Call center agents have to do more than just hear what a customer is saying if they want to deliver superior customer service -- they also have to understand how something is being said. Cogito Corp., a voice analytics software company spun out of MIT, wants to help with that.
Cogito's artificial intelligence technology listens for conversational behavior cues, such as tone, conversational flow, interruptions and pauses, and gives the call agent real-time coaching advice. Ali Azarbayejani, CTO at Cogito, called the technology "augmented intelligence." He said it's based on the evolutionary biology theory that, before language evolved, humans and primates communicated with each other using nonverbal cues.
"When you're communicating with someone, signals -- about your state of mind and your emotions -- are transmitted through your behavior," he said. "We've taken that and turned it into a technology for the call center."
At the recent MIT Sloan CIO Symposium, Azarbayejani spoke to SearchCIO about the value of voice analytics software and why augmented intelligence is better than automated intelligence in the call center. He also discussed the challenge of striking the right tone with the software and the challenge of making the software sound like a coach, not a critic.
Are your customers asking how to measure the ROI on AI investments like Cogito?
Ali Azarbayejani: Some of our customers use NPS [net promoter score] as an indicator. And what we've done with our technology is we've applied it with control groups. Some agents have the intervention and some agents don't have the intervention -- the intervention being a real-time coaching system that helps agents in their conversations. And our customers are finding 10% to 20% increases in NPS of the intervening group with respect to the control group.
There's a lot of ROI that's just evident in NPS and customer satisfaction increases that you see.
Are these metrics the same metrics they would be using otherwise, or are they new metrics used to measure the value of Cogito's voice analytics software?
Azarbayejani: Both. NPS is an industry standard measurement that everybody uses. Companies use it to evaluate whether the technology is helping. We also provide our own conversational experience score. So that's another metric that they can use, and that's based on our behavioral analytics.
Are you a proponent of augmentation or automation?
Azarbayejani: In our industry there are a lot of efforts to build digital agents like chatbots with the idea of displacing contact center workers and automating those interactions. But what we find, and I think what our customers find, is that there's a certain amount of low-hanging fruit that can be plucked by these technologies, but by and large, people still need human interactions to solve complicated problems. Our technology is more like an augmented intelligence that understands what's going on in a conversation and helps the agents and helps the call center have better human interaction.
How does this augmented intelligence work?
Azarbayejani: We provide metrics in a couple of different forms. I mentioned to you our customer experience scores. These scores are a little more nuanced, and they're statistical scores. We generally don't present those scores to agents because it would be too difficult for them to unravel and understand, statistically, what they mean. Those scores are presented, typically in aggregate, to supervisors and call center operations managers, who use them to figure out strategies for improving operations in the call center.
We have a different set of metrics, which are very actionable coaching advice that we deliver directly to the agents in real time. Things like, 'You're interrupting,' or 'You're talking too fast.' Very concrete, actionable things that are immediately verifiable whether they're correct or not and that can be acted on directly.
How does the technology strike the right balance between critic and partner?
Azarbayejani: That's a real challenge. And it's really a UX/UI [user experience/user interface] challenge having to do with the persona of the software. So it can feel nagging if you do it wrong, and our goal is to make it feel more like a coach or a therapist.
Does the software, then, also deliver positive, encouraging messages to the employee?
Azarbayejani: That's another topic of UX/UI that we get into a lot. As it turns out, during a call, positive feedback is not as appreciated because you have to divert attention to see what's going on. So agents and supervisors mostly want to see things that can actually improve the conversation. But between calls, we're really focused on giving positive feedback when agents are improving or they're having good conversations.