Experts are predicting that within the next 20 years, automation could replace up to 40% of the global workforce. While these discussions typically revolve around more traditionally labour-intensive jobs (like shipping and manufacturing), some are wondering if artificial intelligence and machine learning could become so efficient as to replace fraud management teams in the near future.
Can AI really make humans obsolete in the world of fraud? This is the question that has left some industry experts scratching their heads recently. I want to dive deeper into the topic and look at the current state of AI in the industry and how it can best be leveraged to make fraud management more seamless, boost top-line revenue, and reduce customer friction.
The current role AI is playing
Many fraud solutions today leverage machine learning and artificial intelligence in one way or another. Retailers, financial institutions, and creditors often rely on artificial intelligence to create rules for their transactions. At a basic level, these rules are used to automatically decline or approve transactions or applications based on an analysis of certain data points such as email address, name, phone number, or IP address.
Alternatively, companies may utilize AI to flag suspicious transactions and send them to manual review teams within their customer support or loss prevention departments. These teams use the data gathered during the transaction and flags created by machine learning rules to assess whether or not a transaction is fraudulent. Manual review is not only the most expensive aspect of fraud management but also the most subjective. Agents responsible for these reviews rely on their best judgment alongside any data analysis they have access to, in order to make a decision about whether or not to approve a transaction.
These jobs, ultimately, are the ones that some believe automation, machine learning, and artificial intelligence will replace. Studies show up to 82% of transactions in manual review queues are approved, which opens questions about whether or not they should have been reviewed at all. Could AI have done a better, more efficient job of approving these transactions?
Maybe, but…
There is no easy answer to whether or not AI will overpower human jobs in general, but when it comes to fraud management, the answer is probably no. At least not yet. It’s true that AI can and does improve outcomes and efficiency across nearly every industry, but it does not completely eliminate the need for manual reviews or customer support agents that collect additional information before processing applications and transactions.
A key piece of the manual review process that many machine learning proponents miss is the customer service and anomaly detection aspect. AI is only as intelligent as what it has seen previously, meaning these models can detect unusual activity or data outliers, but they cannot explain them. This is one major way humans are still critical to the decisioning process. At Emailage, our decision scientists review shared fraud data in real time to better customize fraud models by analyzing data outliers and applying rules to them across industries.
Another key role that humans play in the fraud management process is in gathering supporting data. Transactions and applications are more than just points of data, they represent human beings attempting to purchase something or gain credit. Humans make mistakes, which AI cannot yet account for. Typos, misspelled or mismatched names, transposed digits in phone numbers could all lead to a risk assessment engine flagging a legitimate customer for manual review causing friction and abandonment. Manual review teams are able to reach directly out to customers to obtain corrected information and update transactions on the fly, reducing friction and stopping loss by making customers feel valued and protected. AI simply cannot replace the human touch.
Conclusion
AI can reduce manual reviews by as much as 50%, improving fraud teams’ productivity and reducing customer friction as well as lost revenue. It’s an amazing tool that should be a part of every holistic fraud management strategy, but it does not outright replace a manual review team or fraud analysts who manage the data needed to improve machine learning. While automation may make human roles in other industries obsolete, they will continue to play an integral part in fraud management for the foreseeable future. Fraud and risk managers are responsible for mitigating risk as much as possible. The right people in the right roles providing a human touch is the only way to successfully get the job done.
About Amador Testa
Amador is Chief Product Officer at Emailage. He is an industry expert in online fraud, identity theft and cybercrime. Before Emailage, he was the head of fraud for card acquisitions at American Express and later led global fraud prevention divisions at Citigroup. Amador enjoys playing tennis, running marathons and traveling with his family.
About Emailage
Emailage, founded in 2012 and with offices across the globe, is a leader in helping companies significantly reduce online fraud. Through key partnerships, proprietary data, and machine-learning technology, Emailage builds a multi-dimensional profile associated with a customer’s email address and renders a predictive risk score. Customers realize significant savings from identifying and stopping fraudulent transactions.
Every day we send out a free e-mail with the most important headlines of the last 24 hours.
Subscribe now