Latest / The Tech Matchmaker / How Krisp Is Changing Multilingual Contact Centers
Transcript
- Gina Gibson: Hello everybody. Today we are chatting with Robert Maute Director of Channel Partnerships from Krisp Robert, welcome back to The Tech Matchmaker. super excited to hear what's new with Krisp. ⁓ So for those that missed your last episode, can you give just a little bit of a background about yourself for us? Bob Maute: Sure, Gina It's great to be here. Thank you for having me back. ⁓ yeah, so so I have I'm a tech junkie. I have been in this space forever, moving and shaking to whatever that next technology is. I love the challenge. I love things that are new. ⁓ and been this for many years ⁓ and joined Krisp ⁓ about a year and a half ago. ⁓ I had done work with them in previous years in a previous life when they were a supplier in our portfolio. So I got to peel the onion back before I really went there. It was pretty neat. I saw them from you know from the from the distributor side and then got to work directly with them. Yeah. Gina Gibson: Excellent. And just refresh everybody's memory. What what is the elevator pitch for Krisp Bob Maute: Yeah. So we focus on allowing ⁓ you know contact centers to operate globally. You don't really have to worry about with Krisp you don't have to worry about where you're serving your customer from, right? ⁓ we basically knock down communication barriers that cause frustration on service calls. So your agents and your customer have a better experience. So we do that with noise cancellation as our primary product. It's bi directional. We also have voice isolation. So if there are multiple people speaking in the room, we can isolate just on that one voice. And that alone has significant impacts in KPIs. And then the next product that we sort of launched was accent conversion. So we can take a heavy accent and we can convert that into a North American accent, a UK English accent, and coming soon an Australian accent. So we can serve multiple markets for you know outsourced. contact center agents that may operate in the in in India, the Philippines, Latin America, South Africa. We can allow them to serve those domestic North American markets, UK markets, and the Australian markets. So we do that, yeah, and we do that all without changing any of the customers' infrastructure or the process. This is a simple application that installs on the user's device. Gina Gibson: Excellent. Bob Maute: And all of those services are available in one application that can be easily managed by the IT team. Gina Gibson: And and I'm gonna give you a plug for non contact center because I use Krisp for the noise canceling as well and it works phenomenally. So, ⁓ absolutely. And and I know we had you on the show last November, I believe it was, and we gotta look at what you do, like you were saying, the noise canceling and and that. but what's new? Because I know you got some new stuff coming out or or that's out already and and that is really impressive. And I you alluded to it a little bit there, but Bob Maute: Yeah. Mm-hmm. Gina Gibson: should we take a look? Bob Maute: Yeah, I'll I'll talk quickly about ⁓ two things. ⁓ the first thing that we're really looking at is we're starting to get into ⁓ the tools that allow ⁓ agents to perform better outside of just speech. So agent assist, ⁓ knowledge chat, and other services like that. And then also supervisor assist. How do we score the call with the agents? And because we do that all on device. Nothing is saved. We do it right there. It's highly secure. Like a an enterprise company doesn't have to give up their call recordings in order to understand how the agents are performing. That's coming down the pipe. ⁓ you know, we're we're we're there right now. We're we're trialing and working with several ⁓ BPOs and customers today. But where we're really sort of hitting ⁓ hitting the ball out of the park right now is with the voice translation. ⁓ it what was a, you know. a new product ⁓ you know sixteen months ago is now being launched into enterprise customers that are highly you know highly ⁓ concerned about compliance and safety and we're basically replacing language line we're replacing ⁓ transperfect we're basically ⁓ also filling the multilingual needs for contact centers. And it is it's synchronous. So one to two seconds into that conversation, okay, the AI is gonna come in over the top of the agent's original voice and translate from you know my English to Spanish and the agent's Spanish to my to English for me. Real time, synchronous, and we're starting to see it in some really big opportunities. Gina Gibson: And so with that, do we wanna show everybody watching here today what we're ⁓ what we're gonna see? Bob Maute: Sure. So if what you'll see