Contact centers can have a difficult time staffing agents and adopting modern technology to help them provide better CX. Call volumes and customer expectations continue rising, and compliance requirements are ever-more challenging. All of this can be managed with a good plan, the right technology partners, and an honest effort to do right by the customer – and your agents.

As noble as all this sounds, it’s not enough to address one of the darker sides of today’s technology, namely telecom fraud. This activity is far more prevalent in the world of telemarketing, but it also permeates the contact center in ways that may not be apparent. According to the 2022 Omnichannel Authentication Survey from Neustar, more and more fraudsters are targeting agent-led authentication methods, leading to a $5.8 billion increase in customer fraud losses in 2021, which is a 70% increase from 2020.

Telecom-based fraud takes a mind-boggling variety of forms that would require many posts to catalog. While only some of it applies to the contact center, we’ve all experiences multiple types of fraud as consumers.

Telecom-based fraud is painful and causes customers and organizations a lot of friction, but it’s also not going anywhere. It’s important that contact center leaders acknowledge this reality and try to manage it as best as possible. This post focuses on inbound fraud, where bad actors are calling in under the guise of a real customer for some form of illicit activity.

Fraudsters have many ways to steal personal information, either directly or via dark web sources. Even with just a few simple pieces of ID, they can easily dupe seasoned agents into doing their bidding.

On the surface, these “customer” queries seem like reasonable interactions, but unless vigilant measures are taken, they will keep occurring, and eventually the cost to your organization will become significant. There are a multitude of approaches contact centers can take to improve their security posture, and artificial intelligence (AI) is now playing a larger role. Here are two use cases where AI can help mitigate – and even prevent – this particular type of fraud in the contact center.

AI Fraud Use Case 1: Speech Biometrics

Voiceprint has long been used to authenticate callers and bypass other forms of screening, so it’s especially helpful in quickly connecting customers to agents. This tool isn’t AI-based and because it requires an opt-in process to set up, most customers don’t bother. With AI, biometrics can be used to build profiles of customers based on previous phone interactions, for both speech and voice recognition.

Agents will not typically know the customer from the sound of their voice, but AI tools can match up a current call in real-time with the customer’s profile. This use case is about speaker recognition, where speech biometrics can accurately validate the caller’s identity. The caller isn’t aware this is happening in the background, and if an anomaly is detected, the agent and supervisor will get real-time alerts.

So, regardless of how legit the caller may otherwise appear, this detection could trigger requests for other forms of customer identification and the fraudster will likely end the call at that point. The key here is AI’s ability to process speech biometrics in the moment, and avert fraud that the agent would not have been aware of.

AI Fraud Use Case 2: Pattern Recognition

Speech recognition is about language versus the speaker’s voice, and uses AI-powered machine learning to detect. Fraudsters have varying degrees of sophistication, but they generally follow a tightly scripted approach.

When fraudsters have obtained a list of your customers, complete with personal ID, they will try schemes with the entire list. Some will try the same scheme every time, and more sophisticated operators will try different schemes to avoid detection. So, one might call in attempting to make a purchase, another may call seeking a credit or refund, and others may call about a special bonus or offer that you’re promoting.

The role of machine learning is to work from call recordings, where patterns can be identified in real-time. As with the first use case, alerts can be given where a fraudulent scheme has been detected. As noted, fraudsters will follow their scripts, and while the first few rounds may have been successful for them, the contact center will now have the blueprint for the language and phrases used.

The key takeaway here is these AI capabilities are not supported in legacy contact centers. If this type of fraud has been on the rise in your contact center, you may have just found your best reason to modernize and evolve your contact center and adopt AI.