Speaking at Dreamforce 2018 in San Francisco, Dan O'Connell, general manager of VoiceAI at phone and conference system provider Dialpad, said that two trends are converging to make that scenario a reality:
- "The transition to the cloud is making businesses more productive. 60 percent of enterprises have some kind of cloud technology," said O'Connell.
- "AI is making businesses more effective. We're in the early days, but machine learning (ML) is augmenting tasks already," he said; phase two will come when AI and machine learning are fully automating really boring tasks.
The result will be new opportunities to leverage voice: as an input for all of our devices and as a dataset in itself. “Voice is the last offline data set," said O'Connell, describing it as a “dark" dataset that's mostly inaccessible. Dialpad is a cloud-based platform for sending and receiving phone calls, and in addition to logging calls and managing contacts, it can analyze calls in real time, transcribe them and let the agent annotate them. That can help open up the data contained in calls.
Business-to-customer communication is still 60 percent voice-based, said O'Connell. Given that, how do you use voice interactions to enhance your brand? “Make sure every interaction is positive," he said, and speech recognition and speech-to-text transcription will help.
We don't want to build a better tape recorder. We want to remove the tape recorder.
—Dan O'Connell, general manager of VoiceAI, Dialpad
Doing this properly is about more than just recording conversations, though. “We don't want to build a better tape recorder," said O'Connell. “We want to remove the tape recorder." Recording and analyzing calls takes a lot of time—the goal is real-time understanding.
Thanks to ML and advancements in NLP, O'Connell said, speech recognition is nearing human-level accuracy. That has the potential to change how we work by augmenting and automating interactions. If you can achieve real-time understanding of customer conversations—what their underlying sentiment is, how critical the situation is and so on—you can react faster to ensure a positive interaction.
This approach provides after-the-call benefits as well. Once you have a data set of transcribed and properly characterized interactions, you can start to analyze what your customers are usually calling about. You can track real-time sentiment and eliminate the “I don't know the answer to that, I'll have to get back to you" moments from your call center.