Aaron Marsh 

Hello, and welcome to the bugs of bank automation news podcast. I’m Associate Editor Aaron Marsh. And I recently had the chance to speak with Rana, good Raul, CEO of behavioral signals, which has automated tone detection, that is detection of emotion and intention in customer service calls. And it’s so far targeting its product at banks and financial institutions. I caught up with Rana to hear about the technology and how effective it’s been. You guys are automating some different things with with with voice work, not necessarily recording voice, but recording recording like or an automating the recording and processing of tone of voice. Is that right?

Rana Gujral

It’s definitely a little bit more about our tech and what we’re doing with that tech, in terms of business use case. So we, at the core were deep tech AI company that was spun out of USC, the early AI research and the models. And the breakthrough that our founders and researchers focused on was all within the umbrella of signal analysis and interaction labs, sale for short, which is, which is part of USC and Stanford. And one of the founders is actually also an executive director and founder of sale.

And the focus back in the day was to extract a variety of intelligent signals from the acoustics of a conversation, which is the total variance in donations, pitch and tonality, prosody, etc. So, now, the way to think about it is that in our conversation, there are two big elements. One is the spoken word, and then everything else, which is behind the spoken word, or beyond the spoken word, which is what we call the acoustics, the pitch and tonal variants, etc. And so what we do is, we focus just on the acoustics, we don’t process the spoken word, and from the acoustics, we extract a variety of signals live real time, in the moment.

And these signals can be put into three different buckets, I mean, the first would be emotions, like anger, happiness, sad, sadness, etc. And then there will be signals that represent behaviors such as engagement, empathy, politeness, and so on. And the third group or bucket would be advanced signals that are more, you know, more sort of in line with an industry KPIs such as such, as, you know, scoring a customer’s satisfaction on that call or an engagement meter for the agent, or predicting churn or propensity signals have a certain action that we can predict, for example, will the client buy or not buy? Or will the debt holder pay or not pay? So all the signals we can derive, from what we process, the acoustics of the conversation. So that’s, that’s a core tag. I mean, we’re, we’re one of the first to mark it in this field. I mean, the one of the ways to describe this in this sector is behavioral signal processing BSB. And we really sort of pioneered this arena.

And the engines that we built, have won many awards, at his speech at other major events and industry benchmarks, and currently perform better than anything else in the world today. I mean, for example, one, one way to self measure, the performance of a deep Tech engine is using something called an F score, which is really a combination of precision and recall. And our F scores are always measured, in comparison to how a human escort would look like. So humans ability to extract the signals from the tone of voice is about point eight, that’s the F score. And our engines are performing right now at point nine, two. So it’s actually much better than average human in terms of the ability to extract the signals. Just the core technology. Yeah, I mean, like another way, another sort of bedrock there is like we can in a conversation between an agent and a client can bring date if the client will buy or not buy within the first 20 seconds of a call. And that predictions like between 80 to 85% accurate on a given day.

This is done without necessarily processing any spoken word. So that’s the core technology. Now, one of the things we’re doing with that core tech is we’ve built a very special product towards the goal of maximizing the effectiveness of a conversation that needs to happen. And that’s a product that we call AI mediated conversations or AMC. And what it does is it it takes these engines that can extract the signals, and creates a behavioral profile of all the participants in the conversation, which is typically an agent and a client, or there could be multiple clients on an agent review multibody conversation. And these profiles are like conversational bio friends, it’s essentially it’s, it’s codifying the unique converse, converse conversing style of an individual.

And it’s a factor of many different attributes about 75 or so that range from how fast you speak to the amount of energy you exude to various emotions and behaviors represent during a conversation. And when you understand that, and when you have that profile, then you can identify who are good matches for that person’s profile. From a conversation standpoint, it’s not a personality match or something else, it’s more towards, we’re going to be having a conversation are we going to be struggling to have a good conversation or is going to be flowy clicky conversation just because we are matching.

And that’s what that engine makes happen. And it’s, the system has predefined matches based on previous interactions, often agent and the client, and then when a next conversation needs to happen, whether it’s an inbound or an outbound, then it’s done. Using those matches, inbound, you know, you’re redirected to an agent that already has been pre identified for you as your best match. And if it’s not available, then you go to the second and the third, and fourth, so on. And when you do that, you’re significantly improving the dynamics of that engagement or that conversation and you have amazing results come out of it.

So for example, we improve the revenue recovery or collections, or also sales anywhere from 12 to 18%. By deploying our solution, we also significantly improved customer satisfaction, we also give back on other contact center metrics, such as reducing the average handle time, improving the percentage, of course call resolution, improving engagement, engagement, and morale, all of those things get positively impacted. And so that’s, that’s been our primary focus as a company. That product is largely being sold to financial institutions, so banks and other collection houses and BPOS or other financial, you know, intermediaries that, that have, you know, the need to improve with these KPIs and improve the dynamics of the customer engagement. And of course, collections is a very important use case for us because it certainly helps with the collection metrics.

Aaron Marsh
You’ve been listening to “The Buzz,” a Bank Automation News podcast. Thanks for your time, and be sure to visit us at BankAutomationNews.com for more automation news in financial services. Please don’t hesitate to rate this podcast on your podcast platform of choice.