In the last two years, AI and machine learning have evolved well beyond the most ambitious expectations. With tech giants like Google, Uber, and Tesla (to name a few) continuing to accomplish increasingly impressive feats, the number of AI and machine learning applications seems limitless.
When you consider emerging trends such as “unsupervised learning,” where machines actually learn from data without the need for human input, it’s clear that we’re on the cusp of a full-fledged AI revolution that’s bound to shake up many industries, including fraud prevention.
For payment decision makers, effectively detecting and mitigating financial crime in the digital age is an especially pertinent issue - one that AI can tackle. To deliver the most impact in fighting financial crime in 2019, AI platforms should deliver on the following four advancements.
In 2019, intelligent automation is key
As shortage of data scientists continues to grow, automation of AI and machine learning processes will prove to be invaluable for organizations’ role in the fight against financial crime. However, its important to distinguish between simple automation, such as a computer program that repeats predefined tasks (think Adobe Photoshop’s ability to automate a specific predefined workflow), and intelligent automation where computer programs can adjust their strategies and understand data in context to complete more complex tasks (think AI that can adjust its strategies to beat professional video game players not once, but 10 times in a row). This will greatly increase working efficiency, making 2019 the year of intelligent automation.
Advancements in automation will increasingly continue to free up data scientists from repetitive tasks, such as AutoML, which automates much of the data science workflow and frees them from time-intensive tasks. While some major advancements, such as auto-data cleaning, have yet to be perfected, platforms like Feedzai, which are capable of automating many labour-intensive portions of the data science process, will prove to be an important asset.
In 2019, AI and ML decisions need to be explainable
In 2019, AI is expected to make significant strides into the anti-money laundering (AML) space, fueled by incentives from regulatory bodies such as the US Treasury Department. Understanding the logic used by AI to arrive at a certain decision is paramount in the AML use-case:
with the increasing regulatory scrutiny, the ability to explain AI’s logic behind its scoring method is now paramount to avoid corporate liability;
in a recent McKinsey article, the authors show how explainable AI offers opportunity in AML to restructure traditionally duplicative structures in order to increase operational efficiency;
fraud analysts gain more confidence in their own decisions when they oversee an AI-based model and evaluate its justification for blocking a transaction.
The importance of explainable AI doesn’t end with AML. Explainable decisioning also gives fraud analysts more insight into the logic behind why a transaction is or isn’t blocked. This allows for increased traceability and transparency, enabling analysts to see the logic for themselves and make decisions with more ease and certainty.
In 2019, data is the differentiator
Investment into the AI and machine learning space is at its highest levels in history, according to IDC. With great leaps forward coming seemingly daily, how does a platform leverage these advancements in order to retain its position at the top of its field? Open platforms that support outside AI and machine learning models are a great answer, but that means that algorithms and models are no longer the core differentiator. Instead, in 2019 we will see data as the key differentiator in the fight against fraud and financial crime.
Financial institutions and merchants often have the same two problems:
Not enough data sharing across internal teams
Lack of data from external sources
While many organizations are breaking down silos and sharing data between internal teams, gaining a holistic view of financial crime via access to outside data has proven to be difficult. Solving for this is especially important for FIs and merchants who need to build increasingly-robust data profiles. Consortium options offer a holistic view of an entity, which greatly increases risk identification accuracy.
In 2019, we are expecting to see more AI and machine learning platforms that provide consortium options powered by data from the entire payments ecosystem.
In 2019, deep learning moves to the forefront of the AI arms race
First, what exactly is deep learning? While machine learning relies on heavy feature engineering, deep learning automates that process:
Feature engineering is automated by breaking down complex concepts by defining them as the aggregate of smaller, simpler concepts (ie breaking a human face down by identifying the eyes, ears and mouth) without the need of a data scientist identifying the features for the machine.
So what does this mean for anti-financial crime efforts as we move forward in 2019? Most importantly, deep learning greatly decreases the resource demands that feature engineering brings with it. The feature engineering process often requires both time and expertise, thus making it a heavy resource sink. Deep learning can completely automate that process by breaking down fraudulent transactions and identifying patterns that might not even be visible to a data scientist’s eye.
Deep learning provides extraordinary potential for the future development of anti-financial crime platforms and we will only see the beginning of its true potential in 2019.
The Future of Fraud, ML and AI
ML and AI have come a long way in the past few years, and the industry is advancing quickly. With more data to crunch than ever before and more regulatory scrutiny facing companies that own and process it, there are certainly some unique challenges coming in 2019.
Fortunately, recent advancements in ML technologies, particularly as they pertain to risk management, have already proven that fraudsters have a tough road ahead of them, as well. New AI-driven approaches to risk detection, pattern recognition and task automation are changing the fraud game as we know it. As time goes on, we’re eager to see what other applications will arise.
About Saurabh Bajaj
Saurabh Bajaj is Senior Vice President and Head of Product at Feedzai, where he leads product and strategy for all Feedzai Products. Prior to Feedzai, Saurabh was head of fraud and risk products at Neustar, a leading identity resolution company. Saurabh holds a Bachelor of Engineering from the University of Mumbai and an MBA from the University of California, San Diego.
About Feedzai
Feedzai is the market leader in fighting fraud with AI. We’re coding the future of commerce with today’s most advanced risk management platform powered by big data and machine learning. Founded and developed by data scientists and aerospace engineers, Feedzai has one mission: to make banking and commerce safe. The world’s largest banks, processors, and retailers use Feedzai’s fraud prevention and anti-money laundering products to manage risk while improving customer experience.
Every day we send out a free e-mail with the most important headlines of the last 24 hours.
Subscribe now