AI vs. Human Judgment: The Future of Credit Risk Decisions

Nearly 75% of Atradius credit decisions in the U.S. are made automatically. That figure shows just how much technology already influences credit risk assessment. Advanced analytics can evaluate financial performance, payment behavior, industry conditions, and broader economic signals quickly and consistently.

But credit risk is rarely black and white. What happens when the numbers do not tell the whole story?

We spoke with Patrick Scardina, Senior Risk Underwriter Manager at Atradius, about where technology adds value, where human experience still matters, and why the future of credit decision making may depend on both.

Let’s start with the big question. How much of a credit decision today is data, and how much is human judgment?

Patrick: Every credit decision is a blend of technological intelligence and human insight. I like to think of it this way. Data provides the map, but experience helps navigate the terrain.

Nearly 75% of our credit decisions in the U.S. are made automatically. Our models draw on years of credit experience, payment and claims history, default trends, industry benchmarks, and economic indicators. Automation is particularly valuable for straightforward decisions, including smaller credit requests or businesses with clear risk profiles. It allows us to respond faster while evaluating similar risks more consistently.

But the more complicated decisions often live in the gray areas. That is where experience becomes important.

An underwriter can look beyond the financial results and ask what is actually happening within the business. We can evaluate circumstances that may not fit neatly into a model. The strength really comes from combining the two. Technology gives us speed, scale, and consistency. Human judgment gives us context and perspective.

AI can process enormous amounts of information. Where is it actually making the biggest difference for an underwriter?

Patrick: I think one of the biggest opportunities is helping us make sense of all the information available. Think about what goes into assessing a company. You might have financial statements, payment histories, industry reports, news coverage, economic developments, and other information coming from different places.

Working through all of that takes time.

AI can help connect those sources quickly, identify emerging trends, and flag information that deserves a closer look. It might surface a developing risk or pattern that is difficult to recognize when you are looking at each piece independently. For an underwriter, that changes where we can spend our time.

Instead of concentrating so heavily on gathering and organizing information, we can focus on interpreting it. What is the story behind these numbers? Is this trend significant? Is something changing within this company or its industry? I see AI as an amplifier of what an experienced credit professional can do, rather than a replacement for that expertise.

So, what can you recognize as an underwriter that AI might miss?

Patrick: A lot of it comes down to understanding why something is happening.

AI is very good at recognizing patterns and identifying unusual information. But data can show you what is happening without necessarily explaining the reason behind it. That distinction becomes especially important with privately held companies, where publicly available information may be limited.

As underwriters, we can speak directly with finance leaders, business owners, and other decision makers. Those conversations can uncover information that might never appear on a balance sheet.

A company could be planning an acquisition, dealing with customer concentration, preparing for succession, or navigating supply chain pressures. There might be a major strategic investment affecting short-term performance.

We can take that context and test it against financial performance, payment behavior, industry conditions, and other information. That is where experience adds difference. You are not replacing the numbers. You are determining what those numbers mean within the bigger picture.

Can you give an example where that context could actually change how you view the risk?

Patrick: Imagine a manufacturer whose sales suddenly decline by 30%.

If you are looking only at the data, that decline could resemble the beginning of financial distress. You see a significant deterioration and understandably want to know whether the company is becoming a greater credit risk.

But then you speak with management.

Maybe the company temporarily reduced production to install new equipment, modernize its operations, or expand capacity. Revenue has declined in the short term, but the business is investing in its future. That does not mean you ignore the 30% decline. It means you start asking different questions.

Why did sales decline? How long will production be affected? How is the project being financed? What does demand look like when capacity comes back online? Those answers help you distinguish between a temporary disruption and a more fundamental deterioration in the business.

There is another layer to this too. Underwriters are not only evaluating the financial information. We are also considering how credible and complete that information is. Direct conversations with management, industry knowledge, and broader market intelligence can sometimes identify concerns that historical financial data has not captured yet. That ability to the story behind the numbers is still incredibly important.

What concerns you most about businesses relying too heavily on AI for credit decisions?

Patrick: One concern is explainability.

If a model recommends declining a credit limit or identifies a business as higher risk, we need to understand why. You need to be able to explain what is driving the decision. The other challenge is what happens when the environment suddenly changes. Models can identify historical patterns extremely well. But those patterns become more difficult to rely on when something fundamentally changes the business landscape.

COVID was an extreme example. Industries were disrupted almost overnight. Revenue disappeared, operations stopped, and some traditional indicators of credit strength suddenly had to be viewed differently. Changes in tariffs and trade policy can create a similar challenge. Companies may face higher input costs or make changes to sourcing, pricing, and supply chains. Those are situations where you must ask whether yesterday’s patterns still apply to today’s risk. That is where human judgment remains important. An experienced underwriter can question the signal, consider what has changed, and determine whether something represents temporary disruption or a longer-term structural issue.

For me, the future is not about choosing between AI and human judgment. AI helps us process more information and recognize signals faster. The underwriter provides context, skepticism, and accountability.

AI can help us see more. Human judgment helps us understand what we are seeing.

Get the Latest Trends & Insights

Be the first to know about economic and political changes affecting you and your customers’ business, delivered right to your inbox.

Sign Up