How Conversational AI Can Filter Contact Centre calls of Sensitive Data

FREMONT, CA: People contact customer service centres daily and speak to representatives while giving them sensitive information, such as credit card numbers. There is now a technique to take that information out of calls while still allowing data to get through for transactions, thanks to conversational artificial intelligence (AI) systems that use natural language understanding capabilitie

This is crucial since managing any personally identifiable information (PII) necessitates adhering to numerous security and privacy standards, which can differ depending on the country. Additionally, there is a non-trivial chance that private data could be stolen or disclosed. In reality, there are documented cases of malevolent actors recording verbally supplied credit card information, resulting in bad results.

According to the head of AI research at conversational AI vendor Interactions, there was an incident where an enterprise customer came to with a real-life story saying, this happened, someone noted down the credit card numbers, and those things were leaked in the open market. That prompted us to consider the technology and ways to redact personally identifying information in real-time with minimal delay and without degrading the user experience. To that end, Interactions created a fresh technology called Trustera, which is now generally accessible.

Taking a Hybrid AI approach to Conversational AI

A startup called Interactions creates platforms for conversational AI technologies for businesses. Contrary to popular belief, Interactions has not largely adopted the methodology that conversational AI technology is generally associated with: human interactions with bots. Bangalore said that his business had adopted a hybrid AI strategy.

With the hybrid AI concept, conversational AI is used in conjunction with humans to provide user experience in a frictionless manner. For instance, the Trustera system is not bot-driven and is designed to function in settings where a consumer calls a customer care number and then chats with a human.

Redacting PII during human-led discussions is more difficult than during interactions solely bot- and digital-driven interactions in an interactive voice response (IVR) system. He pointed out that because PII transmission is a necessary step in IVR or both dialogues and is started by the system, the system is aware of when it is happening.

When PII is requested or transferred during human-led talks, it doesn't necessarily happen at the same moment in the dialogue. It's important to comprehend the actual human speaker and the PII being sent. The AT&T Bell Labs skills form the foundation of the AI technology that Interactions has created for its conversational AI platforms.

The on-call model from Interactions has been taught to recognise when various human speakers convey personal information (PII). The model is dynamic and is always being revised. People use a self-supervised auto ML technique, in which they take the calls from the previous day and use a notional confidence metric to suggest that these are data components that they can feed back to the model.