5 Voice Agent Failure Modes You'll Hit in Week One — Venky B, Plivo

Let's just do a couple of quick Let's just do a couple of quick questions and then we'll jump right in. questions and then we'll jump right in. questions and then we'll jump right in. Uh, how many of us in the room here have Uh, how many of us in the room here have Uh, how many of us in the room here have built voice AI agents? built voice AI agents? built voice AI agents? Okay, that's a that's a pretty good Okay, that's a that's a pretty good Okay, that's a that's a pretty good audience here. And how many of you guys audience here. And how many of you guys audience here. And how many of you guys have built AI agents that have been have built AI agents that have been have built AI agents that have been deployed in production? Not bad. Okay, cool. So, uh we'll talk Not bad. Okay, cool. So, uh we'll talk about what typically happens, right? about what typically happens, right? about what typically happens, right? Like everyone's talking about wise AI Like everyone's talking about wise AI Like everyone's talking about wise AI agents. Uh the agents. Uh the agents. Uh the you know, one pill solution to pretty you know, one pill solution to pretty you know, one pill solution to pretty much everything in the world today uh is much everything in the world today uh is much everything in the world today uh is is wise agents. So, everyone's building is wise agents. So, everyone's building is wise agents. So, everyone's building one and trying to deploy that. They one and trying to deploy that. They one and trying to deploy that. They sound great when you're sort of building sound great when you're sort of building sound great when you're sort of building that in your dev sort of landscape and that in your dev sort of landscape and that in your dev sort of landscape and then the moment you take this to from a then the moment you take this to from a then the moment you take this to from a proof of concept to production things proof of concept to production things proof of concept to production things start failing. Uh so we'll walk through start failing. Uh so we'll walk through start failing. Uh so we'll walk through these five different angles of like how these five different angles of like how these five different angles of like how uh or what we have seen uh at PO with uh or what we have seen uh at PO with uh or what we have seen uh at PO with VIA agents but just before that a quick VIA agents but just before that a quick VIA agents but just before that a quick uh intro from from my side uh I am Wenke uh intro from from my side uh I am Wenke uh intro from from my side uh I am Wenke the founder and CEO uh she used the the founder and CEO uh she used the the founder and CEO uh she used the title agent engineering manager u I'm title agent engineering manager u I'm title agent engineering manager u I'm calling myself chief agent officer uh calling myself chief agent officer uh calling myself chief agent officer uh from a from a title standpoint u okay so from a from a title standpoint u okay so from a from a title standpoint u okay so what is what is uh you know why why are what is what is uh you know why why are what is what is uh you know why why are even qualified for this this discussion even qualified for this this discussion even qualified for this this discussion and and like uh what what are we seeing and and like uh what what are we seeing and and like uh what what are we seeing that a lot of companies don't get to that a lot of companies don't get to that a lot of companies don't get to see? I I'll talk a bit about our journey see? I I'll talk a bit about our journey see? I I'll talk a bit about our journey in terms of like how we've uh come along

in terms of like how we've uh come along in terms of like how we've uh come along so far and then jump right in. Uh you so far and then jump right in. Uh you so far and then jump right in. Uh you know we were we've been around for about know we were we've been around for about know we were we've been around for about 14 years. Our journey has been a 14 years. Our journey has been a 14 years. Our journey has been a developer API platform and then now an developer API platform and then now an developer API platform and then now an uh you know an AI agent business. We uh you know an AI agent business. We uh you know an AI agent business. We started with voice and SMS APIs back in started with voice and SMS APIs back in started with voice and SMS APIs back in the day uh 2011 and then uh you know now the day uh 2011 and then uh you know now the day uh 2011 and then uh you know now we are primarily focused on our AI agent we are primarily focused on our AI agent we are primarily focused on our AI agent offering uh the the full stack on our offering uh the the full stack on our offering uh the the full stack on our platform. We we see over a billion voice platform. We we see over a billion voice platform. We we see over a billion voice calls each month across the globe. Uh calls each month across the globe. Uh calls each month across the globe. Uh and which is where we've seen a lot of and which is where we've seen a lot of and which is where we've seen a lot of these uh you know patterns emerge in these uh you know patterns emerge in these uh you know patterns emerge in terms of like how when we work with our terms of like how when we work with our terms of like how when we work with our customers what happens on their voice customers what happens on their voice customers what happens on their voice agents in in production. Uh we're a uh agents in in production. Uh we're a uh agents in in production. Uh we're a uh 90 member team and uh we've we have uh 90 member team and uh we've we have uh 90 member team and uh we've we have uh 50 million funding in the bank. Fun 50 million funding in the bank. Fun 50 million funding in the bank. Fun fact, this is not from external VC fact, this is not from external VC fact, this is not from external VC investors. This is all from being a investors. This is all from being a investors. This is all from being a profitable company having put that cash profitable company having put that cash profitable company having put that cash in the bank over over these years. Uh in the bank over over these years. Uh in the bank over over these years. Uh some customers we we power across the some customers we we power across the some customers we we power across the globe. Uh you know, we've just left some globe. Uh you know, we've just left some globe. Uh you know, we've just left some some logos in there. But primarily from some logos in there. But primarily from some logos in there. But primarily from an offering standpoint, u I would sort an offering standpoint, u I would sort an offering standpoint, u I would sort of cohort this into three different of cohort this into three different of cohort this into three different buckets. One is a programmable AI agent buckets. One is a programmable AI agent buckets. One is a programmable AI agent offering. We call it uh I mean it's a offering. We call it uh I mean it's a offering. We call it uh I mean it's a speech pipeline, not a true speech pipeline, not a true speech pipeline, not a true speech-to-pech product yet, but that's a speech-to-pech product yet, but that's a speech-to-pech product yet, but that's a that's a programmable offering. We also that's a programmable offering. We also that's a programmable offering. We also have an AI agent studio. It's a no code have an AI agent studio. It's a no code have an AI agent studio. It's a no code visual uh builder. And then we like I visual uh builder. And then we like I visual uh builder. And then we like I said, we started with voice APIs. So we said, we started with voice APIs. So we said, we started with voice APIs. So we obviously have built this out over the obviously have built this out over the obviously have built this out over the last 14 years, the SIP trunking and the

