#292 Peeyush Ranjan: How Nurix Is Redefining Voice AI for the Enterprise

Discover how Neurix is building the next generation of voice AI that sounds and reacts like a human in this conversation with Peeyush, former Google technologist and co-founder of Miracle Labs. Peeyush shares how Neurix is solving the toughest challenges in conversational AI — from eliminating latency to creating natural, real-time dialogue that mirrors human interaction. He explains why voice is the hardest frontier in AI, how Neurix's proprietary models manage conversation flow, and what it takes to integrate voice agents into enterprise systems at scale. Learn how Neurix is combining low-latency speech recognition, dialogue management, and large language models to deliver seamless, multilingual customer experiences. If you're a business leader, product builder, or AI professional interested in how human-like voice agents are transforming customer support and enterprise communication, this episode reveals the future of intelligent conversation. Stay Updated: Craig Smith on X:https://x.com/craigss Eye on A. I. on X: https://x.com/EyeOn_AI

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Speaker 00

  The number one thing which we really invested in NewX is actually in making sure that the latency is really low. Let's say you're talking to a model which is gonna perform certain tasks for you. There is in chat and all, it's very clear, you can have more latency, even in voice when it comes over, there's expectation built. But when you call into a business and you're talking to a person, there's no expectation latency. It's like almost human-like behavior requires, as I finish, you talk. And not only that, there is a little bit of, you know, parallel conversation going as well. For example, if I'm speaking for a little bit longer and you're saying, mm-hmm, and all those things, like, you know, those are markers of like, yeah, you're nodding along and you're following me.

• 0:45Speaker 00

  Those are also part of the conversation. And if you just keep quiet for too long, then also I feel like, oh, are you even listening? Are you there? My name is Peeyush. I'm a technologist at heart, you know, and I've been working on a variety of different problems. It's had like an intersection of technical depth and, you know, worldwide impact, thanks to a long-time association with a company called Google, where I joined like nineteen years ago. I left them just end of March and over the period of time, you know, have worked on a variety of problems from search to, you know, workspace.

• 1:24Speaker 00

  And most recent and fintech and most recently, past three years, I was working on AI as part of Google Assistant and the Gemini app. And as of March, you know, I left Google to join with her ex-colleague and a friend for last ten years, Mokesh Bansal, in creating, in partnering in what is called Meraki Labs, which is about incubating and building, you know, AI companies. You know, we all see that it's changing the world and NewX is, you know, first of those companies which, you know, I have been involved with.

• 2:11Speaker 00

  I'm on the board and obviously giving a lot of technical input and architectural input on how things can work and how they can scale. NewX itself, yeah, go ahead.

• 2:25Craig S. Smith?

  No, go ahead. So Newrix is one of the companies incubated.

• 2:30Speaker 00

  Uh huh. Yes, yes. And you know, I'm just starting. I mean, this is, this is the first and in a really important space, which is, you know, of conversational agents. So i- if you ask me like about Newrix itself, I think the the toughest problem which I feel like Newrix is handling, which is required is actually having a world-class voice AI solution. You know, most of the conversational agents, there are many different ways conversations are had, right? You can have text and email-based or chat-based and voice-based, you know, and of course in future like video and other things, but

• 3:12Speaker 00

  currently, voice is the cutting edge of that, which has the hardest set of problems within it. That is one big space which Newrix is going deeper into in variety of ways through new innovations. And then there is, of course, the other area is how do you make this agent work really well for enterprise use cases? So there's a voice set of problems and then there's an enterprise set of problems. And the goal which Newrix has is to, through this, be the smartest, you know, operator for the brand, like, you know, when you are talking to Newrix for, you know, when deployed inside a company,

• 3:55Speaker 00

  it is essentially a representative of the brand that you are talking to them, whether it is from starting from sales all the way to support. And also, at the same time, it's not just that you are talking and getting information. Is the smartest agent which you can talk to because it integrates deeply into the enterprise and is able to help you through all your workflows. So those are the two big facets of it.

• 4:20Craig S. Smith?

  Yeah, I've been talking to a number of people about voice and about

• 4:27Craig S. Smith?

  translation also. I just talked to a company called Deep L. I don't know if you know them, but they do, they do translation. And then I, I, earlier I had on the podcast, a company from.

• 4:48Craig S. Smith?

  I always get their name mixed up because there's so many similar names. I think it's Speechmatic. Does that sound, yeah, they do very uh natural sounding. Their emphasis is on very natural sounding speech.

• 5:06Craig S. Smith?

