to lending: Build – Test – Scale

At Snappi, we have adopted a simple principle: learn faster than we lend. Our first engine relied mainly on traditional credit-bureau information, but our early experience taught us that static scores and rigid rules are not enough. Tightening rules may reduce risk, but it can also exclude too many people. Transaction activity, extended data, customer engagement and the maturity of the relationship can provide a more current picture. Rigid rules initially help to contain risk, but they can also push approvals too low; behavioural evidence offers a way to restore the balance between prudent growth and protection.

1. Build: create a better customer picture

We combine Tiresias information with verified eGov income, Snappi transaction activity, a short customer questionnaire and, progressively, PSD2/Open Banking data. The purpose is not to collect more data for its own sake. It is to answer three separate questions: Is this genuinely the customer? Can the customer afford the loan? Is the customer likely to repay?

Each is an independent gate. A strong credit score cannot compensate for an unaffordable loan or a serious fraud concern.

2. Test: start small and learn from behaviour

We begin with manageable limits and observe actual behaviour: whether customers repay on time, maintain sufficient cash flow and engage responsibly with the bank. Our experience shows that repeat borrowers perform better and that timely, clear communication improves repayment. Engagement is not merely a marketing measure; it can provide meaningful evidence of financial behaviour. Every repayment teaches the model something new. Lending therefore becomes a continuous learning process, not a one-off judgement made at application.

We also test our models for accuracy, stability, fairness and unintended consequences. New data or models must prove their value before they are scaled. This requires continuous improvement, not a scorecard frozen in time: rules should evolve with fresh evidence.

3. Scale: reward evidence and responsible behaviour

A thin credit file should not automatically result in permanent rejection; it should determine the starting point. Customers who demonstrate reliable repayment, financial discipline and sustained engagement can gradually access higher limits and broader products. Exposure can also be reduced if affordability weakens or risk increases. Trust is earned gradually: start small, observe responsible behaviour and expand access only when the evidence supports it.

We intend to make this progression visible through responsible gamification. Customers should understand how on-time repayment and sound financial behaviour help them build credibility and improve future access. But gamification must reward financial discipline—not the volume or frequency of borrowing. We do not want it to become a softer term for encouraging more debt.

4. The benefit for society: fairer access and stronger credit outcomes

We want the work we do to leave society better than we found it.

The first benefit is fairer economic participation. A viable freelancer, returning professional, seasonal worker or young applicant should not be excluded merely because their financial life does not resemble that of a traditional salaried borrower. Better information can reduce false rejections and allow limits and pricing to reflect actual risk more accurately.

The second is protection against over-indebtedness. A fuller cash-flow view does not only identify customers who may safely receive credit; it also identifies when a proposed loan is unaffordable or when a customer’s position is beginning to weaken. Better inclusion and better portfolio quality can reinforce each other.

The third is a more competitive, digital, mobile and cross-border financial system. Customer-controlled evidence can help people carry their financial credibility between banks and, eventually, across borders. This matters in a Europe where people and capital already move freely but positive retail financial evidence does not travel nearly as easily.

Shared value creation

Technology alone cannot deliver all of this. Access also depends on product design, risk-sharing and public policy. Lending is a shared responsibility.

Responsibilities, benefits and risks are all shared. Lenders must design affordable products, explain decisions and support customers in difficulty. Regulators must enable secure data sharing and competition while enforcing consumer protection. Customers must provide accurate information and repay when able, with a credible path back after genuine hardship. Trust must be earned on both sides.

Conclusion

The future of lending should not be defined by a choice between wider access and risk management. With better information, banks can do both. Cash-flow data, open banking and accountable AI can reveal the reality behind a thin credit file, helping lenders distinguish between customers who cannot safely afford debt and those who are difficult to assess through traditional methods.

But technology is only part of the answer. Every decision must remain grounded in affordability, fraud prevention and transparency. Models must be tested, customers must understand how their information is used, and trust must be earned gradually. A responsible lender should be able to say “yes”, “no” or “not yet”—and provide a path forward.

For Greece, this is an opportunity to protect the hard-won improvement in credit quality while building a financial system that reflects how people live and work today. Freelancers, returning professionals, young people and internationally mobile families should not be invisible merely because their lives do not fit an old template.

Our objective is not easier lending or greater risk appetite. It is better underwriting: moving beyond judging the past alone, understanding the present clearly, and helping creditworthy customers build a stronger financial future.