last 14 years, the SIP trunking and the last 14 years, the SIP trunking and the audio streaming layers. So we don't rely audio streaming layers. So we don't rely audio streaming layers. So we don't rely on other folks for the telefony or the on other folks for the telefony or the on other folks for the telefony or the carrier layer. Like that's the carrier layer. Like that's the carrier layer. Like that's the breadandbut business we've built over breadandbut business we've built over breadandbut business we've built over all these years and and that's on top of all these years and and that's on top of all these years and and that's on top of which our AI uh agent platform sits. which our AI uh agent platform sits. which our AI uh agent platform sits. Okay, with that uh let's get into this, Okay, with that uh let's get into this, Okay, with that uh let's get into this, right, which I'm I'm sure since you guys right, which I'm I'm sure since you guys right, which I'm I'm sure since you guys have all built AI agents, you've all have all built AI agents, you've all have all built AI agents, you've all seen this or you know built this in in seen this or you know built this in in seen this or you know built this in in one manner or another and we'll spend one manner or another and we'll spend one manner or another and we'll spend more time on this in terms of like how more time on this in terms of like how more time on this in terms of like how uh the entire pipeline looks, right? Uh uh the entire pipeline looks, right? Uh uh the entire pipeline looks, right? Uh what we see with customers is and and what we see with customers is and and what we see with customers is and and I'm sure you guys can all relate to this I'm sure you guys can all relate to this I'm sure you guys can all relate to this is you know anyone thinking about AI is you know anyone thinking about AI is you know anyone thinking about AI agents what they do is they pick a bunch agents what they do is they pick a bunch agents what they do is they pick a bunch of these orchestration frameworks and of these orchestration frameworks and of these orchestration frameworks and they do a pretty good job live kit or a they do a pretty good job live kit or a they do a pretty good job live kit or a pipecat you know build their AI oen on pipecat you know build their AI oen on pipecat you know build their AI oen on top of that uh they think they can just top of that uh they think they can just top of that uh they think they can just sort of orchestrate these different four sort of orchestrate these different four sort of orchestrate these different four layers speechtoext lm uh and and TTS layers speechtoext lm uh and and TTS layers speechtoext lm uh and and TTS with turn detection in between and we're with turn detection in between and we're with turn detection in between and we're off to the races like my AI engine agent off to the races like my AI engine agent off to the races like my AI engine agent works in a in a P and it's good to work works in a in a P and it's good to work works in a in a P and it's good to work in production. Uh typically that's what in production. Uh typically that's what in production. Uh typically that's what happens. They sort of measure their happens. They sort of measure their happens. They sort of measure their latencies and you can see some latencies and you can see some latencies and you can see some indicative latencies on on this slide at indicative latencies on on this slide at indicative latencies on on this slide at at each layer and they're like yeah this at each layer and they're like yeah this at each layer and they're like yeah this this uh seems good for me for what I this uh seems good for me for what I this uh seems good for me for what I need. So let let's position production need. So let let's position production need. So let let's position production and then the the production w uh sort of and then the the production w uh sort of and then the the production w uh sort of start to kick in and and you see all start to kick in and and you see all start to kick in and and you see all sort of failure modes which we are going sort of failure modes which we are going sort of failure modes which we are going to spend you know most of the time on on to spend you know most of the time on on to spend you know most of the time on on in this talk at least. Uh I've kept some

in this talk at least. Uh I've kept some in this talk at least. Uh I've kept some time at the end for Q&A if you guys want time at the end for Q&A if you guys want time at the end for Q&A if you guys want to have uh you know questions but we'll to have uh you know questions but we'll to have uh you know questions but we'll jump right in from from this to uh you jump right in from from this to uh you jump right in from from this to uh you know different failure modes we see. know different failure modes we see. know different failure modes we see. Let's start with Let's start with Let's start with you know the first one which everyone you know the first one which everyone you know the first one which everyone talks about like this is the most spoken talks about like this is the most spoken talks about like this is the most spoken about failure mode which is latency. U I about failure mode which is latency. U I about failure mode which is latency. U I think we have a few AI agent talks today think we have a few AI agent talks today think we have a few AI agent talks today or AI agent talks today. Um I'm pretty or AI agent talks today. Um I'm pretty or AI agent talks today. Um I'm pretty sure like everyone everyone's going to sure like everyone everyone's going to sure like everyone everyone's going to touch upon this specific failure mode touch upon this specific failure mode touch upon this specific failure mode which is why I'm bringing this right up which is why I'm bringing this right up which is why I'm bringing this right up uh in in terms of uh you know some like uh in in terms of uh you know some like uh in in terms of uh you know some like how this entire experience is for uh how this entire experience is for uh how this entire experience is for uh users right uh typically most folks users right uh typically most folks users right uh typically most folks measure this by time to first audio so measure this by time to first audio so measure this by time to first audio so the time when you user stop speaking to the time when you user stop speaking to the time when you user stop speaking to your agent starts speaking right and I your agent starts speaking right and I your agent starts speaking right and I think you've you've probably seen this think you've you've probably seen this think you've you've probably seen this if you guys have built voice agents on, if you guys have built voice agents on, if you guys have built voice agents on, you know, what uh good or natural feels you know, what uh good or natural feels you know, what uh good or natural feels like, what uh sort of annoying feels like, what uh sort of annoying feels like, what uh sort of annoying feels like or noticeable feels like, and then like or noticeable feels like, and then like or noticeable feels like, and then what annoying feels like, which is, you what annoying feels like, which is, you what annoying feels like, which is, you know, different tiered steps. Uh we know, different tiered steps. Uh we know, different tiered steps. Uh we notice, you know, most people want to be notice, you know, most people want to be notice, you know, most people want to be under 550 cuz that's what's advertised under 550 cuz that's what's advertised under 550 cuz that's what's advertised by, you know, platforms or uh you know, by, you know, platforms or uh you know, by, you know, platforms or uh you know, solutions or or or or layers. But I solutions or or or or layers. But I solutions or or or or layers. But I think most end up between 750 to 1.2. uh think most end up between 750 to 1.2. uh think most end up between 750 to 1.2. uh that's where most of the folks end up that's where most of the folks end up that's where most of the folks end up at. Uh the really bad performing ones at. Uh the really bad performing ones at. Uh the really bad performing ones end up you know more than 1.2 and then end up you know more than 1.2 and then end up you know more than 1.2 and then you start to see users uh hang up. Uh you start to see users uh hang up. Uh you start to see users uh hang up. Uh now I I'll share with you like what