  So it, one of my questions is,

• 5:11Craig S. Smith?

  you know, the underlying tech uh exists, it came out with uh with the advent of transformers. OpenAI does, I talk to OpenAI's chat GPT when I'm in the car and it's remarkable, even the standard voice, how natural it sounds. Uh, so how do you, how does a new startup fit into that crowded market? What, what, what differentiates, I guess?

• 5:45Speaker 00

  That's, that's a, that's a great question. I think that there are a couple of things which uh uh have to be kept in mind. So let, let me first talk about the voice part of it. Uh, the voice part of it, the number one thing which uh we really invested in Neurix is actually in making sure that the latency is, is really low. You know, uh, when you are talking to, uh, let's say you're talking to a, a model, uh, which is gonna perform certain tasks uh for you. There is in, in, in chat and all, it's very clear. You can have more latency, even in voice, when it comes over, there's expectation built that, well, I asked something, it is gonna take a while for it to get

• 6:28Speaker 00

  to the answer. But when you call into a, a business and you're talking to a person, uh, there's no expectation latency. It's like almost human-like behavior requires, as I finish, you talk, and not only that, there is a little bit of, uh, you know, parallel conversation going as well. Uh, for example, if I'm speaking for a little bit longer and you're saying, mm-hmm, and all those things, like, you know, those are markers of like, yeah, you're nodding along and you're following me. Those are also part of the conversation. And if you just keep quiet for too long, then also I feel like, oh, are you even listening? Are you there? Right? So there are these kind of psychological aspects of it, uh, which are above and beyond the

• 7:09Speaker 00

  actual uh, uh, likeliness to a human uh, speech. This is actually about conversation. So we invest a lot of time in, in making sure that the latency does not drop. And one of the reasons why it is important is when you're calling into an agent, well, the two things, one is when you're calling into agent, you might ask the agent to go do something. Right? Therefore, your, your pipeline of activities can have latency introduced in many places, but we don't want your voice to be uh, your uh, voice interaction to have any kind of latency. So we have invested a lot of uh, uh, engineering in there to actually be able to respond to you super fast irrespective of what is happening behind. That is one.

• 7:52Speaker 00

  The second thing is what I was talking about, that human interaction and dialogue between two people is actually a little more nuanced uh, than the uh, you know, over and out Roger, you know, kind of like walkie talkie style. So here we have our own proprietary dialogue manager, which also is trained to understand that, okay, this conversation which is going on, is the person stopping and thinking about because you suddenly ask them, hey, what are your last four? Or what is your driving license number? And the person is quiet, it's probably because they're fishing for their driving license. As opposed to, you know, uh, it's your turn to talk again, or if the hmm is happening, then that's what that means. So we have, that is the second

• 8:31Speaker 00

  uh, uh, investment in there. Uh, and the third thing, which is beyond a single person to person conversation is like, imagine a, a call center, right? Now there are like, uh, five hundred people calling simultaneously. It naturally from system design, right? Engineering wise, it can have impact on latency of each one of them, depending on how you architect it. So we have to think about that architecture as well, that when the load goes up, it does not mean every people, every person feels that their agent suddenly becomes slower. So all of these are aspects, you know, this is the uh, one of the things. Which I feel we are feeling is is the more you get deeper into the domain of how the models are being applied.

• 9:16Speaker 00

  The more these nuances start to show up and you have to go solve them because the models are very, very capable and they'll only become more capable later. But you have to understand these nuances and you have to solve them properly. So that you can deliver a world-class experience on the other side. So, so the point you made about. These capabilities which are there, yes, they are there and we leverage them wherever they are. And then we solve these problems so that. The person calling into the that business feels like the other person is, you know, really as human-like as possible.

• 9:46Craig S. Smith?

  Yeah, and one of the things you're talking about is endpoint detection, right? Knowing when the other side has completed a sentence or something. And uh uh it and what are the underlying models or or do you are used? Is is Newark sort of model agnostic because you know there are new models coming out all the time? Uh, or or are you working off of open source models and fine tuning open source models? Um, yeah, I'm curious about the architecture.

• 10:24Speaker 00

  Yeah, so there are parts which are our own proprietary models, like the stern-taking piece is our own proprietary model. Uh, you know, we need to really make sure that we understand the conversation which is happening, which are many times business-specific conversations. And also we understand, uh, you know, when is it that because. We also operate in variety of noisy environments and all that, so we have to make sure that we we understand when really there's an interruption versus there is not. So that is proprietary model. But then when you start deploying that uh into the uh, you know, a particular customer.