here is our full portfolio, noise cancellation and voice isolation. Then we'll transition to accent conversion. And then we'll show live voice translation, ⁓ English to French with two of our ⁓ agents. Gina Gibson: Wow, that's impressive, Robert. ⁓ what impacts are you seeing it have in the space? Like I'm I'm envisioning, you know, our talent pool for hiring just got a lot larger, and the over like ⁓ I ⁓ language, ⁓ multilanguage agents are often more costly for contact centers. ⁓ That with this too. What are some of the other benefits? Bob Maute: Yeah. So accent conversion alone really expands the agent pool for contact centers. ⁓ you know, there's there's a limited number of agents that are available outside of the US that have the English, you know, competency to actually serve a North American market. By by eliminating those accent concerns, that agent pool increases significantly, customer satisfaction goes up. So Enterprises that either have their captive agents overseas or off or they ⁓ outsource can now see big improvements in both their KPIs, their average handle time, their first call resolution, but also have a much larger ⁓ you know agent pool than they previously had. For multilingual, what we're seeing is the cost associated with that. So if you look at language line and transperfect. Gina Gibson: Yeah. Bob Maute: Right. Both of them are third parties and and they've been the standard for years. ⁓ healthcare companies utilize them extensively because they are required by law to serve a certain number of languages for their customers, or they can be fined. Right. So some of those languages they may get three or four calls a month, some of them they may get hundreds, right? Or thousands or whatever it may be. A language line has been the go to tool, but what with language line, you're bringing in a third. Gina Gibson: Absolutely. Yeah. Bob Maute: part per person to the call. So you have to bridge that up. It's not immediate. It can take minutes, you know, sometimes double digit minutes to get somebody to the call, depending on the language. And then you have that activity of like there's a person there that's doing that. I don't know what the security risks are in that situation, but what we see is Gina Gibson: Mm-hmm. Bob Maute: We have enterprise customers that are basically looking at this as a replacement for language line and transperfect. And the savings can be north of 65%. Couple of yeah, so it's it's being adopted rapidly. We're showing 95% accuracy in our translation rates. This is from the customer, right? ⁓ what we're seeing is we're replacing 90% of the calls that previously went to a third party with AI. Only 10% of the calls need to be escalated to a live person in order to handle that. That's reducing average handle time by 30%, because it's synchronous versus the consecutive model that is a third party, where the third party listens to the entire comment from the one side and then translates it to the other and does the same thing back and forth. We're synchronous. There's a one to two second delay both ways for those agents where the AI comes in over the top. ⁓ zero concerns with one of our recent ⁓ case studies about safety or compliance with our AI. So this is like a game changer for you know contact centers that are looking to serve multilingual customers and the challenges with hiring the right number of speakers in a particular language, availability, churn. Gina Gibson: Mm-hmm. Bob Maute: You know, all of the things that are typical to a call center, but now throw, you know, make all of your agents multilingual. And what is really, really cool is traditionally we've done this by the IVR basically sending a Spanish call to a Spanish person, you know, person that has the their setup for Spanish. in I think it's 10 days we're gonna have ⁓ language auto detection. So rather than have to either route it through an IVR or push whatever. Gina Gibson: Right. Bob Maute: We will detect the a the ⁓ customer's language in a couple of seconds, notify the agent and configure everything so that that conversation can go forward without any interaction from the agent side. Which typically was choose the language that you're looking for or have it preset and have the R IVR send it to you. So just lots of really good things going on in ⁓ the voice translation space that are really impactful both cost wise and KPI wise. Four contact centers. Gina Gibson: I love that. What would be like one of the best use cases for for your new functionality? You've covered a lot. Like you've talked about BPOs, you've talked about, you know, ⁓ the Spanish and and being through the IVR. Is there a a use case you'd like to highlight for us? Bob Maute: Yeah, so it talked a little bit about it as we were talking. We've