now I I'll share with you like what now I I'll share with you like what we've seen practically in in uh we've seen practically in in uh we've seen practically in in uh production with uh customers using this production with uh customers using this production with uh customers using this with at at different layers and then you with at at different layers and then you with at at different layers and then you know solutions to uh some of these. The know solutions to uh some of these. The know solutions to uh some of these. The way we want to think about this layer is way we want to think about this layer is way we want to think about this layer is sort of a balance between these three sort of a balance between these three sort of a balance between these three which is cost, intelligence and latency, which is cost, intelligence and latency, which is cost, intelligence and latency, right? And and and why do I bring these right? And and and why do I bring these right? And and and why do I bring these three up? Because they're sort of three up? Because they're sort of three up? Because they're sort of interrelated. I think one of the things interrelated. I think one of the things interrelated. I think one of the things I was just chatting with uh you know a I was just chatting with uh you know a I was just chatting with uh you know a couple of folks outside one of the couple of folks outside one of the couple of folks outside one of the things last one year we've seen lot of things last one year we've seen lot of things last one year we've seen lot of innovations lot of intelligence spike on innovations lot of intelligence spike on innovations lot of intelligence spike on the LLM side of uh things right and most the LLM side of uh things right and most the LLM side of uh things right and most of the you know intelligence has come in of the you know intelligence has come in of the you know intelligence has come in in terms of thinking or uh you know in terms of thinking or uh you know in terms of thinking or uh you know reinforcement learning and and so on and reinforcement learning and and so on and reinforcement learning and and so on and so forth the irony with voice agents is so forth the irony with voice agents is so forth the irony with voice agents is like almost always your the the LLM or like almost always your the the LLM or like almost always your the the LLM or the agent that's talking has to have the agent that's talking has to have the agent that's talking has to have thinking turned thinking turned thinking turned Right. So all the advancements we've had Right. So all the advancements we've had Right. So all the advancements we've had in the LLM layer in the last one year in the LLM layer in the last one year in the LLM layer in the last one year like none of that even apply here now.

like none of that even apply here now. like none of that even apply here now. Right? You obviously you have you know Right? You obviously you have you know Right? You obviously you have you know better models that can do you know better models that can do you know better models that can do you know better instruction following or tool better instruction following or tool better instruction following or tool calling but pretty much all of your calling but pretty much all of your calling but pretty much all of your intelligence that's been built in on the intelligence that's been built in on the intelligence that's been built in on the thinking layer is all off by default if thinking layer is all off by default if thinking layer is all off by default if you want it to be fast enough. So so you want it to be fast enough. So so you want it to be fast enough. So so that's one of the ironies that we come that's one of the ironies that we come that's one of the ironies that we come up with. So then how do you sort of up with. So then how do you sort of up with. So then how do you sort of balance intelligent cost and latency? balance intelligent cost and latency? balance intelligent cost and latency? Let's let's look at some of these uh you Let's let's look at some of these uh you Let's let's look at some of these uh you know options uh that are out there in know options uh that are out there in know options uh that are out there in the market right so and I'm specifically the market right so and I'm specifically the market right so and I'm specifically picking LLM because if you looked at the picking LLM because if you looked at the picking LLM because if you looked at the previous chart LLM is u you know sort of previous chart LLM is u you know sort of previous chart LLM is u you know sort of your highest latency bucket that adds to your highest latency bucket that adds to your highest latency bucket that adds to this right and uh if you look at you this right and uh if you look at you this right and uh if you look at you know frontier models which I think most know frontier models which I think most know frontier models which I think most folks start by default your your openi folks start by default your your openi folks start by default your your openi your clouds your geminis u you know p50 your clouds your geminis u you know p50 your clouds your geminis u you know p50 ttfftd is roughly around 450 to 500 on ttfftd is roughly around 450 to 500 on ttfftd is roughly around 450 to 500 on on a good day and it can get spiky, on a good day and it can get spiky, on a good day and it can get spiky, right? It can it can uh you know P90 P95 right? It can it can uh you know P90 P95 right? It can it can uh you know P90 P95 can go easily upwards of 1.2 1.3 seconds can go easily upwards of 1.2 1.3 seconds can go easily upwards of 1.2 1.3 seconds even uh and and that's not good for the even uh and and that's not good for the even uh and and that's not good for the overall agent experience.

