Banks Need ‘Change Management Checklist’ in Battle Against Fraud

Artificial intelligence (AI) holds the promise to transform all manner of industries — helping banks, in particular, improve their anti-money laundering (AML) efforts.

But as Chris Caruana, VP of Strategy at Hawk:AI, Ramon Ramirez, director of AML/KYC Operations at Western Alliance Bancorporation, and Miguel Navarro, head of Client Identity Verification and Authentication at KeyBank, told PYMNTS, anti-financial crime teams within those banks are faced with a challenge:

Knowing if the AI technology they’re considering will, in fact, suit their needs.

An AI Arms Race

The pressure’s on — and using AI’s not really a question of whether, but when.

After all, the fraudsters themselves are using AI to create new attack vectors with speed and success. The banks are locked in an arms race with the bad actors. AI, the trio of panelists told PYMNTS, has become accessible enough so that it can be consumed and used by just about anyone — for good and for nefarious purposes.

“You can buy a synthetic identity for around $15,” on the dark web, noted Navarro, who added that “It’s a scary world out there … and we need a bit of assistance from AI to help us, in our businesses, make sure that we’re treading safely” as banks conduct their day-to-day activities and manage risk.

For the financial institutions (FIs) themselves, Ramirez and Navarro said, risk management poses a significant challenge. Navarro termed it the “defender’s dilemma,” where the bank must strive to be successful 100% of the time against fraudsters — but the criminals need to be successful only one time out of 100 attempts to inflict damage. 

“There’s just so much data out there,” Ramirez said, “that for a human being to sift through it becomes an insurmountable task.” 

Against that backdrop, Caruana said, FIs — from the community banks to the largest, global marquee banks — have mandates in place to explore how AI can be harnessed to improve efficiencies and to leverage the technologies to examine and improve existing anti-crime efforts.

“AML,” said Caruana, “is all about risk appetite and risk tolerance.” And that risk appetite, for an organization, needs to find its way into model-based decisioning. Ideally, he said, transaction-level data and other information can be fed into those models (which are constantly learning), which in turn create actionable insights across the organization. 

The Approach

Banks, like any organization mulling new technology, grapple with the age-old question of whether to “rip and replace” legacy technologies. Ramirez, Navarro and Caruana were quick to point out that FIs are loath to replace entire systems.

Ramirez stated that executives examining new transaction monitoring systems are doing so in order to augment the tools that already work well. Large, complex organizations have already invested significant resources in people and processes. There’s also the need to keep operations humming even in the midst of embracing technological changes.

Said Ramirez: “Banking does not stop just because you decided to change transaction monitoring systems. … Everything has to keep moving at the same pace.” Building parallel systems is enormously expensive, and might be reserved for only the largest, global banks. Rip and replace might appeal to only to relatively smaller FIs. Ramirez advocated an approach where AI is deployed “on top of” existing systems, in incremental fashion.

Added Navarro: “Rip and replace, from my perspective … sounds very aggressive,” and an evolutionary approach is ideal, with staff training along the way, and with the goal of breaking down silos within the FI.

“There always needs to be investment, when it comes to the people side, the process side, and the tool side,” added Navarro.

Building the Checklist

The incremental, evolutionary approach to AML technology, starting with a goal and working backwards — and in selecting providers (Hawk:AI among them), said Caruana, can render change management a bit less daunting.   

Fashioning a change management checklist, Caruana said, should tick several boxes.

“Whether you’re rip and replacing, or you are augmenting,” he said, AI-driven AML technology “should be easy to deploy and easy to use. That should be on your change management checklist.” When examining which providers to use, a few key questions are critical: How much experience have they had? Are they going to actively manage the solutions or will that be done via in-house teams?

“The disruptive technology,” Caruana said, “cannot disrupt the day-to-day operations of the financial institution.”  A checklist, Ramirez said,  also will need to examine the “chokepoints” within a bank’s current systems, where time and money are being consumed, and where AI can be most easily used for operational efficiencies.

Now’s the time for AI to find its place within AML efforts, Navarro said, because “The simplest implementations of AI are happening today, and it’s all going to get more complicated.” The FIs that don’t invest in AI now, he said, risk falling behind, in the competitive sense against their peers, and in the battle against the fraudsters. 

As Navarro said, “It’s not about plucking the ‘fruit’ from AI today, but rather about ‘planting the tree’ today.”