AIMinds #026 | Benjamin Gleitzman, CTO & Co-Founder at Replicant
About this episode
Benjamin Gleitzman is the CTO and Co-Founder of Replicant, a Contact Center Automation platform that uses AI-powered natural conversations to fully resolve customer service calls, allowing agents to focus on more complex issues. He previously created Hunch, a collective intelligence system acquired by eBay, and worked as a visiting researcher at Google, where he developed App Inventor for Android. Gleitzman also created The Algorithm Auction, the world's first auction of computer code, showcasing his innovative approach to blending technology with creativity.
Listen to the episode on Spotify, Apple Podcast, Podcast addicts, Castbox. You can also watch this episode on YouTube.
In this episode of AiMinds, host Demetrios engage in a thought-provoking conversation with Benjamin Gleitzman, CTO and Co-founder of Replicant. Gleitzman details his journey from a computer scientist and AI researcher to founding an art gallery dedicated to computer code, emphasizing the historical and aesthetic significance of algorithms. He shares a unique anecdote about DJing for LCD Sound System, reflecting his innovative blend of creativity and technology.
The discussion shifts focus to Gleitzman's work in conversational AI at Replicant, explaining the design challenges and achievements in creating AI that effectively mimics human interaction to improve customer service and operational efficiency. He illustrates how Replicant's partnership with Deepgram has enhanced voice transcription, significantly reducing response times.
Gleitzman also touches on the human aspect of AI, citing Replicant's role in reducing repetitive tasks in contact centers and consequently improving agent well-being, evidenced by a 70% drop in disability insurance claims. He further explores the need for greater inclusivity and empathy in voice recognition technology.
Overall, the episode highlights Gleitzman's passion for merging technology with artistic elements to evolve AI platform functionalities, providing insights into his entrepreneurial journey, including his experiences with Hunch and eBay, and his vision for advancing AI applications in business.
Fun Fact: Benjamin started an art gallery specifically focused on computer code, showcasing items like the oldest algorithm written on a Babylonian clay tablet, emphasizing the intrinsic beauty and artistic value of code.
Show Notes:
00:00 Users shared predictive data, raising ethical concerns.
06:01 Ancient procedure for cistern volume calculation.
09:18 Startup founder defends decision on technology's use.
11:46 Desire for Spotify to connect like-minded users.
14:06 Creating a repeatable process for developing startups.
19:31 Advanced technologies rewrite customer voices to aid agents.
22:51 High success in contact centers, expanding services.
26:03 People prefer speaking with agents over machines.
29:43 Agents struggle with outdated technology in contact centers.
More Quotes from Benjamin Gleitzman:
Transcript:
Demetrios:
Welcome to the AI Minds podcast. This is a podcast where we explore the companies of tomorrow being built. AI first, I'm your host, Demetrios, and this episode is brought to you by Deepgram, the number one speech to text and text to speech API on the Internet, trusted by the world's top conversational AI leaders, startups, and enterprises like Spotify, Twilio, NASA, and Citibank. In this episode, we are joined by none other than Benjamin, the CTO and co founder of Replicant. And I have a funny question for you, Benjamin. You've got a story about Ben Dorgan that you can kick off. I think you were doing some dj work there, which is not related to anything that we were planning on talking about today, but I got a keep it rocking because we were having such a great conversation before I hit record.
Benjamin Gleitzman:
Yeah. Thanks so much for having me. It's really a wonderful opportunity to speak with you. Big fan of Deepgram. Been using the service for a number of years, and it's been a great partnership with replicant. I'm coming to you live today from Portland, Oregon. I live in a small town called Corvallis, and I was recently in Bend, Oregon, to open for lcd sound system at a major amphitheater, which is such a dream come true, is really fantastic. And I've had a love of music and the arts for many years.
Benjamin Gleitzman:
Although I had my start as a computer scientist and also an AI researcher, I took a little bit of a left turn after I had sold a company to eBay, and I said, you know what? These big company cultures are not exactly what I'm looking for. And so I started an art gallery of computer code, and we would really celebrate code for code's sake. What if you could just make code and have it be kind of beautiful? Not for a quarterly profit or to think about the amount of money it could make, but just for the beauty that it would bring to the world. And so it's really a phenomenal way to work when you can be inspired in that way.
Demetrios:
So many questions come up around the art gallery, around doing AI back in the day. But the biggest one, my burning question is, how did you get the call to be the DJ for LCD sound system?