overall agent experience. overall agent experience. So so that so that's your frontier So so that so that's your frontier So so that so that's your frontier model. Now there's another options which model. Now there's another options which model. Now there's another options which is your your cerebrus or or the gro that is your your cerebrus or or the gro that is your your cerebrus or or the gro that is famous and popular for spitting out a is famous and popular for spitting out a is famous and popular for spitting out a lot of tokens or or tokens very fast, lot of tokens or or tokens very fast, lot of tokens or or tokens very fast, right? uh these work but for you to get right? uh these work but for you to get right? uh these work but for you to get dedicated latency or time to first token dedicated latency or time to first token dedicated latency or time to first token on these you need dedicated capacity and on these you need dedicated capacity and on these you need dedicated capacity and that is really expensive that's where I that is really expensive that's where I that is really expensive that's where I spoke about the cost uh as as being one spoke about the cost uh as as being one spoke about the cost uh as as being one of the things to balance right it's of the things to balance right it's of the things to balance right it's really expensive and then like you talk really expensive and then like you talk really expensive and then like you talk to anyone from the gro team or the to anyone from the gro team or the to anyone from the gro team or the cerebrus team they'll tell you you need cerebrus team they'll tell you you need cerebrus team they'll tell you you need to book 12 months in advance for to book 12 months in advance for to book 12 months in advance for dedicated capacity they're booked out dedicated capacity they're booked out dedicated capacity they're booked out for the next 12 months so so that's for the next 12 months so so that's for the next 12 months so so that's that's a pretty expensive option and that's a pretty expensive option and that's a pretty expensive option and then you really need to be sure that the then you really need to be sure that the then you really need to be sure that the model you're deploying on some of these model you're deploying on some of these model you're deploying on some of these infra layers uh will be here 12 months infra layers uh will be here 12 months infra layers uh will be here 12 months from now and and it's a it's a big from now and and it's a it's a big from now and and it's a it's a big investment and a big unknown. So, so investment and a big unknown. So, so investment and a big unknown. So, so what's a realistic option for production what's a realistic option for production what's a realistic option for production grade uh grade uh grade uh agents that are that are good quality agents that are that are good quality agents that are that are good quality and end up balancing uh three of these u and end up balancing uh three of these u and end up balancing uh three of these u this is what has worked for us u which this is what has worked for us u which this is what has worked for us u which is the open source models u there are is the open source models u there are is the open source models u there are obviously a lot of them in terms of like obviously a lot of them in terms of like obviously a lot of them in terms of like the variety and and variations you can the variety and and variations you can the variety and and variations you can pick I'm specifically talking about the pick I'm specifically talking about the pick I'm specifically talking about the two we work with u quen 3.5 and gemma two we work with u quen 3.5 and gemma two we work with u quen 3.5 and gemma four. These are uh you know kind of four. These are uh you know kind of four. These are uh you know kind of cutting edge open source models right uh cutting edge open source models right uh cutting edge open source models right uh out in the market right now and we've out in the market right now and we've out in the market right now and we've done a lot of benchmarking around this

done a lot of benchmarking around this done a lot of benchmarking around this in how they work. It it can be scary to in how they work. It it can be scary to in how they work. It it can be scary to think like okay I have the models now I think like okay I have the models now I think like okay I have the models now I have to host them you know run them on have to host them you know run them on have to host them you know run them on my own GPUs and so on and so forth but my own GPUs and so on and so forth but my own GPUs and so on and so forth but if you are consistently targeting under if you are consistently targeting under if you are consistently targeting under 300 ms u this we've seen this to be a a 300 ms u this we've seen this to be a a 300 ms u this we've seen this to be a a great option to balance between latency great option to balance between latency great option to balance between latency cost and intelligence now some more deep cost and intelligence now some more deep cost and intelligence now some more deep dive here if you're doing only English dive here if you're doing only English dive here if you're doing only English uh quen 3.5 or GMA both work fine but if uh quen 3.5 or GMA both work fine but if uh quen 3.5 or GMA both work fine but if you're doing multilingual uh right you're doing multilingual uh right you're doing multilingual uh right international audiences different international audiences different international audiences different languages uh Gemma 4 is a much better languages uh Gemma 4 is a much better languages uh Gemma 4 is a much better model for that uh we've seen uh token model for that uh we've seen uh token model for that uh we've seen uh token fertility evals essentially what that fertility evals essentially what that fertility evals essentially what that means is if if I were to dejargonize means is if if I were to dejargonize means is if if I were to dejargonize that is like how many tokens does it that is like how many tokens does it that is like how many tokens does it take to generate one word in that take to generate one word in that take to generate one word in that language okay so Gemma is much much language okay so Gemma is much much language okay so Gemma is much much better at least 2.5 to 3x better than better at least 2.5 to 3x better than better at least 2.5 to 3x better than quen 3.5 from that perspective so your quen 3.5 from that perspective so your quen 3.5 from that perspective so your time to words is much faster on Gemma or time to words is much faster on Gemma or time to words is much faster on Gemma or everything else equal right on a on a everything else equal right on a on a everything else equal right on a on a multilingual basis. Now what sizes do multilingual basis. Now what sizes do multilingual basis. Now what sizes do you pick at the LLM layer? Uh the you pick at the LLM layer? Uh the you pick at the LLM layer? Uh the mixture of expert usually works fine. Uh mixture of expert usually works fine. Uh mixture of expert usually works fine. Uh the three or four billion mixture of the three or four billion mixture of the three or four billion mixture of expert usually works fine. The the expert usually works fine. The the expert usually works fine. The the problem with mixture of expert is like problem with mixture of expert is like problem with mixture of expert is like if anyone goes down wants to go down the if anyone goes down wants to go down the if anyone goes down wants to go down the direction of fine-tuning that can be a direction of fine-tuning that can be a direction of fine-tuning that can be a challenge uh because fine-tuning mixture challenge uh because fine-tuning mixture challenge uh because fine-tuning mixture of experts models are not easy. Uh you of experts models are not easy. Uh you of experts models are not easy. Uh you can end up breaking the model uh a a lot can end up breaking the model uh a a lot can end up breaking the model uh a a lot of times. So, so that's one challenge we