• 11:03Speaker 00

  Then there are specific models depending on a what is the task which they want to apply it for. You know, we are able to integrate with OpenAI, we are able to integrate Azure implementation of that. We are able to integrate with Gemini over GCP through Vertex. All of these are available and they are all integrations depending on the deployment which we are going after. There are other places where we also have custom models like the speech to text. I think I mentioned that, like, you know, the real world noise detection and saying, okay, what exactly is this person saying? And understanding the speaker is talking into the phone as opposed to somebody else. Or there is two other people in the background talking.

• 11:48Speaker 00

  All of that is actually done using a proprietary model because we want that experience to be perfect. So that and the dialogue back and forth is done by our own proprietary model. And when deploying into the business, then we can choose what is the best model for, you know, that particular use case.

• 12:05Craig S. Smith?

  Yeah, and and this this product, it's not a consumer product. It's it's going into the tech stack for various other applications as as voice engine.

• 12:21Speaker 00

  Yes, it is actually an enterprise product. It is still an end user product in the sense that the end user interacts directly. They'll dial into or they will call into or they'll message a Newric's agent. But that Newric's agent is deployed for a customer, like if it's a retail customer, will deploy it. And the retail business will deploy for. And the end customer of that retail will call in and say, where is my order? And that call will be fielded by Newric's agent. So that solution is built by us and it is available, but they feel like they're calling into the business. And then it is Newric's agent which then picks it up, talks to them. And based on how it has been configured and deployed, it will actually go look up in the order managing system or something.

• 13:06Speaker 00

  And then be able to give them the answer back or or, you know, act on that order depending on what the operating procedure is.

• 13:13Craig S. Smith?

  Yeah, and and again. You know, am I wrong that this is an increasingly crowded space? I mean, I just from the number of people that contact me or the various conversations I've had.

• 13:30Speaker 00

  Uh huh.

• 13:30Craig S. Smith?

  It's like there are a lot of people out there trying to tackle this problem. And is is the market large enough that uh that it'll support many, many solutions? Or are companies going to start? Uh, well, one of the other questions is these guys at D Bell are doing translation uh between language pairs.

• 13:58Speaker 00

  Uh huh.

• 13:59Craig S. Smith?

  Uh, and they're not doing, actually, they're not doing uh voice to voice. Uh, they're doing text to text and voice to text. But um presumably you need uh models fine tuned for each language. Uh, is Nurix focused solely on English? Or are they um uh looking at other languages?

• 14:30Speaker 00

  Um, Nurix is focused on multiple languages. We are currently available in 20 plus languages. And you know, um, uh, we have customers in in US and in India and in in both of these countries. Multiple languages are needed, uh, you know, especially like in US, you at least have like English and Spanish. And and in India, uh, you know, almost every state has its own language. So, uh, you know, it it's multilingual. I will uh say one thing, uh, it's not a language translation experience. It is actually a uh conversational agent experience. So you can talk to this Nurix agent in any language. And primary value you get out of it is the fact that it can actually add.

• 15:14Speaker 00

  On the business's behalf, as your customer support or your customer sales agent, while you are talking to it and you can talk through like a chat or a voice or whatever. Now, like you said, it's a very, you know, happening space, a lot of people are entering. Generally, the two things that I will say, that yes, it it kind of shows that the market is ready to accept solutions. And we are seeing the same thing, like you said, that we have over a quarter million, you know, calls which were done. Just, you know, this year and and we have five X quarter over quarter growth, right? So seeing this kind of growth, it's not surprising that a lot of people are coming because they're like, market is ready.

• 15:58Speaker 00

  Now, the question is, just like we have seen it many times in the past, that when a market starts to become ready, a lot of people are there. But right now, I would not say that the market is saturated because there's so many people who actually have call centers and voice-based, you know, representation of the representatives, both outbound and inbound, which still are, you know, not AI agents and that that whole thing can be much better. The cost of the business will be lesser. The error, because these are very repetitive at time tedious jobs, the error rates can go down.

• 16:40Speaker 00

  This churn in in terms of users, so productivity can be up because you don't have to train as much. At the same time, simple things like I'm sure Craig, you have experienced as you call a number like, all right, you are on the hold music, right? There is not going to be hold music anymore, right?

• 16:55Speaker 00

  So there there are those kind of things and I think that the world will look different if we just, you know, and I'm I'm sure that you will agree with me, in AI era that we are, world looks different, you know, six months from now, right? So there will be a different world, but right now, people are entering and it's validating the market and and it's not done yet. I actually feel very confident that we are building a solution which actually is at the edge of this, like this voice problem you Talking about like making human like voice possible out of the system is non trivial. Uh. There is a lot more psychology involved in it. We spend a lot of time making sure squeezing out every bit of latency out of it.

