Context is everything with Voice UX driven by AI

By Subramaniam Vaithiyalingam & Bill Fleming

In the next few years, Zero UI voice-activated customer service will become much more prevalent. It’s not just because the technologies are getting better—it’s largely because the need is being driven by consumer adoption and increased usage of voice-based interfaces.

It was only about a decade ago that effective AI interaction through voice was mostly in the realm of high-tech academic research. Now it’s nearly taken for granted. Voice-activated AI is a growing part of our everyday lives with the likes of Amazon’s Alexa, Apple’s Siri, Google Assistant, Microsoft’s Cortana and hundreds of other players. Increasingly, young consumers and digital natives prefer an anytime/anyplace voice-activated interface.

It’s not hard to build a closed-domain voice-interactive system today, because there are readily available building blocks for speech recognition and natural language understanding. Yet, creating a personalized customer service experience via voice-activated UI is only possible by understanding who the customer is and what their query context is, in order to build conversations that enable solutions driven by AI and Machine Learning.

Context is everything

Voice UX should be based on who is asking the question and the context in which the question is being asked. An example of a complex contextual question is if someone were to verbally ask: “How do I take $5,000 out of my 401k plan?”, the response would need to take into consideration the customer’s age, working status, and the specifics of the particular 401(k) plan that the person is enrolled in. It’s a seemingly simple query, with a not-so-straightforward answer—yet it needs to be answered with accuracy.

Intelligence goes deep

While voice recognition is beginning to mature, there is much work to be done to create building blocks for artificial intelligence. The landscape of interactive intelligent systems is very dynamic, with new algorithms, tools, libraries, services, and startups emerging every day. Though algorithms to train the models are often out-of-the-box, obtaining domain-specific data to train these models is a challenge.

Deep Learning is the epitome of an intelligent system, and organizations with historical and live transactional data have a lot of intelligence with mining potential. Moreover, a deterministic approach to build intelligence, versus probabilistic knowledge-based systems, is more beneficial to ensure accuracy. Huge data, combined with massive computing power and Deep Learning, can build intelligence that can serve investors appropriately.

Yes, voice UX and AI is a complicated matter. But it’s happening and accelerating faster and faster these days. Zero UI voice interfaces are now part of our daily lives, because of the recent advances of Machine Learning. The future of customer service everywhere is about to get more sound.