of times. So, so that's one challenge we of times. So, so that's one challenge we see with Make sure experts, but usually see with Make sure experts, but usually see with Make sure experts, but usually out of the box, it gets you 90% closer out of the box, it gets you 90% closer out of the box, it gets you 90% closer to where you want to be like even to where you want to be like even to where you want to be like even without any fine-tuning or or or custom without any fine-tuning or or or custom without any fine-tuning or or or custom work done on the model. Uh so that's the work done on the model. Uh so that's the work done on the model. Uh so that's the advantage of mixer experts. Uh now, if advantage of mixer experts. Uh now, if advantage of mixer experts. Uh now, if you want to fine-tune and and you you you want to fine-tune and and you you you want to fine-tune and and you you want to go deeper and say like look, I'm want to go deeper and say like look, I'm want to go deeper and say like look, I'm working for a specific domain, working for a specific domain, working for a specific domain, healthcare, what have you, right? uh and healthcare, what have you, right? uh and healthcare, what have you, right? uh and I want to make sure I I'm able to I want to make sure I I'm able to I want to make sure I I'm able to fine-tune my model. You want to start at fine-tune my model. You want to start at fine-tune my model. You want to start at least with uh the 8 billion 12 billion least with uh the 8 billion 12 billion least with uh the 8 billion 12 billion at least uh from where we are today. at least uh from where we are today. at least uh from where we are today. Maybe maybe six months from now a 4 Maybe maybe six months from now a 4 Maybe maybe six months from now a 4 billion 4 billion model beats the 8 billion 4 billion model beats the 8 billion 4 billion model beats the 8 billion model uh hands down. But for billion model uh hands down. But for billion model uh hands down. But for today uh what we've seen is you minimum today uh what we've seen is you minimum today uh what we've seen is you minimum need a 8 billion or 12 billion model. Uh need a 8 billion or 12 billion model. Uh need a 8 billion or 12 billion model. Uh cuz you're looking for two things in cuz you're looking for two things in cuz you're looking for two things in these models. One obviously fast tokens these models. One obviously fast tokens these models. One obviously fast tokens but uh good instruction following. Okay. but uh good instruction following. Okay. but uh good instruction following. Okay. And the second thing is like very high And the second thing is like very high And the second thing is like very high uh success ratio in tool calling because uh success ratio in tool calling because uh success ratio in tool calling because if you can do these two things well then if you can do these two things well then if you can do these two things well then you are on to like 70 80% there from not you are on to like 70 80% there from not you are on to like 70 80% there from not even having to fine-tune it fine-tune even having to fine-tune it fine-tune even having to fine-tune it fine-tune any model like models will work out of any model like models will work out of any model like models will work out of the box right u so so that's uh been our the box right u so so that's uh been our the box right u so so that's uh been our recipe we've actually uh we run two recipe we've actually uh we run two recipe we've actually uh we run two flavors one a fine tune model flavors one a fine tune model flavors one a fine tune model for specific industries and then for uh for specific industries and then for uh for specific industries and then for uh you know most generic use cases uh MOE you know most generic use cases uh MOE you know most generic use cases uh MOE model just works out of the box. Uh model just works out of the box. Uh model just works out of the box. Uh there are a few more tips and tricks there are a few more tips and tricks there are a few more tips and tricks we'll talk about in the upcoming slides we'll talk about in the upcoming slides we'll talk about in the upcoming slides where we see failure models, but but where we see failure models, but but where we see failure models, but but that's where we stand from a from a