• 17:35Speaker 00

  So you feel like you're talking to a real person. There's a lot more dialogue management involved. This is why our proprietary models are there. So a lot of these, the science is going on and it's obviously, hopefully, you know, the, the fact that we're investing so much in the science will help us, you know, uh, see traction. And I think we are starting to see that and, and hope this quarter will quite growth continues, right? So, yeah.

• 17:59Craig S. Smith?

  Yeah. And then how, um, uh, do you think the market will then eventually coalesce around a few big players? Or, or, or do you think that companies will end up specializing in, uh, like there'll be the, uh, the dominant player for India or the dominant player for China, the dominant player for South America.

• 18:28Craig S. Smith?

  Or do you think one company will cover all that of and become dominant globally?

• 18:36Speaker 00

  Um, I don't know how it eventually will be, but I can tell you the, uh, current heterogeneity that exists and that heterogeneity, like somebody who covers that really well can become dominant. Or it can become the way the market gets segmented, right? So the biggest heterogeneity is that, you know, and this we have learned from our scaling that you really have to, uh, go in and, and, you know, be, uh, like, get into the business and deploy, right? It's not that you're from outside, you throw something over the wall and the business says, all right, let me go click, click, click and be done.

• 19:20Speaker 00

  Because if you make mistake, it actually affects the operations of the business. It affects the profitability and, and things like that, customer retention. So you have to go deep into it and, and deploy. And there's a lot of learning which is happening in that place. Like I will give you example, uh uh just to give you examples of how current technology cannot just be thrown over the wall. It tells me that you cannot today have a business which goes, all right, I got it, go everybody and and take it over. Here is an example, you you must know about RAG, right? Retrieval, Log, Vintage Generation. Yeah. Now, people take documents and they kind of stick that inside that. It's a very standard technique. However, if you're going inside the insurance business and you have put all the policies in there, right?

• 20:04Speaker 00

  It's not as trivial to tell a customer that, yes, your current claim is supported or not supported. Because if you make a mistake in your RAG, if you did not pull the right chunk out, you might end up denying something which should have been covered, or you might end up covering something which should have been excluded in the fine print. Right? So we invest a lot of time in understanding the structure of the document based on the domain that is, so that we can chunk it properly and we can retrieve it properly. Uh, you know, we even spend time in like rewriting the query to like, you know, when the person comes and says, well, uh, is this covered by policy? Then we know how exactly do we rewrite. Now, this is all domain-specific work.

• 20:43Speaker 00

  Now, the reason I give you this example is if you take all over the world, this work has, this is not easily generalizable. And and this is what we are seeing with everyone, not just in New York, like everybody else, is that they have to go in almost Palantir style, like they have to go into their customers', you know, building and kind of work with them. And that means the answer to your question that will there be one which will win all, I think, till technology gets to that place. And maybe, I mean, New York is building a platform called New Play, right? We will, we have a platform which will make it super easy for you to pick and choose. And when those kind of platforms start emerging, we are starting to offer one. Uh, and those become very capable, that is when we will be able to answer this question.

• 21:27Speaker 00

  I I it's hard to tell right now, but right now it's to keep the quality high, it's a lot of work. We have been able to scale the deployment and that is why we are pretty excited about it.

• 21:40Craig S. Smith?

  And and I'm just curious, your sales operation.

• 21:49Craig S. Smith?

  How do you figure out who might need the solution? And and then, uh, yeah, I mean, because there there are all these different sectors and all these different players. I mean, as you said, customer service, customer service alone is is a very fragmented market. So, yeah, I mean, what's the strategy? Do you go after one vertical? Do you, uh, do you sort of have a massive sales team that's calling everybody who could possibly need voice?

• 22:30Speaker 00

  Yeah, well, Craig, we are still a startup, right? So we are like, uh, maybe, uh, I think six months or a year. Uh, my involvement has been less than six months, but the thing is that we are still a startup. So there are two main things which I will say.

• 22:47Speaker 00

  You would not turn down a customer who comes in and says, we want your stuff, right? But at the same time, when reaching out, we got to go where momentum is and where we are actually like, uh, we have offensive customers. And we are seeing a lot of traction in in specifically like, you know, in e-commerce market uh, and and in financial services market. Uh, the example I gave you, insurance was a financial services and e-commerce is the example I gave you, the order return thing. Uh, we are seeing a lot of, uh, you know, uh, traction in there and and, you know, but I'm I'm thinking that given the growth pattern we are in, we are also not like, you know, that that doesn't mean that all our revenue is coming from just those two. Uh, you know, we are a startup. Why would we say no to more customers coming in, right? Like, scale as fast as you can.