Benjamin Gleitzman:
Oh, it was. My heart was racing, for sure, and I was actually only given about two days notice to prep for it, which was just kind of wild. So right out of college, I had joined a company called Hunch, and it was a recommendation engine. And this was back when online quizzes were really hot. And so, you know, it was fill out this quiz and we'll tell you what festival you should go to. Fill out this quiz and tell you what your career should be, or, you know, what shoes you should werewolf you are. Yeah, exactly. Yeah.
Benjamin Gleitzman:
Which character from the Muppets are you?
Demetrios:
Wiley.
Benjamin Gleitzman:
And people were very, it was interesting how much information people would share about themselves. These were questions that were written by our users and then the AI would figure out, okay, which of these are predictive and which of these are not, but our customer base would answer 125, 150 questions about themselves, which was a very eye opening. And I think it was at the time that people started to think about, well, what can you actually do with this kind of data? And now I think we've seen maybe a little bit of the darker side of that data where Facebook can take all this information about you or meta can know everything that you're looking at. And it's so good now that people think they're being listened to. But really, the profiles are so rich of this data that we've put out into the world. But this was still the early days. I think it was before Cambridge Analytica and before some of the major scandals around data. And my CTO at the time, Matt Gaddis, who's brilliant engineer, brilliant researcher, really kind of opened my eyes when it came to what we could possibly do with AI.
Benjamin Gleitzman:
His partner is the keyboardist and a musician in LCD sound system and is just phenomenal. He plays the analog. Since her name is Abby, shes very phenomenal. And so she asked me to come and play. It was a real honor and kind of, it was wonderful to be able to look out into the crowd, see so many happy, excited people just living in the moment.
Demetrios:
Wow, what a special time to go out there. And Im blown away by that because the people that understand who lcd sound system are, that are listening probably are very jealous right now. The people who do not, probably do not care.
Benjamin Gleitzman:
Go listen to their album Spotify, please. It's, yeah, they're really phenomenal and also fantastic people. And so that's, it's great when you get to meet your idols and they're actually wonderful individuals.
Demetrios:
Yeah, that helps. That helps. So you have been kind of like moonlighting as a dj, but that's not your main gig. And you also had a gallery for code, just for code's sake. How did you present that to people?
Benjamin Gleitzman:
Yeah, it's a great question because there's obviously you think about art and you think about visuals, you think about flemish painters, you think about the way that they use light and the way they use pigment. I'm interested in provenance. I'm interested in the history. Where did this piece of code come from? Why was it created? Who went to jail because of it, who became rich because of it? What were the motivations behind it? And so we had everything from, we did have some visual piece. We had the oldest algorithm that we could find. It's on a babylonian clay tablet from 4000 years ago. And it describes a process, an algorithm for how you would calculate the volume of a cistern. And so it's this beautiful cuneiform writing on a small tablet.
Benjamin Gleitzman:
And at the bottom, after describing how to calculate the volume of a particular cistern, it says, this is the procedure. And you see this? This is the procedure listed at a number of these different clay tablets. And it means, I've told you how to do something once, but you can extrapolate this. And this is the procedure for how you will do it for know, a cistern of any width and length and depth. And I found that very beautiful, the way that it was presented, and not only visually, but also the history of where it came from. Now, I've been told by Egyptologists that the Egyptians actually had a lot of algorithms before the Babylonians, but they were written on papyrus. And so I think that's why storage medium also matters. And so remember to back up your cds because they may not last forever.
Benjamin Gleitzman:
But, yeah, it's wonderful to see the through line of the babylonian clay tablet all the way through to. We had a virus that was written specifically for the auction. And what was interesting was the auction house refused to sell our virus. They'll sell duchamps toilet. They'll sell very outrageous pizzas, sometimes very offensive pieces. But the virus was a little bit too hot for them. And so I thought that was interesting to see where the line is drawn there with what the art world is ready for.
Demetrios:
Yeah, I think I remember something making news where they sold a glass of water.
Benjamin Gleitzman:
Yes.
Demetrios:
And like, they'll sell that.
Benjamin Gleitzman:
Yeah, they'll sell. They'll sell a Banksy that, you know, destroys itself, you know, quickly after. Quickly after the sale. And, you know, I think it was also a simpler time. It was before nfts. It was before a lot of like, I think, grifting kind of came around into the, into the art and technology world. And so it did feel a little bit more pure. And it was an eye opening way of working because it pulled in not just the hard technology and people who were excellent data scientists or excellent AI researchers.