that's where we stand from a from a that's where we stand from a from a latency LLM standpoint. Um all right, latency LLM standpoint. Um all right, latency LLM standpoint. Um all right, I'm running tight on time, so I'm going I'm running tight on time, so I'm going I'm running tight on time, so I'm going to fast track this. U now there are a to fast track this. U now there are a to fast track this. U now there are a couple of other flavors in this. Uh couple of other flavors in this. Uh couple of other flavors in this. Uh people build agents with a a mixture of people build agents with a a mixture of people build agents with a a mixture of models. What they do is you know for u models. What they do is you know for u models. What they do is you know for u the the talking part of it they have a the the talking part of it they have a the the talking part of it they have a conversational model which is a much conversational model which is a much conversational model which is a much lower smaller model and then you know lower smaller model and then you know lower smaller model and then you know maybe even a three billion model and maybe even a three billion model and maybe even a three billion model and then for tool calling they have a much then for tool calling they have a much then for tool calling they have a much larger model so that they have a larger model so that they have a larger model so that they have a improved tool calling success ratio improved tool calling success ratio improved tool calling success ratio there. Uh sorry the second one is uh assume your sorry the second one is uh assume your transcriptions are going to be brittle transcriptions are going to be brittle transcriptions are going to be brittle like that's that's uh something you want like that's that's uh something you want like that's that's uh something you want to sort of uh live by when you're to sort of uh live by when you're to sort of uh live by when you're building AI agents even if you have the building AI agents even if you have the building AI agents even if you have the best transcription engine out there and best transcription engine out there and best transcription engine out there and I I I'll show you why right like the the I I I'll show you why right like the the I I I'll show you why right like the the the state-of-the-art transcription the state-of-the-art transcription the state-of-the-art transcription engines out out in the market u you know engines out out in the market u you know engines out out in the market u you know sort of get you to four to 6% word error sort of get you to four to 6% word error sort of get you to four to 6% word error rate right and this is on known eval rate right and this is on known eval rate right and this is on known eval sets on real world noisy calls with you sets on real world noisy calls with you sets on real world noisy calls with you know sort of uh accents like people know sort of uh accents like people know sort of uh accents like people having different sort of accents uh having different sort of accents uh having different sort of accents uh domain vocabulary and so on and so forth domain vocabulary and so on and so forth domain vocabulary and so on and so forth like those usually end up in the double like those usually end up in the double like those usually end up in the double digits from a word erate perspective digits from a word erate perspective digits from a word erate perspective right uh now you obviously you can right uh now you obviously you can right uh now you obviously you can fine-tune you know pick up an open fine-tune you know pick up an open fine-tune you know pick up an open source model and fine-tune uh but we see source model and fine-tune uh but we see source model and fine-tune uh but we see typically like what breaks here often typically like what breaks here often typically like what breaks here often and there are patterns s here in terms and there are patterns s here in terms and there are patterns s here in terms of what breaks. So, proper nouns,

of what breaks. So, proper nouns, of what breaks. So, proper nouns, jarens, uh phone numbers like random jarens, uh phone numbers like random jarens, uh phone numbers like random missing digits with phone numbers, uh missing digits with phone numbers, uh missing digits with phone numbers, uh wrong substitutions. I I'll walk through wrong substitutions. I I'll walk through wrong substitutions. I I'll walk through some examples of like how you solve for some examples of like how you solve for some examples of like how you solve for these addresses when you're trying to these addresses when you're trying to these addresses when you're trying to collect a long address. Uh you know, the collect a long address. Uh you know, the collect a long address. Uh you know, the the transcription engine could just end the transcription engine could just end the transcription engine could just end up missing some parts of it. up missing some parts of it. up missing some parts of it. Code switch languages. I I'll just take Code switch languages. I I'll just take Code switch languages. I I'll just take a example of a language I speak because a example of a language I speak because a example of a language I speak because that's was easy for me to put on the that's was easy for me to put on the that's was easy for me to put on the slide. uh where you know like if you slide. uh where you know like if you slide. uh where you know like if you were to sort of take English but written were to sort of take English but written were to sort of take English but written in a different script uh that's what's in a different script uh that's what's in a different script uh that's what's used for Hindi right like this is used for Hindi right like this is used for Hindi right like this is English written in that script right English written in that script right English written in that script right whereas like the actual English version whereas like the actual English version whereas like the actual English version of this is hello how are you so if if of this is hello how are you so if if of this is hello how are you so if if I'm addressing an audience in a I'm addressing an audience in a I'm addressing an audience in a different country where I have code different country where I have code different country where I have code switched languages and I start getting switched languages and I start getting switched languages and I start getting my English in a different uh sort of my English in a different uh sort of my English in a different uh sort of script everything starts breaking from script everything starts breaking from script everything starts breaking from the transcription engine to the LLM the transcription engine to the LLM the transcription engine to the LLM layer and then beyond because your LLM layer and then beyond because your LLM layer and then beyond because your LLM starts then producing output in that starts then producing output in that starts then producing output in that sort of script a lot of times and then sort of script a lot of times and then sort of script a lot of times and then your TTS messes up. Okay. So, so this is your TTS messes up. Okay. So, so this is your TTS messes up. Okay. So, so this is uh very important to be careful about uh very important to be careful about uh very important to be careful about and if you want to build your agent and if you want to build your agent and if you want to build your agent independent of the transcription engine, independent of the transcription engine, independent of the transcription engine, you need to build a layer that you need to build a layer that you need to build a layer that normalizes all of this, right? We'll normalizes all of this, right? We'll normalizes all of this, right? We'll talk about solutions in a minute. And talk about solutions in a minute. And talk about solutions in a minute. And there is the other case which is Hindi there is the other case which is Hindi there is the other case which is Hindi in Latin or or or you know Roman, right?

