Have you ever been on a conference call furiously taking notes only to forget what these notes actually meant in the context of the call?
As keyboard and computer warriors, we’ve all been there. And after years of conference calls running Mutual Mobile, working in real estate and angel investing, Sam Gaddis was getting close to his breaking point as he kept missing valuable information on those calls.
“When I was running the marketing team in mobile, you might have five calls a day and several things happen," he said. "By the end of the day, your brain is fried, and you might miss something in between taking detailed notes."
He was trying to figure out ways to automatically transcribe calls. But what really pushed him over the edge was an episode of "Black Mirror."
“There's just one episode where it's this future where these contacts record everything... that happens in their lives and they use a mechanism to go back and see exactly what happened,” Gaddis said. “So you can imagine all the horrible ways that couples fight when they can go back and say 'I actually said this.'”
While Gaddis didn’t want any part of that "Black Mirror" episode, he was interested in what could happen if this type of technology was restricted to primarily audio.
How Jog.ai works
The voice technology behind Jog.ai runs on something called acoustic diarization. It is a software algorithm that looks at differences in pitch and tone and tenor of the voices on the recording and then tries to guess who is speaking. It's essentially an extra layer on top of a basic Google text transcribing service, where if you tried to analyze words, it looks more like a screenplay rather than a whole block of text.
But that’s just one bit of the technology. Gaddis had some disheartening moments early on, where the software sometimes struggled to pick up who was actually speaking at a particular moment in time.
“There are two main things we focus on with the transcription engine: does it accurately guess the words that are being spoken, and does it accurately guess who is speaking those words?” Gaddis said.
With feedback in hand, Gaddis and his team set out to create an interface applicable to everyone. But there were more aspects of the technology that needed to be cleaned up. They wanted 100 percent speaker splitting accuracy with the conference calls, or else, as Gaddis put it, they were going to have to shut it down.
“Essentially, what we're doing is taking individual audio from each side of conversation and running that through a transcription engine independently,” Gaddis said.
“From this call, I'll have a portion of the call that's just you and not me speaking, and its the same thing for me not speaking," he said. "We build around each of the transcription engines and then do some pretty intense math to chunk it out so that it does read like a screenplay rather than a wall of text.”
But what if you hear something interesting on the call? How can you revert back to it? You can “flag” a piece of the conversation by clicking the space bar. The flags show up on the right hand side of your Jog.ai screen.
Who Uses Jog.ai
There are plenty of people who are using the platform.
"One thing I hear from clients is this is a really great feature for those people who are running through an airport but their client is saying something important,” Gaddis said.
Journalists have been another group using Jog.ai. Gaddis finds that many are using Jog.ai for test cases on over-the-phone interviews, but that the people that are having the most success from the software are those in any sort of client service roles who are doing consultancy work and employees in digital marketing agencies.
While it may not be for the traditional transactional sales rep, those who work in channel sales, where you may not talk to a client for six months, find benefits in seeing the strategic nature of each call and remembering personal details that they may have forgotten otherwise.
Gaddis also noted that real estate agents and hiring managers make up a growing portion of the user base.
“Those are pretty different use cases and Jog seems to be working for them,” he said.
The Future
While Jog.ai has a variety of features to work on, one interesting aspect of their roadmap pertains to an integration, which many of their customers have been inquiring about. With this integration someone could send data from Jog to any service, whether that’s Evernote, Salesforce or Google.
In addition, the company is focused on using Jog.ai as a storytelling mechanism, and also a teaching tool. Gaddis recalled a story where a customer was getting slammed by his competitors, and was trying to figure out how to better position his company. The user used the tagging tool as a way to mark customer feedback when they mention their competitors.
“He was able to show his boss that it wasn’t just him,” Gaddis said. “Here's the customer saying exactly what you can hear their voice and you can really understand what he's trying to communicate.”
For Gaddis, a feature to consolidate that sort of feedback and share it would be something interesting for the platform down the line.
Jog.ai is free to try, with a 14-day free trial. Currently English is the only language supported, but more are on the way in 2018.