Benjamin Gleitzman:
It was painters and it was sculptors, and it was people who had a very fantastic way of speaking and a command of language. And as I think about the AI of the future and where we're going, it's not always code, it's now. How well can you write a prompt? How well can you get an idea out of your own mind and into, I'll say, the mind of the machine? I don't want to personify it. I think machines are more than human. I think, I don't want to be ethically correct here in saying that a lot of the creations that we're making are, they're not people, they are machines. And I think that it's interesting to think about how the command of language comes into that.
Demetrios:
So you spent time in the art world, and eventually you were inspired to start another company and go through the whole rigamarole again. Are you a bit of a masochist, or what's in it that you wanted to go through this? Cause you know how painful it is, right? I imagine you had been through it. Was it that you had so much time behind you, you forgot it? Kind of like what I hear happens to women in childbirth, where it's painful in the moment, but then afterwards they romanticize it.
Benjamin Gleitzman:
You look at the faces of your children, and you say, oh, yeah, I remember why.
Demetrios:
That's exactly what happened to me. That is very true.
Benjamin Gleitzman:
Yes, yes. So, yeah, I did throw my hat back into the ring after, I think, swearing off any kind of a venture backed startup. But that company hunch, that was the recommendation engine that was eventually acquired by eBay, and it was a great exit for the company. But I had a little bit of a different idea of how I wanted a technology like that to be applied. You could find your tastemates, you could find your nemesis, you could learn a lot about yourself. And today, it does power billions of recommendations for eBay, but it's selling shoes, it's slanging watches. And so I think it's a little bit different than the ethics and the goal of what I had for an algorithm like that. But in the process of that acquisition, I got to work with Jack Abraham, who's one of my co founders here at Replicant, and we had a particular way of working where we would take a very small team.
Benjamin Gleitzman:
Actually got a call from him around midnight one night, and I had not met him before. I'd heard of him, just from the world of startups, and he said, benjamin, I've heard some great things about you, and we're on a secret mission from the CEO of eBay, John Donohoe at the time, and we're going to go down to Australia and we're going to completely reinvent what eBay is, how the recommendations work, how the items are sold. Let's modernize it in a way that it really hadn't been modernized since the late nineties. I said, wow, that sounds great. When do we leave? And he said, 06:00 a.m. i. So this was a kind of, I loved the immediacy, I loved the excitement of kind of, you know, I didn't even know who I would be traveling with. I knew Jack would be there.
Benjamin Gleitzman:
And so showing up at the airport, kind of one by one, the main characters were revealed. A phenomenal designer who is now head of design at Meta, amazing computer scientists that have gone on to create, you know, phenomenal aiworks and Jack, that way of working where we would take a small team, go to a place, I think we spent about three weeks down there, developed a prototype product, brought it back to a lot of fanfare, but then also the difficulty of working within the system, the startup within the startup, trying to go and push this forward was another interesting path.
Demetrios:
Just a little bit of a tangent, because I think you're gonna appreciate this one where you were saying, oh, I would love to know who my nemesis is or who I can pair with. I've been thinking a ton about how Spotify has all this incredible data on me and what type of music and podcasts I listen to. It would be awesome if I could know who else is like this special snowflake like I am, that listens to this certain type of music, but also enjoys these podcasts and that type of thing, but they don't do any kind of social, which is weird. They used to kind of be able to see what your friends are listening to, but that's if you add your friend or you have your list, the playlist, and you can see someone who's made that. But I know that they have their embeddings of me, and I represent an embedding which is, you know, in vector space somewhere, and there's probably a lot of people right around me that I would enjoy meeting. And so it's weird to me that they don't do that.
Benjamin Gleitzman:
Yeah, I feel a little crestfallen with the trajectory of Spotify because there were a number of companies I love to do hackathons. It was a major excitement back in the day. I still love it. And you could upload a song and it would say, hey, here's the number of beats per minute. Here's where the downbeat is. You could, like, remix things really easily. Spotify went and, like, snapped up and purchased a lot of these companies and then kind of shut those features down. You know, they really didn't extend it.
Benjamin Gleitzman:
And what we found at hunch was that if you're looking for someone you can really connect with, yes, it's good to find people who have things in common. But the best way is if you collectively dislike something together. You know, like, if you both don't like some particular type of music, that's going to be a better indicator of your long term friendship ability than liking the latest Taylor Swift or something like that.
Demetrios:
Oh, that is hilarious. So tell me what you're working on now.
Benjamin Gleitzman:
Jack and I had worked on this project for eBay. We kind of both went our separate directions. I started the art gallery. He started, actually, an incubator that took that particular way of ideating and starting companies.
