In the modern American economy, credit scores are the north star—simple for lenders, standardized for regulators, and central to consumers’ lives. With the proliferation of AI, lenders are now facing an uncomfortable truth: credit scores are flat.

They capture a snapshot of the past, not a moving picture of the present. And in today’s volatile economy—where income streams are fragmented, careers are non-linear, and financial behavior changes faster than models can keep up—that flatness is becoming a serious problem. AI underwriting has emerged as a new dimension layered on top of credit scores—one that promises greater nuance, inclusion, and accuracy, while raising fresh questions about fairness, transparency, and regulation.

The limits of traditional credit models

Traditional credit models were built for a different world. A world where people held long-term jobs, paid bills by mail, and followed relatively predictable financial paths. These models rely heavily on historical data, such as payment history, credit utilization, length of credit history, and types of credit used. That works reasonably well—until it doesn’t. Consider two increasingly common scenarios:

• Volatile economies: Inflation spikes, layoffs, gig work, and sudden shifts in spending behavior can quickly make historical data misleading. Someone with a pristine credit score last year may be struggling today, while someone with a past hiccup may now be financially stable.
• Thin-file customers: Millions of people—young adults, immigrants, gig workers, and underbanked populations—either lack sufficient credit history or have none at all. Traditional models simply don’t know what to do with them, often defaulting to rejection.

In both cases, lenders are flying blind with two-dimensional tools in a three-dimensional financial world.

Adding the third dimension: AI underwriting

AI underwriting addresses this visibility gap by analyzing not just static data but also dynamically tracking and interpreting ongoing financial actions. Its capacity to process complex, real-time information directly illuminates a borrower’s present and adaptability, moving beyond what traditional models offer. Instead of asking, “What has this person done with credit in the past?” AI models ask, “What is this person doing financially right now—and how likely are they to adapt?”

To answer that, lenders are turning to alternative data and generative AI models. This can include:

• Cash flow data from bank accounts
• Income consistency rather than income level
• Utility, rent, and subscription payment behavior
• Employment patterns and industry stability
• Real-time spending and saving habits

Generative AI uniquely interprets complex, unstructured data that traditional systems overlook. It can recognize behavioral patterns, simulate changing economic scenarios, and predict borrower responses to stress—adding insights unattainable with previous scorecards. The result is underwriting that feels less like a score and more like a profile.

From static scores to living models

One of the biggest shifts AI underwriting introduces is time sensitivity. Credit scores update slowly over months and years, while AI models can update continuously. For example, instead of penalizing a borrower for a missed payment six months ago, an AI-driven system might notice that their income has stabilized, their spending has become more conservative, and their savings balance is growing again. That context matters. It allows lenders to differentiate between temporary disruption and structural risk.

In a volatile economy, this flexibility can mean the difference between unnecessary rejections and smart approvals. For borrowers, it can feel like being judged on who they are now—not who they were during a rough patch.

However, while AI adds nuance, it brings a critical challenge: explainability.

Of course, none of this works if decisions feel like they’re coming from a black box. Historically, one of the biggest criticisms of AI in finance has been explainability. Regulators and consumers alike want to know why a decision was made, not just what the decision was.

New AI underwriting models are being designed with explainability at the core. Instead of spitting out a yes or no, they can highlight key contributing factors. Some lenders are even using generative AI to translate these factors into plain-language explanations for consumers. That transparency builds trust—and gives borrowers actionable insights on how to improve their financial standing.

One of the most promising outcomes of AI underwriting is its potential to expand access to credit without increasing systemic risk. Thin-file customers, in particular, stand to benefit. When alternative data is used responsibly, it can reveal reliability that traditional models overlook. A gig worker with irregular income but strong cash management habits may be a safer bet than their credit score suggests.

That said, inclusion can’t come at the expense of fairness. Alternative data can introduce new biases. Spending patterns, for instance, can correlate with socioeconomic factors. That’s why lenders invest in bias testing, audits, and governance frameworks.

Regulators are watching — very closely.

As AI underwriting gains traction, regulators are paying attention. In the U.S. and globally, regulators are asking tough questions. Can decisions be explained in human terms? Are models compliant with fair lending laws? Is alternative data being used ethically and consistently? Who is accountable when an AI model gets it wrong?

So far, the regulatory stance has been cautious but not hostile. Most regulators recognize that static models have limitations—and that innovation is necessary. The emphasis is on transparency, auditability, and consumer protection rather than outright restriction. In many ways, this scrutiny is healthy. It’s forcing the industry to build AI systems that are not just powerful, but responsible.

Looking to the future, consider the evolving role of credit scores: from gatekeepers to signals.

Credit scores aren’t disappearing anytime soon. They’re too embedded in financial infrastructure for that, but their role is changing. Instead of being the final word, credit scores are now one signal among many. AI underwriting adds depth and adaptability, making credit decisions more conversational than final.

In a world where financial lives are increasingly complex, that shift feels overdue. The challenge now is balance: leveraging AI’s power without losing human oversight, expanding access without amplifying bias, and innovating fast enough to keep up—without outrunning regulation.