As Artificial Intelligence both simplifies and changes fraud activity, SotaTek is seeing great success with SotaShield, a platform that helps banks screen merchants, detect emerging risk patterns and reduce manual investigations. CEO Meng Khong Tong said one key is for companies to improve their onboarding security procedures.
During the onboarding process, criminals can deploy synthetic identities, shell entities and AI-generated business fronts that are not detected by traditional fraud detection systems, which focus on transactions. Agentic AI, OSINT, sanctions screening, and transaction anomaly detection are among the tools that are combined to spot risky merchants before they act. SotaTek blends AI, blockchain, and cloud-native architecture.
Tong said AI has upended traditional approaches to everything from software development to security. Companies have employed thousands to develop and maintain their data systems, but AI has reduced the needed headcount and deployment time. Fraudsters have leveraged AI to build synthetic identities and conduct scams at unprecedented rates. This allows them to target all areas of the traditional customer relationship, including onboarding, spending, and cross-border payments.
Financial institutions are in a pickle, and they know it. They have to modernize their tech stacks across the board (not just fraud detection systems), but in the past it would take many years, legions of staff, and plenty of cash. Given the rapid pace at which technology progresses, their finished product may be out of date before it debuts.
How AI and fintechs have changed onboarding
Tong said AI reduces that development time from as much as eight years to maybe one. System maintenance and upgrades are easier, so companies can keep pace more easily. That combination of strong benefits, manageable risk and necessity has them acting en masse.
“Is it risky?” Tong asked. “Yes, but the risk of not doing it is higher.”
AI also helps the new participants maximize the volumes of internal data that has long sat in multiple, unconnected storehouses. SotaTek helps them connect these many systems, regardless of programming language and other previous barriers.
“This data was never shared,” Tong said. “Different departments and vendors were attacking the issues by themselves.”
Fintechs have also forced the banks to act. Historically, traditional players onboarded customers in traditional ways. Business account creation required branch visits.
The combination of digital development and the pandemic forced a shift, but one that made the fraud fight much harder. Fraudsters set up fake companies online at scale and pivot rapidly. Merchant processors might have a few dozen human screeners. That’s not a fair fight.
“If an AI-based system isn’t going to catch you, humans probably won’t either,” Tong said.
Now, institutions can create a series of AI agents, each tasked with a specific job: one handles facial recognition, another AML, and so on.
“But the challenge is to not automate the whole process, because RPA can do the same thing,” Tong cautioned.
The trick is to provide a smooth user experience while maintaining security and accurately depicting the entire merchant lifecycle. SotaTek helps this by creating a merchant social graph that maps the merchant’s behavior and its activity with all other merchants. That process is living and evolving, absorbing company changes like sanctions and personnel changes.
“Waiting for things to happen and transactions to fail, then you realize it’s fraud?” Tong asked. “Better still, you build this into the infrastructure stack so that when you set up the company you’re monitoring, you’re monitoring five or six different parameters. The AI learns the whole life cycle of the merchant and decides if this is something risky or not.
“It’s a journey of learning. The AI system is learning to strike a balance between not introducing too many hurdles or processes versus a good customer experience. We don’t want to keep validating things.”