Fighting Fraud: How AI is Closing the Gap

Fraud in equipment finance isn’t new. In fact, it’s one of the hottest topics in the industry.

And for good reason. Research indicates that every dollar lost to fraud costs North American financial institutions $5 – a poor return on investment by any measure.

What is new and what’s encouraging is the rapidly expanding, changing breadth and scope of AI solutions lenders can use to detect and mitigate fraud. We talked to several ELFA members to get their views on AI’s current use in the industry, find out where they are experiencing the greatest demand for services and learn about the types of fraud they are seeing most frequently. 

The Burning Issue: Human Judgment

First and foremost is the question that’s on everyone’s mind: Is AI taking away human interaction and judgment in the equipment lending and financing process? Is it completely changing the fabric of the industry? It’s worth addressing this overarching issue before diving into specifics.

And sometimes, the best use of AI isn’t AI alone. Tom Ware, President of Tom Ware Advisory Services, LLC, finds AI is appropriate when used in conjunction with traditional, transparent mathematical models – and human judgment. “What AI is best at, and can be even brilliant at, is making suggestions, rather than making decisions,” he says. “AI can suggest which deals to look at more, or less, when it comes to detecting possible fraud.” 

Demand

The strongest demand for AI in the industry today is around originations, credit operations, document-heavy processes and portfolio oversight, according to Corazzi.

Lenders, equipment finance companies, and private capital all want to move faster without weakening control or creating exposure, says Corazzi. That is creating interest in automating repetitive validation, reducing manual review and using data more continuously to identify exceptions or unusual activity.

Multi-dimensional, Multifaceted

Experts agree: In every functional area, fraud is multi-dimensional and multifaceted. Document and identity fraud are becoming significantly more sophisticated. Altered financial statements, manipulated invoices, identity impersonation and increasingly convincing synthetic content all create new challenges for lenders.

One of the most important dynamics to understand, says Corazzi, is that AI is operating on both sides of the problem. The same technologies that help lenders analyze documents and identify anomalies can also help bad actors create more convincing documents, identities and communications. Fraud techniques are becoming faster and more scalable, which means controls cannot remain static.

“The fraudsters are getting sophisticated,” agrees Dolan, who finds third-party fraud the most prevalent. A fraudster can make very subtle changes to a document, and can impersonate a real company with real documents and real identities they’ve purchased on the dark web, he says. Exacerbating the problem is when the fraud hits multiple deals. It’s one thing to identify fraud in a one-off deal, he explains. But it becomes very costly for a lender processing multiple deals from a particular vendor.

Fraud Categories

Fraud can occur anywhere in the lending lifecycle, but following are key areas where fraud is occurring most frequently in the industry, according to the experts.

• Identity theft and mismatch. Synthetic identities and misrepresentation in the origination process lead experts’ lists of areas where they see most instances of fraud. Basically, Sharma explains, it involves someone trying to apply for a loan on another person’s behalf. Biometric verification and machine learning-enhanced document verification can identify synthetic identities. 
• Document manipulation. In the application process, submission of tampered documents or falsified financial information is common, “more common than you’d like to believe,” according to Sharma. It could be bank statements, invoices or even tax returns. AI software now can trace something as small as an edit to a PDF that a fraudster used, the time they made an edit and the exact values they changed.
• Entity fraud. Seen often in underwriting, entity fraud occurs when someone uses a stolen, manipulated or fake business identity. Companies involved in fraud detection and mitigation are using AI to verify status of entities in real time, looking not only at Secretary of State websites, but also performing wide-ranging research to analyze an entity’s entire web presence, from social media reviews to news media coverage. Sharma notes that in one case, an AI system captured an instance of entity fraud from a Yelp review. 
• Asset fraud. Attempts to deceive a lender on the value, ownership, condition – and even existence – of equipment can be devastating, says Dolan. AI is helping with the ability to do significant research on the equipment itself, thereby enabling it to identify potential cases of overvaluing.