Demetrios:
You mean flying to Australia at 06:00 a.m. with no heads up?
Benjamin Gleitzman:
Yeah. But also thinking about how can we make this a repeatable process? Can we get a finance team, can we get a team of lawyers for incorporating the ideas? Can we get an HR department that can help with hiring and essentially have a studio that turns out a number of different ideas and sees which of them have legs? And so Jack and I would get together every couple, six months, and we'd talk about what was new with him, what was new with me, and is there any overlap that we could find a way to work together again? And there were a number of fantastic startups that came out of atomic, but they never quite had the spark that I was looking for, which was bringing together the really hard technology and the arts as well. And it wasn't until we started to discuss language and voice and conversational AI that I thought, and this was back in 2017, we have these amazing assistants, or what I saw as amazing assistance at the time. We've got Google home, we've got Alexa. Why is it that I've never had a very good conversation with a machine? Something seems like it's missing there, and I can tell that the technology is going to get better and better, but it seems like there is an artistic side to this that's missing, that Google doesn't seem like they've cracked yet. Amazon doesn't seem like they've cracked yet. Watson from IBM doesn't seem like it's quite cracked yet. And I would say, even up to now, here we are, seven years into the founding of the company, it's still kind of an unsolved problem.
Benjamin Gleitzman:
We're getting better and better, but it will require a lot of innovation to create the great conversations of the future.
Demetrios:
I often wonder if it's because they haven't cracked it or they didn't think it was that important, like, there isn't a clear value on it for them or they can't, like, associate. All right, well, yeah, we'll sell more ads if we have better conversations, that type of thing. I imagine it's not that cut and dry, but I often wonder that. And so if I'm hearing you correctly, you said, we don't have that. I want to go and build it. I want to make that a thing.
Benjamin Gleitzman:
Yeah. Unfortunately, I think the motivation or the alignment of motivations is perhaps not there at some of these larger companies because it's a difficult task to create these conversational agents, we call them thinking machines. What it leads to is people not being able to talk in the way and express themselves in the way that they wish they could. I grew up in West Virginia. I have a pretty strong accent when I go home. And I want the machine to be able to understand the West Virginian just as well as the Californian, just as well as the person from Germany who's speaking English as a second language. You should be able to speak in the way you want to speak. You should be able to use not only the accent that you have without code switching, but also the word choice that you want to use and have the machine understand that as well for you.
Benjamin Gleitzman:
And if you miss out on that, if you don't allow people to speak in the way they want to speak, then you're kind of capping yourself at a certain level of resolution. I only hear white males well, and I don't hear anyone else. This is an issue with me in my own home, where the Google home hears my voice a lot better than my partners. And so shes often asking me, hey, can you tell Google to turn off all the lights because its not listening for her voice?
Demetrios:
No way. Thats frustrating.
Benjamin Gleitzman:
Yes. Yes. So replicant is a conversational AI platform, a contact center automation platform, where were really uplifting agents in the contact center by taking away a lot of the minutiae that they need to do on a day to day basis. I used to work in a contact center. That was how I helped pay for college. And I loved the very interesting questions that would come in, all the kind of things that required empathy, things that required creativity, things that required. Hey, I've never heard of that problem before. Let's do a little bit of research for me.
Benjamin Gleitzman:
And then, on the other hand, there was so much minutiae of the same questions that I would get asked a couple, a couple 10,000 times over the course of a couple of months. And I wish that I had a machine at that time to be able to automate some of this, because it would allow me not only to free up my time to work on more interesting problems, but that I wouldn't be so fatigued at the end of the day that I wouldn't feel like, wow, I've just repeated myself 100 times.
Demetrios:
Yeah, I know that. Even when I was just at a conference or when I go back home to visit my family, and I see all the old friends around the city, and it's like telling the same story of, what are you up to these days? And you kind of, like, try to spin it differently, but there's so much you can do on it. And I can only imagine when you're working at a call center or a place where you're getting the same questions over and over. By the end of it, you're like, man, I can't say this one more time, otherwise I'm going to fall asleep as I'm talking.
Benjamin Gleitzman:
Yes, yes. And there's interesting technologies that are being built now that take furious, frustrated customer voices and actually rewrite the pitch and the tone into someone that isn't as frustrated. And that's what the agent hears. Because imagine being spewed all of this vitriol all day. It could really wear people down. And one of our customers, we work with a number of the AAA's and canadian AAA's, they told us that after they had deployed replicant to be able to augment their agents in the contact center and take on some of these more mundane tasks. They saw a 70% reduction in workplace disability insurance claims. Now, I don't think this was all replicant.