got a healthcare company that ⁓ we did a that they required a pretty extensive ⁓ proof proof of ⁓ value, right? As we went through it and we started to ⁓ work with them. We out of like nine hundred calls, okay, we handled eight hundred and ten of them with AI at a ⁓ reduced average handle time of thirty s thirty percent and an average cost savings of s north of sixty five percent versus what they were paying. ⁓ as we said there there was zero concern about compliance or safety in that those translations. And it's it's really cool. I mean we we have a feature in there that that will score every single call, identify where there might be a mismatch in terms of translation and allow you to, you know, allow the system to learn and coach the system on, you know, what that how that translation should appear going forward. Gina Gibson: It's s the numbers are are mind-boggling. Like I can't even imagine the dollars behind it for the large enterprises. Tell me what you've mentioned one to two seconds or whatever. What does the latency look like and what could impact the latency in production? Because I I think it's locally stored, correct? And so, like, is there anything on the customer side that could impact the latency that might end up hurting CSAT Bob Maute: Yeah. So because we our product operates on the device, of course we do extensive testing up front ⁓ or ⁓ research up front to make sure that the device can handle the the ⁓ the footprint or the resource requirements of our software, right? And every quarter our team the the rate at which we're which we're innovating and iterating is incredible. We keep reducing you know, that requirement on a regular basis. So we very rarely run into a device that that's not going to function as long as the agent doesn't have 16 tabs open and you know something running in the background. So so no that would be the biggest risk. But really the way it you know the the core function is on the device. We leave the device. It's the only s solution that we have where we actually leave the device and everything's not done there. Voice translation goes out. There's it ⁓ it taps into our AWS on the back end, hits the like large language model. We haven't been able to shrink that down so it fits on a laptop yet. And then it it sends the ⁓ translation back. So that's you know, you'll see a one to two second delay in when I start talking and when the AI comes in over the top. What we do is we we have a we are we can either eliminate the voice altogether of the original speaker. Or adjust it as we see appropriate. What we find is if we leave a little bit of that voice in there, okay, the agent who's listening to the customer can hear when they're talking in their original language. They don't understand it, so it just becomes white noise. And the translation over the top, which they understand, becomes the dominant it becomes what they hear. And they know during, you know, there's that one or two seconds where there might be some dead air. There's voices. They know there's voices in there. They're not, you know, they're not hearing dead air at any time. Everybody's talking through at a good consecutive clip. Gina Gibson: Excellent. Are you seeing any languages or regions ⁓ in particular that are the the demand is higher right now? Bob Maute: ⁓ so the demand is huge for Spanish, of course. We see a lot of French, German, ⁓ we've got some Russian. ⁓ but there's there's about ten core languages and we've got sixty one. So we conserve in sixty one languages, ⁓ we test those languages for accuracy. ⁓ you know, our top twenty seven have been fully tested, you know, by us. And then there's others that we where we've You know, the the sub the LLM model that we're using has been confirmed for that, right? So it's it's a ⁓ it's a super reliable, very functional, easy to use product when it's in a one-to-one communication environment. Gina Gibson: Excellent. What's next for Krisp? I mean, for since November that's a huge a huge amount of change already. So what what's next? What's on the roadmap? Bob Maute: Yeah. Yeah, the pace is quick. We've got a couple things that we're doing around deep fake detection and agent voice identification, which is important for the space. But I think one of the things that that's really neat about what we're doing is obviously we we've got we've got the tools that allow voice-to-voice communication on that front end of the CX journey, right? What we've also done is we have an SDK product. That SDK product sits in the infrastructure of conversational AI platforms, ⁓ some of the large telco CCaaS providers, and provides noise cancellation, can provide accent, it can provide voice translation. But one of the products we have called Viva is basically a coach the AI