A new approach to contact centres which utilises “natural language understanding” is forcing a rethink of traditional contact centre analytics, according to a new whitepaper from Which-50 and Genesys.
Traditionally, contact centres measure the number and duration of calls as a key metric, with an eye to getting more customers’ calls completed as quickly as possible. But that’s not necessarily the best approach, according to a new whitepaper, How Conversational AI Improves Customer Service.
The research explores how natural language understanding – an emerging technology that converts customers’ conversations with voicebots to text – is uncovering new insights about customer sentiment which may otherwise be lost.
Sharon Melamed, founder of Matchboard, an independent marketplace for contact centre and technology-based service solutions, says some voicebots conversations are opening up new data sources.
“Companies might never hear from some customers who don’t like waiting in a queue to speak to a human call centre agent,” she says. “Voicebots give people a new way to engage with brands where they reveal their issues, frustrations, likes and dislikes. These insights help companies identify problems, fix broken processes, and even understand what is making customers angry, as voicebots can detect raised voices and emotions”.
Importantly for businesses, this information can be combined with analytics to better understand the cost or revenue impact of addressing specific issues they may never have learned about without the voicebot.
Dave Flanagan Director, Digital & Conversational AI – ANZ region at Genesys says customer experience and operational efficiencies are what organisations should have in mind when looking at contact centre and voicebot analytics.
“Intent matching is kind of an obvious one. How often can the voicebot successfully handle the customer request? This capability will grow over time, but you need to ensure that your bot has a solid foundation to begin with,” he says.
The way to do this, according to the whitepaper, is to focus on a handful of critical use cases that add real value to the customer. Provide your bot rich data around this core capability, ensuring it does a few things well. Once there is the confidence that the bot can handle the majority of user expressions, organisations can then look to develop new capabilities. Another key metric identified in the research is confusion triggers. How often does a bot not understand the user’s request, and how does it respond when this inevitably occurs?
“It’s important to monitor user engagement and to identify bottlenecks within the conversation that result in errors, abandonment, or escalation to a human agent,” Flanagan says.
“This is how you can really drive increased self-service.”