in Latin or or or you know Roman, right? in Latin or or or you know Roman, right? Which is like this is Hindi but it reads Which is like this is Hindi but it reads Which is like this is Hindi but it reads English which again messes up everything English which again messes up everything English which again messes up everything uh you know downstream. Those are just uh you know downstream. Those are just uh you know downstream. Those are just examples. This applies to, you know, examples. This applies to, you know, examples. This applies to, you know, Arabic, Mandarin, uh, Japanese, what Arabic, Mandarin, uh, Japanese, what Arabic, Mandarin, uh, Japanese, what have you. Uh, pretty much any language. have you. Uh, pretty much any language. have you. Uh, pretty much any language. So, what actually moves the needle with So, what actually moves the needle with So, what actually moves the needle with a at the transcription layer? Uh, for a at the transcription layer? Uh, for a at the transcription layer? Uh, for prop proper nouns, we recommend uh you prop proper nouns, we recommend uh you prop proper nouns, we recommend uh you using not just keyword boosting. I think using not just keyword boosting. I think using not just keyword boosting. I think a lot of transcription engine engines a lot of transcription engine engines a lot of transcription engine engines provide you keyword boosting where you provide you keyword boosting where you provide you keyword boosting where you can put in specific words into their can put in specific words into their can put in specific words into their engine, but doing dynamic keyword engine, but doing dynamic keyword engine, but doing dynamic keyword boosting. What that means is don't keep boosting. What that means is don't keep boosting. What that means is don't keep the keyword for the entire state of the the keyword for the entire state of the the keyword for the entire state of the call. just add that dynamically when you call. just add that dynamically when you call. just add that dynamically when you think you need that as an answer so that think you need that as an answer so that think you need that as an answer so that you get the highest accuracy. Meaning at you get the highest accuracy. Meaning at you get the highest accuracy. Meaning at different states of the call, the different states of the call, the different states of the call, the transcription engine will have different transcription engine will have different transcription engine will have different uh keywords boosted during different uh keywords boosted during different uh keywords boosted during different phases, right? Uh and that's what we've phases, right? Uh and that's what we've phases, right? Uh and that's what we've seen works best because if you just seen works best because if you just seen works best because if you just pollute your context of the pollute your context of the pollute your context of the transcription engine with tons of transcription engine with tons of transcription engine with tons of keywords, it'll start hallucinating keywords, it'll start hallucinating keywords, it'll start hallucinating again, right? So, so that's what we see again, right? So, so that's what we see again, right? So, so that's what we see typically working best. Uh typically working best. Uh typically working best. Uh yeah, post-process post-process your yeah, post-process post-process your yeah, post-process post-process your transcripts with an LLM, right? Cuz your transcripts with an LLM, right? Cuz your transcripts with an LLM, right? Cuz your LLM has domain context. Your LLM has domain context. Your LLM has domain context. Your transcription engine does not. So a lot transcription engine does not. So a lot transcription engine does not. So a lot of words that it would say uh I'll give of words that it would say uh I'll give of words that it would say uh I'll give you some examples may not make sense.

you some examples may not make sense. you some examples may not make sense. This is transcription like a phone This is transcription like a phone This is transcription like a phone number from a transcription engine. number from a transcription engine. number from a transcription engine. Right? Like what do you think that E is? Right? Like what do you think that E is? Right? Like what do you think that E is? Right? If you give it to an LM, it knows Right? If you give it to an LM, it knows Right? If you give it to an LM, it knows that's a three. Similarly, like what that's a three. Similarly, like what that's a three. Similarly, like what that one is, it's a digit one. So, so that one is, it's a digit one. So, so that one is, it's a digit one. So, so your transcription engine a lot of times your transcription engine a lot of times your transcription engine a lot of times could mess that up, but when you could mess that up, but when you could mess that up, but when you postprocess it with the LLM layer, it'll postprocess it with the LLM layer, it'll postprocess it with the LLM layer, it'll instantly correct that from a collection instantly correct that from a collection instantly correct that from a collection standpoint. I mean, uh, and and the last standpoint. I mean, uh, and and the last standpoint. I mean, uh, and and the last one, like I said, uh, transliteration is one, like I said, uh, transliteration is one, like I said, uh, transliteration is your ST output that's sort of u, you your ST output that's sort of u, you your ST output that's sort of u, you know, multilingual also gets normalized know, multilingual also gets normalized know, multilingual also gets normalized using either an NLM you first using either an NLM you first using either an NLM you first transliterated or, you know, use some transliterated or, you know, use some transliterated or, you know, use some kind of a neural uh, transliteration kind of a neural uh, transliteration kind of a neural uh, transliteration engine. There are a lot of them open engine. There are a lot of them open engine. There are a lot of them open source. You can just pick one of them, source. You can just pick one of them, source. You can just pick one of them, right? Uh that would do all of that work right? Uh that would do all of that work right? Uh that would do all of that work for you. Send cleaned transcripts for you. Send cleaned transcripts for you. Send cleaned transcripts consistently independent of the consistently independent of the consistently independent of the transcription engine to your LLM. All right. The third one we typically All right. The third one we typically see is collecting data. This is where I see is collecting data. This is where I see is collecting data. This is where I think 50 to 60% of AI agents mess up think 50 to 60% of AI agents mess up think 50 to 60% of AI agents mess up pretty badly. Uh and like we like to pretty badly. Uh and like we like to pretty badly. Uh and like we like to think of it as think of it as think of it as a UX problem. Uh but just for voice. So a UX problem. Uh but just for voice. So a UX problem. Uh but just for voice. So think data models uh and not a think data models uh and not a think data models uh and not a transcript coming into an LLM and and transcript coming into an LLM and and transcript coming into an LLM and and trying to figure out what the transcript trying to figure out what the transcript trying to figure out what the transcript said. So let's take some inspiration said. So let's take some inspiration said. So let's take some inspiration from uh I'm assuming most of us are from uh I'm assuming most of us are from uh I'm assuming most of us are developers here um you know take developers here um you know take developers here um you know take inspiration from Python's data classes inspiration from Python's data classes inspiration from Python's data classes pantic zod from Typescript or form pantic zod from Typescript or form pantic zod from Typescript or form fields in the UI right like if you start

