Know your customer. Its acronym is KYC, but you couldn’t blame someone for confusing it with KFC because for banks, it’s a tough muddle with all the likeability of spent chicken grease. As for anti-money laundering compliance, or AML, it’s not far behind in the onerous burden department. And while both these processes protect everyone’s best interests, they also amount to A Mountain of Labor (an AML of a different kind).
Yet a new Celent report suggests the banking industry’s days of soul-crushing manual compliance chores – and the multi-millions of dollars spent on them – are numbered, thanks to artificial intelligence (AI).
Still considered novel by many, AI tools are providing early adopters with real solutions to their compliance challenges, according to the paper entitled Artificial Intelligence In KYC-AML: Enabling The Next Level Of Operational Efficiency. It focuses on Tier 1 institutions, as their burdens are especially heavy.
“Smaller and regional banks have simpler needs in their KYC-AML operations,” says report co-author Arin Ray, a Celent analyst. “By comparison, multi-geography banks with multiple lines of business face more complex risks.”
The inefficiencies plaguing KYC-AML are pervasive, to say the least. With onboarding tasks, analysts spend an average of more than 15 hours to review a single corporate client file. But roughly a third of that time (35 percent) goes to skilled analysis; the rest is wasted on low-level and substantively manual knowledge work to collect and aggregate data.
At scale, the price is staggering. The report estimates the largest global banks spend nearly $1 billion annually on KYC-AML alone. Even the bottom 75 percent of banks in developed markets can spend up to $200 million per year.
There can be considerable brand and business impacts as well. “If a significant corporate client experiences onerous or cumbersome KYC processes at one bank, compared to more streamlined and efficient process at another, then this can be a competitive differentiator,” Ray points out.
For leaders, futuristic solutions are already here
This is where AI comes in. The report demonstrates how AI software solutions can provide banks with solutions that exceed what their existing rules-based KYC-AML technology can do. And in some cases, it’s already happening. “AI tools are definitely being used to address unstructured data analysis,” Ray says.
To bear this out, look no further than recent public statements by United Bankshares. According to a Wall Street Journal blog post, Darren Williams, executive vice president and chief risk officer, discussed his institution’s AI trial. It’s aimed at reducing some types of monitoring inefficiencies – from 95 to 50 percent – which “would be real close to a 30 to 40 percent cost savings for us,” Williams says in the piece.
Making four less of a chore
Indeed, reducing monitoring inefficiencies appears to be a major push among early adopters, according to Ray’s research. During a follow-up discussion with BAI Banking Strategies regarding his report, the analyst drilled down to discuss four KYC-AML use cases where AI is taking hold.
Client activity research and analysis. During onboarding, and throughout a clients’ lifecycle, banks must research and analyze external sources, such as traditional media or social media sites, to ferret out client involvement in negative or questionable activities. As this is unstructured data, traditional software tools are poor at matching activities with the proper client and/or client employees. Nor can existing tools “learn” new terms as colloquial speech patterns evolve, requiring humans to constantly upgrade the lexicon.
Communications monitoring. Whether there’s discourse between employees, or when employees and customers interact, AI tools can sift through large volumes of unstructured content, such as emails and texts, to uncover suspicious items. The flagged content could certain words or code words, which will lead to further review by another employee.
Multi-language monitoring. Compared with existing tools, AI solutions offer a vastly superior option to translate or transliterate client names, along with code words and other data in languages and scripts other than those used in languages with Latin roots. For example, bad actors frequently transpose their first, last and middle names, as well as translate them into another language, to fool traditional rules-based software.
Watch list monitoring. Besides constantly scanning government watch lists, banks also rely on internal and industry lists. In particular, industry lists tend to be long and include names in non-Latin languages – which makes ongoing analysis a challenge for existing tools. AI solutions excel in their capability to curate and monitor such lists.
In each case, the benefits transcend simply containing costs: Banking analysts become empowered to focus on high-value tasks rather than low-level sifting chores. “Deploying AI in any of these areas also reduces reputational risks associated with non-compliance,” Ray points out. “If compliance gaps become public, reputational damage can be as significant as regulatory fines.”
A quartet of AI vendor options
For banks seeking to join the AI wave, Ray says plenty of options are available, with new tools continuing to surface. “Some banks are even investing in emerging AI solutions in hopes of reaping the benefits themselves,” he says.
Banks have four basic categories of vendor types to consider:
General AI solution providers. These providers offer AI solutions outside of the banking industry – such as defense and intelligence, healthcare, internet and online retail industries. Yet the tasks at hand resemble what’s needed for KYC-AML, and are therefore easy to replicate in a banking context.
KYC-AML vendors incorporating AI. As AI technology matures, some existing KYC-AML software vendors have incorporated AI into their solutions or have it on their development road maps. In some instances, people financial services and compliance technology backgrounds have created startups to focus on developing AI solutions for KYC-AML and related areas.
Established banking solution providers adding AI. While Ray believes this category of players is still at a nascent stage, he’s noted a few stand at an advanced stage of developing and launching AI capabilities.
AI an “obvious” choice
Beyond the Celent report, researchers with experience in AI and banking substantiate Ray’s findings that the technology is ready for prime time in KYC-AML.
“For the sake of raising alerts that are then verified by human staff, this use of AI is obvious,” says Kentaro Toyama, W. K. Kellogg Associate Professor at the University of Michigan School of Information. “And it’s certain to be cost-effective over having people perform the tedious task of poring over raw data.”
However, Toyama cautions banks against outsourcing decisions to machines: what he calls “the first sin of AI.”
“As the Celent report makes clear, the near-term opportunity of AI systems is in sifting through data and raising alerts that will be verified by trained staff, not in making final decisions that directly affect customers,” he says.
Regardless, Toyama agrees with the Celent report’s basic assertion: The potential for AI in the financial services generally, and KYC-AML in particular, is nothing short of tremendous. And with any luck, those six pesky initials will no longer strike fear, nausea or annoyance into the hearts of hard-working bankers.
Anne Rawland Gabriel is a contributing writer to BAI Banking Strategies. She is based in Minneapolis/St. Paul, MN.