The Journey to Autonomous Fraud Management
December 21st, 2020
This blog series acts as a companion to the sessions at ENGAGE LIVE. Catch up on the other entries here:
- Looking Back at ENGAGE 2020
- Immediate Action Required: Detecting Authorised Fraud
- Stay Ahead of First-Party Fraud & Mule Activity
- Catch Me if You Can: Fraud Digital Identity Challenges
- Future-Proofing Fraud with Advanced Technologies
- Fraud: AI in Action
As we’ve seen across the ENGAGE blog series, there are a number of challenges facing financial institutions.
We’ve talked about the explosion of data, the challenges of digital identity, changing fraud typologies with rise in authorised fraud such as business email compromise (BEC), increasing hidden fraud from first-party fraud, and the mule accounts that underpin many frauds.
We’ve also seen how new technologies, such as Kafka and cloud, along with improvements in AI can help financial services organizations (FSOs) reduce fraud and costs, all while reducing friction.
Bringing this all together, we can move to Autonomous Fraud Management.
What do we mean by Autonomous Fraud Management?
This is really about a paradigm shift from machine assisted humans to human assisted machines. Let’s unpack this to see what is really required here.
At its simplest, we are talking about making the whole process more efficient by using machines and people to do what they are best at. People are best when talking to and understanding customers or investigating complex fraud cases, and machines excel in dealing with large amounts of data and moving it around.
Why is Autonomous Fraud Management Necessary?
With the rising number of channels and increasing use of digital by consumers, we’re seeing more and more transactions. When you see any increase in transactions, cause and effect dictates that the number of alerts increases as well. This unfortunately just can’t scale as we are, because FSOs cannot justify the cost it would take to have X number of people work Y amount of alerts.
Managing a large quantity of alerts is not most peoples’ strength. You want to have your people talking to customers who are victims of complex fraud, getting them back up and running or investigating complex cases, not simply pressing buttons to refund $100 and order a new card.
How do you get to Autonomous Fraud Management?
Start by building a roadmap of where you want to go. This should cover the fraud risks and threats you see and expect to see along with the operational elements desired. Operationally looking at how increased automation in case management and cloud-based infrastructure can lead to easier deployments and lower cost YoY.
Removing the silos built up is critical as frauds are increasingly linked. You should be looking at holistic enterprise fraud management to view activity across the lifecycle, payment rail and channel. One area that can help is to link this to ISO20022. Building out a hub that makes the most of the structured data from ISO20022 and can still take legacy rails too is a best practice when transitioning into modern digital banking.
Moving on from here, we need to look at building out data-driven decisioning capabilities. Utilise AI and ML to not just build out better models that are refreshed regularly, but to make the move to self-learning. Self-learning models that can pick up new fraud trends and the ability to identify new rules to implement means less human input with improved fraud detection.
To make the most of this requires the kind of elastic and scalable infrastructure offered by the cloud to support using unsupervised ML, as well as supervised, to help deliver models faster. If low volume, high value frauds are an issue, then this should also cover how the models can learn from other FSOs with collective intelligence and federated learning.
If we then move into how alerts are managed, we’re talking about increased self service. Better authentication steps means removing people where possible.
Automate fraud refund claims in areas such as cards, where these are high volume, low value processes. Refund a new card order and replenish on xPay, all without an agent for the majority of cases.
This can include ingesting alerts from other detection systems and provides automation to reduce errors and the time to manage cases, allowing skilled staff to focus on investigating increasingly complex cases, as well as working with customers.
Increasing the throughput of alerts means we can deal with more. This can also include different strategies based on the type of payment and authentication. For example, xPay, when onboarded correctly, offers lower risk on e-commerce. This might also be the case where liability and recovery processes have been enhanced. However, if you have ways to reduce chargebacks and losses, then you can take different strategies. For example, with airline tickets, which tend to have high false positive rates, if you know you can use a service to get the refund without a chargeback by notifying the retailer before the fraudster flies, you can decline less by messaging the customer after it has gone through to generate a fraud alert post transaction. This works because in many cases, the ticket is not used for several days.
This relies on processing more data via machine learning, both structured and unstructured. Bring in multiple end point protections, profiling types and data validation that we have covered already in previous blogs.
It’s not just about reducing the fraud; this all improves the customer experience.
In summary, by using automation and making many elements autonomous, we can transform fraud prevention and operations. Dealing with the big challenges around the volume of data, speed of change and increasing customer demands for digital, while reducing fraud, is key.
Agents can spend their time talking to the customer to get them back up and running and share how to protect themselves going forward. As basic tasks are automated, errors and complaints are reduced and the time to resolve each case decreases, increasing efficiency.
Staff are happier, more engaged with customers, and can work more alerts per day. An effective staff leads to increased fraud prevention, high customer service ratings, and victims of fraud have a good experience from a bank resolving their problem. These happy customers tell their friends, which aids NPS in the long run.