By Boris Bialek, VP and Global Field of CTO Industries, MongoDB
Traditional credit scoring models are struggling to reflect a modern, diverse economy. In the UK alone, 5.6 million people are classified as “credit invisible.” This group includes students, migrants, gig workers, and others who can regularly meet their financial obligations but remain shut out of access to credit because they do not engage with conventional credit products like loans or credit cards.
The problem lies in how creditworthiness is measured. Legacy models do not account for present-day financial practices, and instead rely on historical borrowing behaviours. As a result, people who pay their rent and bills consistently may still face rejections from lenders.
Luckily, a fairer financial system is appearing on the horizon, and the industry is starting to broaden its understanding of what constitutes financial reliability. Alternative data, coupled with rapid advances in AI, provides that opportunity.
A new definition of creditworthiness
As financial behaviour changes, so must the indicators of stability. Rental payments, utility bills, and mobile subscriptions are all reliable indicators of someone’s financial responsibility. These payments are regular, essential, and often prioritised by consumers. Including these data points would bring millions into the formal credit economy without lowering standards or introducing risk.
Additionally, it presents a unique opportunity for companies in this sector. By embracing these alternative, innovative models, lenders can tap into and serve new customer groups, reaching millions of potential customers previously left behind. In doing so, they not only expand their customer base but also gain a competitive advantage in an increasingly crowded market.
In many ways, this is a necessity; the UK labour market has changed dramatically, and full-time, salaried work is no longer the dominant model. Today, many people earn income through gig platforms, freelance work, or multiple part-time roles. These workers often lack pay slips or consistent monthly income, which legacy systems see as red flags. But irregular income is not the same as instability. A freelancer with multiple clients and steady earnings is financially viable. Additionally, the punctual payment of utility bills and subscriptions can also reinforce an individual’s financial stability by signaling to lenders that the individual has a steady cash flow and is on top of their essential financial commitments.
The issue is that current models can’t recognise that, and credit scoring must adapt. Assessing real-time cash flow, savings habits, and spending patterns is a more reliable and fair way to understand financial health.
The potential of AI
AI is key to making this shift a reality. With it, lenders can analyse complex data sets, detect behavioural patterns, and build dynamic risk profiles that paint a more accurate picture of an applicant’s financial wellbeing – and their creditworthiness.
This is especially promising with the advent of generative AI. Today’s models can simulate realistic financial behavior patterns for better credit assessments, which addresses one of the biggest limitations of traditional credit scoring: the reliance on historical credit data. By creating synthetic data that mirrors actual real-world financial behaviours, Gen AI models enable a more inclusive assessment of creditworthiness. For consumers, it means access to mortgages, loans and other financial products that may have been inaccessible previously.
Admittedly, financial institutions must consider and mitigate against the risk of hallucination. If the model presents information that is either nonsensical or outright false, it can have very serious consequences for both the consumer and the lender. Although regulators are still sketching out exactly how to approach such use cases, there are several techniques to mitigate this risk. For example, using Retrieval Augment Generation (RAG) frameworks. RAG minimises hallucinations by grounding the model’s responses in factual information from up-to-date sources, ensuring the model’s responses reflect the most current and accurate information available.
Inclusion is good for business
Outside of the clear moral value of facilitating financial inclusion, the reality is that this is an important growth strategy. People previously excluded from credit represent a new customer base. They need mortgages, small business loans, and credit products.
Lenders who recognise this will grow their portfolios and improve risk diversification. As these new methods become more widely adopted, we’re likely to see a broader range of tailored financial products, faster credit approvals, and increased confidence in lending. Moreover, alternative data will help improve risk assessment by helping lenders avoid both false positives and false negatives. It reduces default rates and strengthens long-term customer relationships.
This is an opportunity to do well by doing right. But seizing it requires institutions to rethink internal processes, adopt new tools, and shift mindsets. The tools to modernise credit scoring already exist. AI and alternative data offer a path to more accurate, equitable, and profitable lending. As a result, the financial industry must now act or risk missing out on a competitive advantage.
