FICO’s New Auditable AI Models Can Help Credit Card Issuers Stay Ahead of Fraud and Compliance Risks

Analytics-software provider FICO rolled out new models today to help credit card issuers and other financial services providers use artificial intelligence responsibly. The programs can help issuers use AI with confidence to detect payment fraud and adhere to compliance initiatives.

Artificial intelligence has changed the way businesses operate in the 2020s. A recent McKinsey & Company study shows that 78% of organizations that responded to the company’s survey now use AI in at least one of their business functions. That rate represents an increase from 72% of the share of organizations that reported using AI in 2024.

And only 55% of organizations said they used AI in at least one business function in 2023, indicating just how quickly its use is spreading.

But widely available AI programs can produce mixed results. Artificial intelligence systems aren’t infallible, often yielding outputs that users find lacking.

People so often find inconsistencies in the results AI programs yield that they’ve come up with a term, AI hallucinations, to describe the inaccurate or misleading answers that the systems can generate.

Financial institutions can strengthen their risk management programs with the new solution from FICO.

The FICO Focused Foundation Model for Financial Services (FICO FFM) aims to provide reliable, auditable outcomes for companies in the financial services sector.

Scott Zoldi, Chief Analytics Officer at FICO, said in a press release that the company’s Focused Foundation Model brings a practitioner’s approach to the use of generative AI in the financial industry.

“FICO FFM enables enterprises to use small language models built for their specific business problems, significantly helping to mitigate hallucinations, provide transparency, auditability, and adaptability,” Zoldi said. 

“These domain-specific models can result in 38% lifts in compliance adherence use cases and more than 35% lifts in world-class transaction analytic models, in areas such as in fraud detection,” he added.

The lift potential FICO’s new models promise may catch the attention of credit card issuers that are anxious to use AI to ramp up their fraud mitigation programs but have lacked confidence in the accuracy of certain AI-powered solutions.

Training a language learning model with a significant number of parameters can require a massive amount of computing power. And that power is likely to cost more than the typical credit card issuer has in its budget for new technology expenditures.

But FICO’s models use up to 1,000 times fewer resources than those required by a general-purpose language learning model researchers train on wide-ranging global knowledge, according to the press release.

That means issuers with relatively small budgets don’t have to sit on the sidelines while competitors with more financial resources leverage AI solutions to get ahead.

When it came to forming its new models, FICO built and engineered them in-house. And it focused the models on curated sets of data relevant to the financial services arena. 

The FICO Focused Foundation Model for Financial Services consists of two programs that can help financial institutions better harness the power of generative AI. 

The FICO Focused Language Model for Financial Services can reduce AI hallucinations because the company built it on data that is domain and task specific. The company said that focused language models have a low barrier to entry and serve to streamline a financial company’s operations.

“Built on curated data and responsible AI principles, these models are becoming essential tools for institutions that require precision, transparency, and scalable trust,” Megha Kumar, Research Vice President of Analytics and AI at International Data Corporation, remarked in the FICO release.

The other model that FICO introduced under its FICO FFM umbrella is the FICO Focused Sequence Model for Financial Services. The sequence model may be a game changer for institutions looking for a more effective method to combat payments fraud.

And many credit card issuers find themselves in that position these days. A recent survey from the Association for Financial Professionals reports that bad actors victimized nearly 4 out of 5 organizations with payments fraud or targeted them with a fraud attack in 2024.

But FICO said in its press release that its sequence model can expose transactional behavior patterns that once were not possible to identify.

Some executives at financial institutions may be wary about using programs that rely on artificial intelligence, even if a name that many people in the financial industry trust is behind the solution in question. 

In a new article from American Banker, Andrew Foster, Chief Data Officer with M&T Bank, underscored the dilemma financial leaders face when determining how much their employees should turn to generative AI to help them complete tasks.

“You need to embrace adoption because you become more efficient,” Foster said. “If you don’t, you fall behind peers who are using it. But you need to have responsible usage and retain accountability.”

To make its language and sequencing models more transparent, FICO added a scoring feature that rates how reliable each result is.

Trust scores from FICO can help credit card issuers have confidence in the results they receive from the company’s models.

Known as trust scores, FICO said in the release that organizations can use them to configure their own risk parameters and help ensure their teams use generative AI responsibly.

Now may be the right time for credit card issuers who haven’t incorporated AI tools into their business processes to consider doing so. 

A recent report suggests that leaders in the payments space, including Mastercard and PayPal, are speeding up their AI-powered projects, which could soon put pressure on credit card issuers to match their pace of innovation.

FICO’s new models offer tools that can help issuers adopt AI more confidently and use it in a responsible, effective way.