• 23:32Speaker 00

  Uh, so that is, uh, the answer to your focus and sales question. But we are primarily seeing traction in two big markets. And, uh, I think, uh, we we are not in in a state where we can have like, you know, salespeople everywhere, our founders are actually knocking on the door, showing up in the conferences, talking to people, right? We don't, uh, you know, we are founder-led sales right now. So that is how we are. Yeah, yeah.

• 23:56Craig S. Smith?

  Yeah, uh, and and the dominant player currently seems to be, uh, Eleven Labs. Uh, how how?

• 24:07Craig S. Smith?

  Do you view them as a competitor? And how do you distinguish Nurix from Eleven Labs?

• 24:19Speaker 00

  So Eleven Labs is a great

• 24:23Speaker 00

  technology provider which we can use. The the basic idea is that they do a great job of generating human like voice for for any kind of parameters. Right? But if you notice, we start from like fielding a phone call and having a conversation. Now, in having a conversation, we have our own, you know, speech to text models with all the noise filtering and the dialogue management. But if we have to generate a response, we can use Eleven Labs. It depends like we we use Google's TTS as well. We can use Eleven Labs, we can use others. They're all technology choices to us.

• 25:05Speaker 00

  The overall approach is there is it that the problems have not been solved. We want to go solve them. And wherever they have been solved, we want to utilize them. So we can focus on the layers where things have been solved. So that is how we see it. We don't see it as like Eleven Labs as a competitor. It is actually a supplier and a partner, if you will.

• 25:21Craig S. Smith?

  I see, yeah, yeah, yeah. This is fascinating how these these new industries are developing and stratifying.

• 25:33Speaker 00

  Absolutely.

• 25:34Craig S. Smith?

  Yeah, yeah. So you you mentioned that we someday we won't have to listen to hold music. It still drives me crazy. Yesterday I was calling somebody.

• 25:48Craig S. Smith?

  And it's like, you know, listen closely as our menu options have changed. And you know, why is there a menu?

• 25:56Speaker 00

  Yeah.

• 25:58Craig S. Smith?

  So what's holding enterprises back? I was wondering, is it because these guys have bought a system?

• 26:09Craig S. Smith?

  And and, you know, they they want to wait until they've amortized it down to nothing? Or is it that?

• 26:20Craig S. Smith?

  Companies are just risk averse and they want to see how the the market evolves. I mean, what what will it take for those, what are they called? I I V R, I can't remember.

• 26:36Speaker 00

  Yes, yeah.

• 26:37Craig S. Smith?

  Yeah, for all of that to go away, where you call up and you speak to somebody. Uh, and you can't really tell whether it's a human or not, and they handle the call fluidly. What you you think that's a year away or 10 years away?

• 26:56Speaker 00

  Oh, I definitely don't think it is 10 years away. Uh, let me look at the amortization thing first, and then I can talk about what might hold somebody back. I don't think it's amortization is going to be the biggest issue because

• 27:12Speaker 00

  the savings, you know, not only the cost savings, but also reduction in errors and mistakes. Uh, increased opportunity of better cross selling and up selling when somebody calls in. And above all, all of this, because it is run by New X's AI, we can provide analytics saying, look, these are the kind of things which you could do with the calls. Today, call analytics is actually very, very hard, but we can do a much better job because everything is running through our digital system. So all of this will have enough business value that you will actually justify the new investment. So that is not the primary blocker, in my view.

• 27:56Speaker 00

  But you know, the primary blocker, I think, would be the businesses building confidence in these non deterministic systems. Right? In the I V R, it's like, you know what I mean, right? It's a very, very clear idea tree you are going through. And then there's like, okay, I can do a set of things and then a human comes in. But here, if you really wanted to remove all that tree, you would call and immediately to be like, hey, Craig, this is, you know, Bob from Target, right? If you're calling Target. And then it's like, okay, what do you want? Now you have a non-deterministic system. What are the expected outcomes? What are the the operating procedures? What are the guidelines and guardrails?

