AI at Work: Enhancing Payee Match in Positive Pay

Jim Young

There has been talk for years about getting rid of checks in business banking, but they’re still very much around. Which means check fraud is still very much a problem.

Bankers know this. They also know the industry has been diligently working to solve check fraud via Positive Pay. At CONNECT 25, Corey Gross, VP of Product, AI & Data Solutions at Q2, shared how Q2 is helping in this fight by using AI to power the Payee Match functionality of its Positive Pay Solution, Centrix Exact Transaction Management System (ETMS).

Too many holes in the safety net

The most recent wave of Payee Match technology, Optical Character Recognition (OCR), does save human hours by automating the comparison of the payee on the actual check to the listed payee. But Gross noted that OCR can sometimes struggle with messy handwriting, low-res images, and complex check stock, leading to high false-positive.

Those false positives can place additional administrative burdens on businesses and erode customer confidence.

“You come to trust a solution because of its technical prowess, but it might actually undo the impact it’s making,” Gross said. “It’s the case of the cure being worse than the disease,” Gross said.

An ideal AI use case

Fortunately, improving Payee Match is a challenge well suited for AI. Contrary to much of the public conversation, not every fraud problem neatly pairs with an AI solution. It is important to remember what AI is and what it isn’t.

“AI isn’t the end. It’s just a tool,” Gross said. “We don’t buy hammers to go around finding nails.”
Instead, Gross and his AI team at Q2 follow this three-step process:

1. Identify a high-value problem
2. Select and implement the right tools
3. Observe, measure, and iterate

Committing check fraud by altering payee names certainly qualifies as a high-value problem. To solve it, Q2 implemented a machine learning model as the primary tool to extract data from checks and make comparisons, at scale, to improve Payee Match performance. Now, ETMS’s Payee Match functionality examines each check and sends a signal to the business client about whether the information matches and, if not, where it believes further investigation is needed.

“The (business client) operator is no longer responsible for making the match from reading an OCR,” Gross said. “The operator is now responsible for training and updating a machine learning model as new exceptions occur, so that the model can do a better job.”

This is Step 3 of the process, in which Payee Match continuously improves and ETMS catches more fraud with fewer false positives.

The Susser Bank success story

Susser Bank, a $2.5B institution based in Texas, has been one of the early adopters of this new Payee Match technology. Louann Habenicht, SVP, Director of Treasury Management Solutions & Client Experience at Susser, shared real-world examples of its impact.

A medical staffing company in Dallas-Fort Worth was struggling to sift through a high number of exceptions, to verify whether they were warranted or just a false positive. The AI-powered version of Payee Match changed that.

“They were like, ‘We love you! You have no idea how much time this has saved!’” Habenicht recalled.

Several business clients were dealing with fraudsters stealing checks directly out of their mailboxes and altering the payee information. “Payee Match stopped that, 100 percent,” Habenicht said.

ETMS with Payee Match has also helped Susser Bank clamp down on old-school fraud, in which perpetrators alter the payee names on handwritten checks.

“It could still be a check that has gone through the accounting system and then (fraudsters) take that check and write Susie Q’s name on it, and (Payee Match) catches it,” Habenicht marveled. “It just blows your mind.”

Stories like that have added up to impressive numbers in 2024 for Susser, including.

• 335K check processed
• Less than 1% false positive rate
• 100% of true fraud caught, valued at $12M
• 46% growth in Treasury Onboarding

More than just tech: Susser’s winning strategy

While AI/machine learning beefed up Susser’s Positive Pay performance, they didn’t view AI as “the silver bullet, but as one of the lead bullets that make up a much more holistic solution,” Gross explained.

Susser pushes Positive Pay adoption by urging each new client to add the solution and requiring any non-adopters to sign hold-harmless agreements with Susser. Bankers then keep coming back to non-adopters at later points, continuing to educate skeptical clients on the merits of Positive pay.

Susser trains each client that adopts Positive Pay on how to use the solution and provides a user guide documenting all the steps. Susser also turns on Positive Pay for clients a few days before go-live so they already have exceptions to view and can hit the ground running on day one.

Finally, Susser makes it easy for clients to see and respond quickly to exception notifications by making it available on the home page of their online banking account and via their mobile banking app.

What’s next for check fraud prevention

Q2 is looking at ways to improve check signature detection, The goal over time would be to create a check verification network to “identify common fraudsters as they move from committing fraud at one FI to another, to detect, ‘Hey, this matches a pattern,’” Gross explained.

Coverage expansion will also help. As more ETMS clients adopt the new Payee Match technology, more check stocks and types will be added into the machine learning knowledge base to improve performance.

Finally, Gross explained that the next evolution of Payee Match could change the operator role from training the machine learning model to instead overseeing a digital AI-powered “agent” who will have the primary job of analyzing and monitoring fraud.

“Check fraud’s not going away,” Habenicht said. “But (Positive Pay) adoption will increase with the more bells and whistles we have available through Payee Match.

“It’s really going to make a difference, utilizing AI.”