Benjamin Gleitzman:
They did a number of different things that have led to higher levels of age and happiness. But that ability to come into work, know that you're not going to have to deal with a lot of those minutiae calls. People are actually sick less, which is very impressive.
Demetrios:
So how are you doing stuff like this?
Benjamin Gleitzman:
Art, and it's the science. And in the early days, 2017, we said, let's build it all. Let's build transcription, let's build a really good sounding voice. Things were quite robotic back in that time. Let's build a really phenomenal brain that can comprehend all of the different ways that people can speak. And let's build a graphical way of representing all of this so that you don't need to have a PhD to understand how the conversations are built. And I think that key understanding of the more people that we can get around the table, the better these machines are going to be. Because it doesn't matter whether you are a contact center agent, you work in the QA department trying to make the agents better.
Benjamin Gleitzman:
Whether you're a manager, you understand the business processes, whether you're the copywriter. It's like everybody has a little bit of input on how you can make these machines better. And I in that coalescing people around the problem and getting a lot of different input from a lot of different types of people and people who think in different ways, that's leading to these high resolution rates that we see in the context center.
Demetrios:
And now there was a question around the platform and having it do everything or having it do one thing, and you've taken a stance on that, right? And I would love to hear how you came to deciding what you've decided.
Benjamin Gleitzman:
Yes, I think that the, there's kind of two pieces here. There's the. It seems to me that customers are asking for platforms that can do more and more and simultaneously we're getting technology that's getting better and better from different players. And so even though we still have in house transcription models, we're often using Deepgram because we can get name collection. For example, my last name is Gleitzmann. I dare you to spell it. And so when we switched over to Deepgram, we saw a 17 percentage point increase. So it wasn't 70% better, it was a whole 17 points more at getting name collection.
Benjamin Gleitzman:
And that's phenomenal. That's the kind of success that I want for the callers that are into our contact center. And so it doesn't matter whether I'm calling out to Deepgram, or I'm using my in house model, or I'm using a symphony of models working together, we want to be able to provide the best possible transcription at that turn in the conversation, the best possible natural language understanding, and then a really good quality voice that can speak back the answer that we want and that leads to these high success rates. Now, what that means is that companies that are specialized in a particular area, like replicant, started with voice, and we only did voice. Now we do sms, now we do chats, now we do agent analytics as well. Our customers are inviting us and really giving us permission to do more and more in the context center, and it's less of a headache for them. I talked to contact center managers who they used to fly to the Philippines a couple times a year, fly to South Africa a couple times a year to visit these business process outsourcers, to visit the call centers that they had set up in different countries where people are not making that much money either. I mean, I think that there's kind of this race in the context center world to, oh, did you know, if you picked this particular city, that's not manila, but it's still in the Philippines, you can pay, you know, less and less.
Benjamin Gleitzman:
It's a, there's a number of ethical concerns that, you know, come in here as well, but not having to manage those relationships or deal with all of that travel and where they can come to one vendor and say, look, I trust this vendor to be able to automate all of my calls, chats, SMS emails, and to do agent analytics, but also be listening for ways that we can improve our automation based on the agent calls, all in service of the customers, of our customers being less frustrated when they call in on the phone. That's great to see.
Demetrios:
Well, it feels like you think a lot about conversational design and how to make that the most natural and also the best experience possible. Are there things that you've specifically decided to do to make better conversations and make it so that your customers have an easier time having those better conversations?
Benjamin Gleitzman:
Yes. Conversational design is so key. It's really a, I think, still an undiscovered and kind of untapped and perhaps even unsolved method of creating really fantastic conversations. Uh, one that we arrived on early was around honesty, honesty with our customers, honesty with our callers about, you're speaking with a machine. You know, this is, it's the fastest, the smartest, hopefully the best sounding machine you've ever spoken with. But at the end of the day, this is not a person. And so just having the confidence to upfront at the top of the call saying this call is recorded and it's with a thinking machine, was a yemenite. It was a little bit wild, it was a little bit of a jump in the early days.