agent. Not only does it eliminate noise and voices, so the agent gets a clear signal. And we've showed where if we turn it off in a noisy place, the AI agent starts going crazy and things start falling apart. With the noise cancellation and the voice ⁓ background voice cancellation on, it can operate very clearly there. But what we also do is we will coach it so that when you start talking, if I'm an AI agent and I'm giving you some instructions, you start talking, we immediately tell it's time to stop and listen. Gina Gibson: And Bob Maute: And we also coach it when the proper time is to respond. So intent, inflection, word choices. So if you come to me and and and I'm an AI bot and I say we have a meet we have availability Tuesday, Wednesday, or Thursday at noon. And I go, hmm, Wednesday. A bot's gonna go, okay, great, Wednesday. We'll send you a ⁓ text with the meeting invite, right? That's not ready to respond yet. So we're we're Gina Gibson: Yeah. Yeah. Bob Maute: basically improving the front end of that customer engagement for both the AI ⁓ the AI agents side, on the AI agent side, and on the you know the live person to person voice side. Gina Gibson: Very, very interesting. so it's like getting into the quality side, right? And I think you mentioned sentiment before and everything else. That's a that's a changer for sure and a differentiator in the market. is there anything, Robert, that ⁓ you'd like the audience to know that I haven't asked you about today? Bob Maute: Mm-hmm. ⁓ I I I do I talk about this at every event where I get up and speak and I ask people if they know who Krisp is, and some people raise their hands, most of them because they have kids who are gamers and they use Discord, right? My son thought I was really cool when he found out I was going to Krisp because he he sees it on Discord all the time. But I mean, we're in two hundred we're on 200 million devices today and we cleanse 80 billion minutes of traffic a month, which is pretty significant. And I I ask all the time. Gina Gibson: Yeah. Yeah. Bob Maute: Who drives a Tesla? If you drive a Tesla, right, and you engage Grok, our software is helping make that communication more efficient and more effective. So while while many people don't know about us, we've been around for nine almost nine years now, and we've established a pretty broad ⁓ you know portfolio of products that ⁓ give us some real influence in terms of customer experience and customer success. Gina Gibson: I love it. I love it. ⁓ and to wrap things up, as you know, ⁓ we always finish with a little bit of ⁓ rapid fire. So ⁓ what sound or noise do you love? Bob Maute: Mm-hmm. Sound or noise that I love. I love the the low moan of my dog when I scratch his ears and he just it's like mmm so happy there. Love, love that noise. Gina Gibson: And what sound or noise do you hate? Bob Maute: ⁓ screeching tires. It never leads to anything good. Gina Gibson: And what profession other than your own would you like to attempt? Bob Maute: So I I've ⁓ I really for many, many years I've been a youth sports coach. and I've enjoyed it tremendously. But yeah, it it's a little harder to make a living in youth in coaching sports than it is. So I I I really have always enjoyed ⁓ you know, the coaching side. ⁓ and and it it plays into business for me too. I do do a lot of coaching here for my people and everything else, but I would love to be a a coach. Gina Gibson: Yeah. Bob Maute: in a you know, a youth program somewhere and have you know, have nothing to worry about but, you know, helping those those athletes ⁓ perform. Gina Gibson: Love that. And Robert, if somebody wants more information about using Krisp, who should they reach out to? Bob Maute: You they call you. I mean, you you have a great understanding of what we do in the space in general. ⁓ and thank you for what you do, you know, and the awareness that you bring to others is is fantastic. ⁓ but you know, we're easy to find. ⁓ you know, you can reach out to me, you could reach out I have ⁓ two channel managers right now running East and the West, ⁓ Nick Stocking and Laura Endris. ⁓ we're but I you would be the best person for that because you could sort of guide them down the right road. Gina Gibson: Perfect. Thank you so much again for joining us. ⁓ we really appreciate it. And and I'm just gonna put it out there and say we look forward to having you again the next time you guys come up with something really fascinating and cool. Love it. It's it's one of the ones I love to put under our cool tools because it is ⁓ definitely groundbreaking. So thank you so much for your time and we'll talk soon. Bob Maute: We'll have something pretty soon, I promise you. More appreciate it. Thank you very much. Have a great day. Gina Gibson: You too!