Fraud Detection: A Continuum

While these categories represent frequent fraud targets, fraud detection is not about any single area or checkpoint, but rather a continuum. As Sharma details, AI’s ability to layer data and patterns across processes allows evaluation with broad context, rather than relying on a single static check at one point in time from one source (such as a credit bureau). 

“It's not always like a smoking gun,” Dolan emphasizes. “Any one anomaly may mean nothing, but AI has the ability to look at combinations of things and identify potential fraud.” Factors can get complex, but as a very simple example, he cites a company that just opened a brand-new office and then presents a phone number from a prepaid phone. “There may be legitimate business cases for both of those pieces of information, but when you combine them, you may want to look a little deeper.”

Lenders are increasingly moving toward more continuous monitoring, connecting operational asset and portfolio data to surface potential issues earlier, adds Corazzi. In doing so, they can surface potential issues earlier and direct skilled people toward the areas that warrant investigation. Looking ahead, he believes the future of fraud mitigation will depend less on a single detection tool and more on connected data, continuous monitoring, and the ability to evolve controls as the threat changes.

“The best technology only gets you so far if the data that's fed into it is still siloed,” says Kellershof. In gathering and connecting data from many different sources, AI provides embeddedness across the full workflow. “AI’s real value is its ability to eliminate silos of information and present a complete picture for the credit analyst.”

Dolan also notes that lenders layer in fraud checks as deals progress. Early in the application process, routine AI checks can highlight inconsistencies or other issues quickly. As the process continues, and the stakes increase, a lender may want to invest in more intensive fraud checks.

Beyond Fraud Detection

Service providers are also using AI to speed and streamline processes like application intake and document generation. While not fraud detection work itself, AI is removing a great deal of labor-intensive work. Reducing data entry and re-keying work means fewer errors, less time training and retraining employees, and the ability to process more applications – and major time savings. “It gives lenders the chance to grow faster and more profitably,” Sharma says.

AI is great at flagging data inconsistencies throughout the lending process, she states. Again, the job of AI is to identify the issue, and to present the data in a digestible format. “It’s up to the lender to decide what action to take, whether it’s to engage with their customer or other,” she says.

Lifecycle marketing is another area where AI is making a mark in the equipment finance industry. By analyzing the life of a contract, indicating potential issues and identifying critical communication touchpoints, AI can help develop – or maintain – a solid ecosystem of contracts.

Looking forward, ideas of a fraud consortium are emerging. Kaaj, for instance, is expanding ways to help lenders identify ongoing fraud types through network-connected data. “Helping the industry share outcome data and emerging patterns helps everyone,” says Sharma.

And for those in the industry who are using AI on their own, whether to learn more about a particular area or ask a specific question, Ware recommends a committee approach. “Ask the same questions from multiple AI assistants,” he recommends. “See to what extent they agree or disagree – and if they disagree, ask each what they have to say knowing the conclusion of the other.” The approach, he contends, reduces the risk of unreasonable or, on occasion, bizarre AI results – and can sometimes even result in AI apologies.

Where to Start

When the 2026 Survey of Equipment Finance Activity (SEFA) asked members about their AI plans, more than half of respondents indicated that, over the 12 months, they had either no AI investment planned or investment of less than $100,000 planned. While these numbers may change rapidly, part of the issue is uncertainty about where to start.

For equipment finance leaders, Corazzi suggests beginning by examining where information enters the organization, where it is validated, where manual handoffs occur and where decisions are being made without enough visibility. Those are often the areas where better data, stronger workflows and responsibly deployed AI can have the greatest impact, he says.

Then, when implementing an AI solution, he recommends companies be very clear about what AI is authorized to do and where human accountability remains. “Give people better information earlier, maintain an audit trail, establish clear escalation paths and keep accountability with the institution,” Corazzi says.

Conclusion

Fraud is a real, measurable threat to equipment finance companies. The good news is that fraud detection and prevention is also real, with measurable results. AI will not eliminate all instances of fraud. But it can highlight the areas that need human consideration and judgment.