Benjamin Gleitzman:
And what we saw then was that people would say, well, Benjamin, I want to talk with an agent. I don't want to talk with a machine. I need a person to answer my problem because I've never, ever spoken with a machine that's been good enough to actually deal with my roadside emergency issue with a flat tire and figuring out where I'm located, calling out to tire shops and figuring out who has the best ETI. They just had never seen that before. And so people would say, well, I gotta speak with an agent. And one of our conversation design principles that we discovered early on was, what's in it for me? Why should I actually spend time speaking with this machine when I could wait on hold and spend 15 minutes waiting for a person? And so, being, again, transparent and honest. I know you want an agent, but I can help you right now with no wait time. Can I get your policy number? Can I get your name? Oh, it looks like you called from this particular location before.
Benjamin Gleitzman:
Is that where you are now? You pull people back into the conversation, and you give them confidence that, hey, this is a new kind of technology that perhaps they've never spoken with before, but they're willing to give it a try.
Demetrios:
I am 100% that person asking for the humanity because of the painful experiences I've had in the past. And so I imagine everybody's been conditioned with those robotic, painful experiences where it's not picking up what you're saying, and it's not really understanding. And so you end up frustrated or banging the phone against the wall. And it's funny that you're almost using, like, a persuasive technique. Like, yes, I can definitely give you a human. That's fine, but if there is a better way, let's just try me and you, and where I get stuck, I'll pass it over to the agent. You can wait 15 minutes.
Benjamin Gleitzman:
Exactly. Choice is really key here. You know, let people choose whether it's, nope, I got to get to an agent, or, you know, this is something that I want to explore now, see whether I can get what I want, because that's, I think, how people are going to have their minds changed a little bit. Another key is, I don't want to have to repeat myself. If I tell the machine, this is who I am and this is where I'm located, and we get a tow truck on the way, but then I say, oh, it looks like the bill for that tow truck is going to be a little bit high. We don't want the machine arguing the ins and outs of the payment. So when we escalate you over to an agent, we're using language models to summarize exactly what happened on that conversation. Give a real, glanceable summary over to the agent so you don't get passed along.
Benjamin Gleitzman:
And they say, okay, now re explain everything that just happened on that call. So that's again, it's an expansion of the platform that we're able to offer these agent screen pops, agent analytics, in addition to the core of replicant, which is really that automation over voice.
Demetrios:
I lived in Spain for about ten years, and the funny thing is that just calling Vodafone, my spanish phone provider, I have to repeat myself every time they connect me to a new agent. And so it's like, I wish the humans would do this too, and my information could get transferred from one to the next. So when they transfer me over, it also goes there. It's so frustrating for me that when a new agent picks up the phone, they go, okay, what's your name and what's your id number? And now let's do some security checks. And it's like, really? I literally just did this. I just did this, I promise you.
Benjamin Gleitzman:
And I feel for the agents in these scenarios, too. Yes, it sucks as callers to have this happen, but imagine if that's your job, that the technology that you're given has let you down in such a way that you can't get that information passed along. And it's not like it's that tough to do. I just think there's been some neglect of technology, specifically in the contact center, which is why I think we found so much success working in this space, is that we're taking a pretty antiquated model and we're taking the brand spanking most new AI that we can, and a lot of great conversation design and a lot of good design thinking, and we're applying that to the problems that are in the context center. That, as you said, everybody's had that negative experience. Everyone's gone and had to go outside and take a walk after they've talked with their service provider, because of how awful that experience was. And I think consumers are not willing to put up with that anymore. And they're actually voting with their feet.
Benjamin Gleitzman:
I think they're saying, that was such a bad experience flying this airline. I'm never going to do that again. I'm going to opt for an airline that has better customer service because that's what I deserve.
Demetrios:
I am so happy for you coming on here and doing that, telling me all about it. It is something that I hope my phone service providers get sooner rather than later. It will save me a ton of time and a ton of headache, as you put it, the nice way I won't have to go and take as many walks.
Benjamin Gleitzman:
Yes, yes, exactly. And that's why I'm so excited about this partnership with Deepgram the success that we've seen both on the transcription side, but also what I know you have coming down the pipeline, which is voices as well, really high quality voices, that's been amazing for us to see also on premise deployments, because latency really matters for us as well. Being able to quickly respond to callers, that's the other kind of side of this has got to be really smart. But it also has to be pretty fast because when you get 300 millisecond delay, 500 millisecond delay, that's okay. When you get up into those one, two, 3 seconds, you're like, are you still there? But people start talking over one another. And so the steps I've seen for Deepgram to continue reducing latency, give us options like doing that on premises deployment, it's leading to a fantastic partnership.
Demetrios:
Excellent. Well, thanks for coming on here and explaining everything. I appreciate you.
Benjamin Gleitzman:
Thank you so much.