• 28:38Speaker 00

  And making sure that this system runs within that, they have to build that confidence. And this is one of the spaces, by the way, we invest a lot in. Like I talked about the rag, we have a lot of the simpler parts here are that, hey, you have a a set of backend services. Can we integrate with that? That is system level problem. We can integrate with that. That's easy. You know, MCP is available there, like out of box integrations, L L M's can do tool use, all that is there. But making sure that the L L M actually does not deny a claim, then there was actually a possibility or does not say, oh, yeah, go fill out this form without calling out that, no, you had an exclusion or a series of such things, or does not misunderstand an acronym, which is specific to a particular domain.

• 29:21Speaker 00

  These kind of things actually are problems which are being solved in the industry. We are investing a lot in that as well. And I think that that, because it's as new and being solved, there is a level of comfort which has to be built. But we are seeing a lot of traction in people trying it out, right? So the thing is that the businesses, once they try out and they see the difference, oh, yeah, this is great. My errors have gone down, my business has gone up, I'm able to transparently see what exactly is happening in this part of my call center. Now, naturally, all right, let's do more of those. So I think that that adoption will will happen. And fairly so, because this is their business. They don't want to risk that.

• 30:01Craig S. Smith?

  Yeah, yeah. And and so you guys build not only the voice generator,

• 30:10Craig S. Smith?

  you you you build the.

• 30:15Craig S. Smith?

  Well, yeah, describe again. I mean, you're not simply voice generation. You you build a higher level solution. So it describes sort of end to end what what that solution is.

• 30:29Speaker 00

  Yeah, absolutely. So now imagine that imagine you are a retailer. You know, and you say that, hey, I would like to try out, you know, your service for, you know, people who call in to find out where their order is or want to cancel their order or like, you know, the status of the order, right? Very simple three things. And you have a backend, you have order management system, which has, let's say, HTTP interface or some other sort of internal interface. So what we do is first is that we and we say, OK, what kind of endpoints will you have? Will you allow people to, you know, call you in? Do they want to check in through a chat bot on your website? Or do you want to like allow them to check on WhatsApp or SMS or whatever? Those are the ways they can talk to you.

• 31:14Speaker 00

  Now, the when they call the calling to the voice is the most technically savvy part, of course, here because of the human thing we talked about. When you call in, are the the phone lines obviously end up with us. And then we immediately, you know, figure out what you are saying using our own custom speech to text model. Right now, and nobody in the industry, not just us, is using a multimodal native voice to voice model because those things are harder to engineer to increase the precision which we want in an enterprise setting. They will get there, we will get there with them. But right now, so we quickly take whatever you have said, we filter out the noise, we understand who's talking, and we bring it to a text thing.

• 31:57Speaker 00

  But now you are sitting inside a dialogue management system where actually it knows that somebody has called in. And then whatever you have said goes on to the, you know, essentially the agent brain. But that agent brain is pulling in context from the RAG system. Now the RAG system, there are a series of techniques of RAGs of like, you know, chunking and, and, you know, querying different kind of things, which depends on the domain which you are in, we, we apply. Uh, in this example, uh, probably we, uh, pull policies regarding, you know, cancellation and return, which is, uh, you know, very custom chunking done depending on the documents you have given us, like, okay, where exactly, uh, you know, these policies are. And now it goes into the RAG system, pulls that up, and then it goes to the, uh, model and which actually kind of,

• 32:41Speaker 00

  uh, then does a tool call, gets the, uh, status of the order, makes the decision what has to be done and responds. When it responds, then we actually, the dialogue manager knows now that the model wants to respond. So the dialogue manager turns that into, uh, voice and sends it up. But in the meantime, suppose there's an interruption which is happening from the user, is the dialogue manager sitting in there, listening to the user, which actually says, all right, wait a second, or, you know, passes on the interruption, whatever it is. So that is the end to end in a very simplistic way, uh, how this whole thing works. Uh, now the complexity on the front end, we already talked about, like, you know, how you have to be really human-like. And on the back end, you really have to, uh, you know, stick to what the business wants.

• 33:23Speaker 00

  Right? And that is where a lot of that engineering also goes. So those are the ends of it.

• 33:29Craig S. Smith?

  Yeah. Uh, and then, uh, if, if the model can't answer, it hands off to a human.

• 33:37Speaker 00

  Yes, if a model cannot answer, it can always deflect and say, all right, well, you know, it's like how tier one versus tier two support can happen, right? So suppose you end up asking something else that, hey, uh, yeah, yeah, cancel that and you cancel it. And so this says that, but I want a different replacement, right? Well, suppose the model is not designed to take orders and say, okay, all right, let me get you to somebody. Yeah, obviously.

• 33:59Craig S. Smith?

