Protecting Banking From AI With AI

Banking professional services leader Genpact is using generative AI to protect customers. 

AI is both the threat of and the answer to financial crime. Genpact, one of the most significant providers of professional services to the banking sector, is pioneering the use of AI to detect fraud and comply with sanctions.

Brian Baral, Global Head of Risk Management, Head of Strategy for Banking and Capital Markets at Genpact, led the adoption of Amazon Bedrock into the riskCanvas technology service offered by Genpact.

Genpact can trace its roots back to GE, the conglomerate that makes aircraft engines, locomotives, medical devices, and much more. GE Capital Services, the finance arm of GE, formed an international services firm in India in 1997 to optimize the myriad business processes that financial services require. This included business process management and technology.

“Since then, we have grown significantly and become one of the leading professional services firms,” Baral says.

The Banking and Capital Markets arm that Baral works for accounts for a third of Genpact’s business and employs 30,000 of the 115,000 employees of the business.

“We support large banks like JP Morgan and Goldman Sachs through to Fintech firms like Chime, the mobile banking service provider,” Baral added.

AI The Threat

“With the advent of generative AI, we are seeing more attack vectors coming in, such as synthetic identities,” Baral says. “So it is more of a challenge for banks to really know who their customers are.”

AI is making the already complex landscape of financial crime even more opaque. As Global Head of Risk Management, Baral and his staff of 8,000 help their financial services customers fight money laundering through screening of customers to ensure they are valid and not a shell company.

This requires monitoring transactions and, if there is suspicious activity, notifying regulators. Money laundering is always important to financial services providers, but in a time of heightened conflict and global tension, there is increased pressure on banks to truly know their customers.

“We also fight against fraud. In anti-money laundering, there is a very rigid set of compliance that you have to follow, and if you run afoul of those, you can get into big trouble with your national agencies,” Baral explains. “With fraud, there are regulations, but you are also battling against a loss to your shareholders.”

His team provides analysis and investigation services to its clients, which include major banks Citi, Goldman Sachs and Santander, as well as challenger Fintechs.

Financial services firms are bound by regulations to produce a Suspicious Activity Report (SAR), sometimes called a Suspicious Transaction Report (STR), whenever they believe a transaction is suspect, out of the ordinary or could be a way of hiding other transactions.

Each SAR or STR is sent to the law enforcement agencies in the countries where the bank is operating. “We do see a continual uptake from customers for more support for investigative research, forensics, and analysis,” Baral says.

AI The Answer

Faced with an increasingly complex financial crime landscape, Genpact has partnered with AWS to integrate Amazon Bedrock into riskCanvas, an end-to-end Genpact financial crime technology that is a software-as-a-service (SaaS) platform.

Amazon Bedrock is the generative AI and foundational model technology from AWS. riskCanvas is a platform that supports clients in running their know-your-customer processes, transaction monitoring, and sanction screening; among its users is one of Australia’s largest banks.

“Generative AI does a really good job of summarising, filling in blanks, and connecting dots, so Generative AI is fit for purpose in these areas,” Baral says.

Baral and his team took the large language models (LLM) that are part of Bedrock, and utilized the security within Amazon Bedrock to ensure that no data from the banks that Genpact works for goes out onto the wider internet.

“This ensures that the models are clean, not just using the whole of the internet like OpenAI does,” he says of the benefits of working with the Bedrock AI functionality.

Baral says the introduction of generative AI into  SARs and STRs was a case of partners sharing their ideas and pain points and the trio jointly working together. It is a case study of how AI will begin to transform organizations.

AI requires high levels of compute, which a hyperscale cloud computing provider can access, Genpact has detailed technical insights into a vertical market, and the financial services sector has the business case – satisfying customers while keeping them in compliance with banking regulatory agencies.

Business Benefit

Baral says Genpact is now able to develop generative AI products at the pace that modern financial services businesses require. One customer, Apex, a Fintech provider of clearing services, is already using riskCanvas and generative AI to create SAR narratives and case summaries.

The U.S.-headquartered business expects to see the time spent on case summaries to reduce by 60%. This is a significant saving; Baral says up to four million SARs are produced a year in the USA.

“They take time to do, typically two to three hours, and they have to be written in a format, as it is a very sensitive artifact,” Baral says. 

Baral explains that data and examples of what a good SAR looks like are parsed into the generative AI, which then drafts a SAR.

“There is a real efficiency, whereas with humans, there is a lot of variability,” he says. “This system gives the analysts a near-perfect SAR, but there is still a human in the loop who reviews the output and can edit it. Organizations will save money, and more importantly, they can do investigative work. So analysts can be analysts again as they are free from the grunt work.”

The freedom to do less administrative work is going to become increasingly important in the quest to stem financial crime, Baral notes.

“There are continually more challenges and more bad actors, and they will always find more ways to attack,” he explains. “So you want to find more ways for your staff to apply their intellect to problem-solving.”

Baral uses the analogy of squeezing a balloon, causing the air to move and bulge somewhere else. Financial crime can be squeezed in one area, and the problem moves, but the greatest risk for the financial services sector is not being able to squeeze. Generative AI could provide the financial services industry with a greater ability to apply pressure on those looking to compromise the sector.