“I’m sorry Dave, I’m afraid I can’t do that.” For decades, that immortal line from 2001: A Space Odyssey was what people imagined voice-powered artificial intelligence (AI) would be like. As technology advanced, we got more sympathetic portrayals of voice-powered AI systems, such as Tony Stark’s JARVIS in The Avengers universe and Samantha from Her.
By Ryan Falkenberg, co-CEO of Clevva
Even though both Samantha and Jarvis were featured in movies that came out in the 2010s, truly capable voice-powered AI still seemed like the stuff of science fiction at the time. Sure, we already had the likes of Apple’s Siri, Google’s AI assistant, and Amazon’s Alexa, but their capabilities were limited, to say the least. And yet here in 2024, voice-powered AI isn’t just a reality but an increasingly valuable business tool.
One area where they’re showing particularly high value is in contact centres. Far from forcing customers to speak slowly and clearly, like older voice recognition systems did, today’s voice-powered AIs can easily handle shifts in conversation and context. This, in turn, allows a virtual agent to carry a conversation almost as naturally as most humans.
Part of the reason for the rapid advances in voice-powered AI is the sheer amount of voice data that the technology has been trained on. Just consider the number of Microsoft Teams meetings that are held every day in multiple languages and with varying accents and you can see why voice AI is improving at an exponential rate.
Voice-powered AI’s rapid advancement has surprised the contact centre industry, mainly because their conversations are usually structured and rule-bound. Unlike conversations with Large Language Models (LLMs) like ChatGPT, which rely on predictive logic, interactions with banks or utility companies must follow specific business, system, and regulatory rules. Additionally, these conversations must be compliant, ask the right questions, provide the right answers, and trigger the correct actions.
And that is where most voice bots have failed. They lack the agility to adjust to every customer, and feel almost menu-driven and wooden in their engagement. The same can be said for chat conversations via the website or WhatsApp.
Automating outbound sales conversations is particularly challenging. People often hang up immediately when cold-called by a robotic virtual agent. Even when seeking resolution, tolerance for an automated system that fails to understand or resolve issues promptly is low. This is why many digital engagements end with customers requesting to speak to a human agent.
Fortunately, we can now change this experience. Virtual agents powered by a blend of Conversational Process Automation and Generative AI technology allow us to automate huge volumes of chat and voice calls at the level of a human expert in a consistent, compliant and hyper-personalised way.
Recent advances should lead to high voice automation adoption in contact centres over the next 18-24 months. As we perfect multilingual conversations with better tone and accent accuracy and resolve integration challenges with older systems, more conversations across chat, email, and voice channels will be automated without the need for a human.
The business case for a virtual agent is significant. It not only offers customers immediate access to a customer service expert across multiple channels, 24/7/365, it also allows companies to stabilise and specialise their human agent teams. They can now focus their efforts on the high impact, lower volume calls that really matter.
Virtual agents can be used as the first point of contact for customers, looking to clarify what customers need and channelling them to the right support. Over time, virtual agents will resolve more of the query types themselves, ensuring the calls that do reach live agents require their sense of empathy, care, humour and human touch.
Right now, effectively servicing each customer with a trained and available human agent is prohibitively expensive for most businesses. The sheer financial business case of getting voice-powered virtual agents right will overcome any problems associated with adopting the technology. The technology is largely ready. All that’s left for it to achieve its full potential is greater adoption.