  Yeah. And, and, and, uh, is this, uh, uh, on that handoff, then is it going to a BPO? A call center?

• 34:09Speaker 00

  Uh, it, it put, I mean, the assumption is that the businesses already have a BPO. Right? Uh, because the call center is where these phone calls were coming in before. Uh, so that already, uh, exists and the handoff will be structured accordingly. Uh, you could always, you come up with other ways of handoff, like, hey, let me, text you a link to our app where you can perform this app or something like that. Yeah.

• 34:36Craig S. Smith?

  Yeah, on the on that handoff piece, I had a company on a while ago called Crescendo. I don't know if you know them.

• 34:45Speaker 00

  Yeah, I know Crescendo, yeah.

• 34:47Craig S. Smith?

  Oh, do you? Yeah.

• 34:48Speaker 00

  Yeah, I know a lot of them from like prior life.

• 34:51Craig S. Smith?

  Oh, and then they talk a lot about Sierra as being a big competitor. Crescendo, their thesis is that that handoff is critical. And it's difficult unless everybody's under one roof. So they, they, you know, they bought Product Hero and they have their own call centers that are trained with their AI. So everyone knows what's what's happening.

• 35:29Craig S. Smith?

  And indeed, I've run into cases where, you know, you talk to an agent, a voice agent, and you or certainly on text, you give it all your information, describe the problem. And then you get transferred to a live agent and they start from the very beginning. Yes, you know, it's frustrating. So how do you avoid that problem?

• 35:59Speaker 00

  Yeah, actually, this is very interesting because the thing that you said is not introduced by AI. This happens human to human transfer as well. You know, I was talking to Xfinity once about something which was going on and they kept transferring me. And every new person started from the top saying, oh, let me verify you are a customer. I'm like, I'm talking to a sixth person transferred inside. So this is a systems problem, which is like when you get transferred. So the question really is that when the handoff is happening, is the context being passed around? And the duty which I feel is that if you design the system right, the model will always do what you have, you know, told it to do, right?

• 36:39Speaker 00

  So if you say that Ben Handoff, prepare, you know, a three-page summary and pass it on to the human. Then on the human screen, if the three-page summary is this is what has been done so far. Then that experience which you get is really based on the competency of the human agent who you're talking to. Have they read through that or not? So I, I think that this is a systems problem. This is not a AI to human problem because we see this in human to human as well. And it's how you have structured it. So, and we are very capable of like, you know, generating any kind of summary for any kind of call. As I mentioned earlier, we do analytics just for internal consumptions as well of how these things are. And being embedded in the team, it actually helps us also, you know, make sure that all these

• 37:24Speaker 00

  connections are, or the impedance mismatch is actually handled really well.

• 37:29Craig S. Smith?

  Yeah, yeah. And, and what was it about this group that gave you enough confidence to join the board?

• 37:39Speaker 00

  Well, you know, it is fascinating because these days the thing which I end up feeling which makes a winning combination is, is two things.

• 37:52Speaker 00

  One is very, very strong clarity about what is the problem which the team wants to solve. Right? That clarity that this problem, any of the problem has to be worth solving, but that clarity rather than there's so many things possible that you can go try everything. Like we talked about like 11 large, you know, not being a competitor, but something we could use. It comes from a clarity that, hey, this is the problem you're solving. Let's use everything which everybody in the world is doing for other problems. So very strong clarity, I felt that they had amazing clarity and it is starting to show in the traction they're getting on the two sides. And the second thing is the quality of the, you know, the, the team, technical and sales.

• 38:33Speaker 00

  You know, it's not surprising when these guys are based out of Bangalore and, and you know, i- i- the Bangalore as an ecosystem. Has started to finally spin up. I mean, you know, it's it's not it's second to Silicon Valley. I mean, it's it's not Silicon Valley, but second only to Silicon Valley, right? And I saw amazing collection of these, like, you know, whatever engineers they are. When I was talking to them, I I was like, wow, this is really the bar. Getting such a collection of people in one place in Silicon Valley would actually take a significant effort. And I felt like you are focused on the problem and you have the right people in your camp. So, you know, it's all about execution now.

• 39:15Speaker 00

  So I think that that is what actually, you know, gave me that confidence.

• 39:21Craig S. Smith?

  Yeah. Can you talk at all about the other companies coming out of the incubator? Or or do you really want to restrict yourself to NERX?

• 39:32Speaker 00

  Uh, give me a few months. I think the other one is like just and not not for any other reason. I'm not being coy. It's because just today, as a matter of fact, I came out of a two hour conversation of like, should we do this or should we do that? So we are in the like, what problem should we solve phase of it right now?

