AI in Financial Services: 10 questions you should ask before thinking about leveraging new technologies for AML- Fraud detection
Challenges and opportunities of new technologies
The use of AI in fraud detection has been picking up pace in the past few years due to its ability to detect patterns in data which a human would not be able to see. And with the rise in machine learning and deep learning, these algorithms are going to become much more advanced going forward.
In July 2021, the Financial Action Task Force (FATF) published a report on challenges and opportunities of new technologies for AML/CFT.
In their synthesis, they acknowledged several benefits of new technologies including facilitating data collection, processing and analysis to help supervisors identify, assess, monitor and communicate suspicious transactions more effectively and close to real time.
But they also raised some concerns related to standing operational and regulatory constraints and complexities associated with updating the legacy systems, explainability and interpretability.
According to FTAF, smaller financial institutions in particular often lack internal capacity or confidence to evaluate the effectiveness of a given innovative solution among a large and growing range of competing vendors and products, to determine if it is appropriate for the institution’s risk profile, customer base, and business activities, or to implement models and manage model risk.
Perform a cost analysis
Before starting any projects on new technologies for AML Fraud Detection, financial institutions should perform a cost benefits analysis covering 5 levels
- specific business and risk faced by the organization
- compliance practice : process and governance in place
- IT infrastructure
- Data readiness
- HR
10 key questions
Given our experience in implementing AI MLsolution for AML and fraud detection, here are key questions to kick off this cost benefit analysis:
- What are the current strengths and weaknesses of the current FCC - compliance process ?
- Did specific areas of risk increase in the last months/years ?
- What is the proportion of false positives ?
- How much time is spent by the investigation team on evaluating false positives compared to the high risk transactions ?
- Does a risk based approach segregate the investigation effort against potential ratings ? are you able to implement the risk based approach homogeneously in the company ?
- Regarding your rules engine, is it flexible enough to cover new regulations ? to track new changes in the customer behavior ?
- Did you make an evaluation of the quality and availability of data ? Which effort is required to collect and clean transactions, prospect and client data ?
- What is the ratio of manual data checks vs automated checks ?
- What is the readiness of the team to use new technologies ?
- How do you intend to introduce new technologies ? by adding gradually new components ? by replacing some part of your legacy systems ?
Then based on weakness and strengths observed on the 5 levels come the requirements to design the new solution being data enhancement, redesign of new practices or enhancement of the existing IT systems with new technologies.
Issues related to the costs of new technologies, the ability to ramp up resources on new technologies as well as the cultural shift to move to new technologies are additional items to be taken into consideration.
Towards augmented compliance
As any new technology, Ai should be evaluated from cost benefits perspective and integrated in the compliance process. It is important to understand that AI does not exist in a vacuum : it should streamline the process, improve efficiency and speed, decrease costs and free up capacity. Otherwise he doesn’t play his role towards augmented compliance.
