Blog

The Self Driving Car of Financial Crime Detection

03.17.2021 | By Mark Speyers
 

In 2015 machines could “see” better than humans and in 2016 AlphaGo learned to play Go and beat Lee Sedol. By 2018 Waymo launched the world’s first self-driving taxi service, and in 2020 GPT-3 learned the internet. It is clear that the pace of AI innovation is accelerating exponentially, and this is only a glimpse of things to come.  

The question we are asking ourselves at AyasdiAI is “can we build the self-driving car of financial crime detection?” It’s a bold aim, yes. But is it impossible – absolutely not!  

Illicit money flows in ever greater volumes through the global financial system; fraud is the most common crime in the digital world and criminal sophistication outpaces the industry’s response. Anti-money laundering systems are swamped by false positive alerts forcing banks to employ armies of investigators who spend 99% of their time looking at completely normal financial behavior. Surely the AI revolution can help with this challenge? Yes, it can; of course, it can. 

It is not a simple problem to solve though. With human adversaries operating across borders, insider compromise, complex and changing regulations, and legacy systems that need to be maintained and adapted create a complex environment in which to operate. But this is no more complex than building a car that drives itself while following the laws of the road and avoiding crashes yet self-driving cars appear to be light years ahead of the response to financial crime. 

Self-Driving Car Example  

A self-driving car works by combining an array of different AI components into an overall “intelligent” system that can sense, monitor and adapt to its surroundings to safely navigate from A to B while avoiding hazards. “The building blocks of driverless cars are on the road now,” explained Russ Rader, senior vice president of communications at the Insurance Institute for Highway Safety.[1]  Front crash prevention systems, self-parking, Drive Pilot (Mercedes-Benz), and blind-spot detection have been in use by car manufacturers for years now.   
 

https://www.ansys.com/blog/challenges-level-5-autonomous-vehicles

The car has sensors that sample its local environment such as cameras and LIDAR that combine with AI systems to enable the car to “see” and “recognize” objects, and also “predict” their movements. The car “knows where it is” based on GPS and a stored map of the roads and it also knows where it is trying to go and the potential routes to get there. All of these functions are combined into a set of next actions and the driving strategy is continuously adjusted as new information about the environment is processed – for example, the car slowing down when it detects a child playing on the side of the road. 

Self-Driving Financial Crime Detection 

It turns out that this overall system design that combines local information sensors, object recognition and prediction with a global map and a continuous strategy adjustment is directly applicable to a wide range of problems, including financial crime. AyasdiAI has built this same design orchestration into its new Financial Crime application – Ayasdi Sensa-NetRevealAML™. The difference is the sensors are tuned to money flow and financial behavior, and the map is a map of all the different customer behaviors that exist in a bank’s customer database. The hazards that we detect and avoid are the criminals.  

In the rest of this blog series, we will explore how this new approach can sit seamlessly alongside existing systems, support regulatory review and achieve breakout performance in accuracy and risk coverage. 

In 2015 machines could “see” better than humans and in 2016 AlphaGo learned to play Go and beat Lee Sedol. By 2018 Waymo launched the world’s first self-driving taxi service, and in 2020 GPT-3 learned the internet. It is clear that the pace of AI innovation is accelerating exponentially, and this is only a glimpse of things to come.  

The question we are asking ourselves at AyasdiAI is “can we build the self-driving car of financial crime detection?” It’s a bold aim, yes. But is it impossible – absolutely not!  

Illicit money flows in ever greater volumes through the global financial system; fraud is the most common crime in the digital world and criminal sophistication outpaces the industry’s response. Anti-money laundering systems are swamped by false positive alerts forcing banks to employ armies of investigators who spend 99% of their time looking at completely normal financial behavior. Surely the AI revolution can help with this challenge? Yes, it can; of course, it can. 

It is not a simple problem to solve though. With human adversaries operating across borders, insider compromise, complex and changing regulations, and legacy systems that need to be maintained and adapted create a complex environment in which to operate. But this is no more complex than building a car that drives itself while following the laws of the road and avoiding crashes yet self-driving cars appear to be light years ahead of the response to financial crime. 

Self-Driving Car Example  

A self-driving car works by combining an array of different AI components into an overall “intelligent” system that can sense, monitor and adapt to its surroundings to safely navigate from A to B while avoiding hazards. “The building blocks of driverless cars are on the road now,” explained Russ Rader, senior vice president of communications at the Insurance Institute for Highway Safety.[1]  Front crash prevention systems, self-parking, Drive Pilot (Mercedes-Benz), and blind-spot detection have been in use by car manufacturers for years now.   
 

https://www.ansys.com/blog/challenges-level-5-autonomous-vehicles

The car has sensors that sample its local environment such as cameras and LIDAR that combine with AI systems to enable the car to “see” and “recognize” objects, and also “predict” their movements. The car “knows where it is” based on GPS and a stored map of the roads and it also knows where it is trying to go and the potential routes to get there. All of these functions are combined into a set of next actions and the driving strategy is continuously adjusted as new information about the environment is processed – for example, the car slowing down when it detects a child playing on the side of the road. 

Self-Driving Financial Crime Detection 

It turns out that this overall system design that combines local information sensors, object recognition and prediction with a global map and a continuous strategy adjustment is directly applicable to a wide range of problems, including financial crime. AyasdiAI has built this same design orchestration into its new Financial Crime application – Ayasdi Sensa-NetRevealAML™. The difference is the sensors are tuned to money flow and financial behavior, and the map is a map of all the different customer behaviors that exist in a bank’s customer database. The hazards that we detect and avoid are the criminals.  

In the rest of this blog series, we will explore how this new approach can sit seamlessly alongside existing systems, support regulatory review and achieve breakout performance in accuracy and risk coverage. 

Latest Insights

How to unleash generative AI’s potential in anti-financial crime
 
03.21.2024 Blog

How to unleash generative AI’s potential in anti-financial crime

Financial Services
Successful enterprise AI doesn’t just require great AI. It demands domain expertise.
 
03.20.2024 Blog

Successful enterprise AI doesn’t just require great AI. It demands domain expertise.

SymphonyAI
retail cpg connected retail
 
03.20.2024 Video

Connected Retail

Retail / CPG