• 39:50Speaker 00

  I I think that we will sometime this week be able to at least put a stake in the ground and then start building a prototype. So I give me give me a few months. I will. I promise I'll be back.

• 40:02Craig S. Smith?

  Yeah. And how does someone use NERX? I mean, it's it's a SaaS product. Do they, is there a, I mean, how is it priced? Is there a self-serve element to it on on the web?

• 40:22Speaker 00

  Yeah, if you go to the web, you can experience the the real human likeness of the voice experience, which I was talking about. By the way, I think we are the only voice agent company in the world which puts it out on the web because we are so confident we have the best solution in the world. So you can go and you can experience like you can say, hey, like, I don't know, a collections agent who's calling you saying, hey, I'm going to pay your credit card bill. And you can have a conversation saying, oh, you know, I'm living paycheck to paycheck. What do I do now? You can like talk through and it actually will have a conversation with you. So that is one way, but but as a business, we are actually gonna launch our our platform.

• 41:02Speaker 00

  So far, all the deals are like, you know, we go in, as I said, like, you know, forward deployed engineers go in and kind of build things out. That's how we have gotten all the deployments so far. Now, but we have been taking that learning and we are building a platform thereby in which you will see, I think, when the demo happens, maybe they'll be able to give you a demo of, you know, the platform itself, a sneak peek. But you can choose various kind of agent templates which we have, like, you know, you want a customer service agent versus a sales agent. Like, you can choose various templates, or you can create your own, like you say, okay, this is my agent, and this agent will connect to these particular endpoints. This is my RAG setup, and this is my, you know, we connect to one hundred fifty different kind of tools in the backend.

• 41:46Speaker 00

  So these are my hundred tools, or you can create your own custom MCP server which we can connect to and, you know, all of that. And then you can go try it out. That you can try out, and of course, you know, there is a, this is something we have learned and a lot of industries learning is like, there's a lot of work which has to be done between standing up an agent and making it run in production and be, you know, face of your business. So, but we do all of that, and the the platform you will see definitely will give you a sense of how to stand one up.

• 42:16Craig S. Smith?

  Yeah, and then to integrate it into another solution, it's an API call.

• 42:23Speaker 00

  To integrate into another solution, it is gonna be an API call. We have like one hundred fifty different type of solutions, ticketing systems and all, which we integrated right out of the bat. But we also support MCP protocols, so if you have a different service and you have a MCP server which can be called by the LLM, we will do that, or if you want a custom solution, then we can do code generation and we can call that API that way.

• 42:49Craig S. Smith?

  Yeah. And is it, how expensive is it? I'm just thinking for myself, I'm building a chatbot so you can ask about all of the podcasts that I've had.

• 43:04Speaker 00

  Yeah, so that is something which maybe the sales team can tell you specifically depending on your case. Right now, we are doing it deal by deal. So depending on how involved your product is and, you know, they'll be, I'm happy to connect you with somebody there. It's, you know, it's not going to be super expensive. We are, I mean, this is the thing which we are seeing that having a human on the other side is more expensive. And also more error prone sometimes, but I'm happy to get you, you know, your, your exact scenario looked at. I think you will, you will, you will like it.

• 43:44Craig S. Smith?

  Yeah, no, I'll, I'll go on the website after this and take a look.

• 43:48Speaker 00

  Yeah, please.

• 43:50Speaker 00

  And let me know if there's something which you feel is not working or anything like, you know, I would love any feedbacks. I can pass it on.

• 43:56Craig S. Smith?

  Is there anything that I haven't asked that you think listeners should know?

• 44:01Speaker 00

  No, I think, I mean, you know, the biggest thing which I will leave you with is, which you will see for yourself, that it's, it's amazing. A voice experience and we've invested a lot in like making sure it is human like and low latency. And the other thing is it's proven technology in the sense that we have, you know, over quarter million calls happening through it right now. And, you know, therefore the company is growing at a very rapid clip. So I don't have a whole lot to add. It's amazing technology and these are amazing times we live in. I mean, even scaling the thing which I was talking about, like just seeing how this thing needs to scale is like fascinating technical problems. So we are having a lot of fun. Yeah.

• 44:47Craig S. Smith?

  Yeah, no, it's an exciting world. So I'm looking forward to the day when, when everyone is using Nurix and I.

• 44:56Speaker 00

  Thank you. So am I. Yeah. No more hold music and telling another agent all your history as you get transferred.