fields in the UI right like if you start fields in the UI right like if you start thinking of it from that problem thinking of it from that problem thinking of it from that problem statement we have seen accuracy grow up statement we have seen accuracy grow up statement we have seen accuracy grow up from grow from 30% to like 95% from a from grow from 30% to like 95% from a from grow from 30% to like 95% from a data collection standpoint when you data collection standpoint when you data collection standpoint when you start thinking in that manner. So like start thinking in that manner. So like start thinking in that manner. So like decide your shape before you ask, right? decide your shape before you ask, right? decide your shape before you ask, right? Like instead of keeping it open-ended, Like instead of keeping it open-ended, Like instead of keeping it open-ended, can you keep it constrained? So can can can you keep it constrained? So can can can you keep it constrained? So can can a phone number be a phone number type a phone number be a phone number type a phone number be a phone number type field? The moment you do that, right, field? The moment you do that, right, field? The moment you do that, right, you know like how many digits it needs you know like how many digits it needs you know like how many digits it needs to have. You can do validation on on top to have. You can do validation on on top to have. You can do validation on on top of that, right? And then what sort of of that, right? And then what sort of of that, right? And then what sort of allowed values can even be there. So in allowed values can even be there. So in allowed values can even be there. So in the previous example we saw if an E the previous example we saw if an E the previous example we saw if an E comes in in middle of a phone number and comes in in middle of a phone number and comes in in middle of a phone number and you know it's a phone number you you know it's a phone number you you know it's a phone number you instantly know like either you smart instantly know like either you smart instantly know like either you smart guess that to three and confirm that guess that to three and confirm that guess that to three and confirm that with a user or you know that's an error with a user or you know that's an error with a user or you know that's an error and then you validated that and asked and then you validated that and asked and then you validated that and asked the user to repeat again right so so the user to repeat again right so so the user to repeat again right so so that's I think one of the common that's I think one of the common that's I think one of the common patterns we've seen here from from a a patterns we've seen here from from a a patterns we've seen here from from a a collection pattern name I think is the collection pattern name I think is the collection pattern name I think is the is the interesting one I've just picked is the interesting one I've just picked is the interesting one I've just picked a you know a a hard to pronounce name a you know a a hard to pronounce name a you know a a hard to pronounce name like There's no way a human is going to like There's no way a human is going to like There's no way a human is going to get this right and and no way a get this right and and no way a get this right and and no way a transcription engine will get this transcription engine will get this transcription engine will get this right. How many ever times you do this right. How many ever times you do this right. How many ever times you do this right? So the moment you start thinking right? So the moment you start thinking right? So the moment you start thinking of this as fields and then have rules of this as fields and then have rules of this as fields and then have rules and then confirmation mechanisms on on and then confirmation mechanisms on on and then confirmation mechanisms on on spelling this uh you know sort of uh spelling this uh you know sort of uh spelling this uh you know sort of uh letter by letter only then you kind of letter by letter only then you kind of letter by letter only then you kind of get it right otherwise it's going to get it right otherwise it's going to get it right otherwise it's going to mess up pretty badly in terms of how you mess up pretty badly in terms of how you mess up pretty badly in terms of how you collect this on a voice call and and collect this on a voice call and and collect this on a voice call and and that's just an example of you know what that's just an example of you know what that's just an example of you know what u I'm talking about in terms of the the

u I'm talking about in terms of the the u I'm talking about in terms of the the data collection piece of it. data collection piece of it. data collection piece of it. Another place where it goes badly Another place where it goes badly Another place where it goes badly dramatically is relative uh values. Date dramatically is relative uh values. Date dramatically is relative uh values. Date being one of the examples. If somebody being one of the examples. If somebody being one of the examples. If somebody says next week uh Wednesday 8, 8 could says next week uh Wednesday 8, 8 could says next week uh Wednesday 8, 8 could mean 8:00 a.m. 8:00 p.m. and then mean 8:00 a.m. 8:00 p.m. and then mean 8:00 a.m. 8:00 p.m. and then figuring out what that date actually is. figuring out what that date actually is. figuring out what that date actually is. Again, now becomes a very constrained Again, now becomes a very constrained Again, now becomes a very constrained problem. If you knew this was a datetime problem. If you knew this was a datetime problem. If you knew this was a datetime field and I I'm collecting a datetime field and I I'm collecting a datetime field and I I'm collecting a datetime field and then you take the current date field and then you take the current date field and then you take the current date and then figure out what this value and then figure out what this value and then figure out what this value would be bases that, right? So, so would be bases that, right? So, so would be bases that, right? So, so that's how you want to make sure like uh that's how you want to make sure like uh that's how you want to make sure like uh you do this with a combination of the you do this with a combination of the you do this with a combination of the LLM with the tool calling and the tool LLM with the tool calling and the tool LLM with the tool calling and the tool calling is doing a lot of this heavy calling is doing a lot of this heavy calling is doing a lot of this heavy lifting for you from a from a field lifting for you from a from a field lifting for you from a from a field standpoint. Yeah. And then you make you you run like Yeah. And then you make you you run like this from a unit test perspective. So this from a unit test perspective. So this from a unit test perspective. So all of your u evals need to start all of your u evals need to start all of your u evals need to start treating these fields as unit tests. And treating these fields as unit tests. And treating these fields as unit tests. And as long as your unit tests uh sort of as long as your unit tests uh sort of as long as your unit tests uh sort of validate and pass, you know, your agent validate and pass, you know, your agent validate and pass, you know, your agent is going to be uh sort of reliable and is going to be uh sort of reliable and is going to be uh sort of reliable and repeatable. You don't, you know, run uh repeatable. You don't, you know, run uh repeatable. You don't, you know, run uh hundreds of end to end agent test cases hundreds of end to end agent test cases hundreds of end to end agent test cases just to find out, you know, one field just to find out, you know, one field just to find out, you know, one field collection is broken. You do your eval collection is broken. You do your eval collection is broken. You do your eval at a field level and a unit test uh at a field level and a unit test uh at a field level and a unit test uh level.

And then yeah, like I said, I think u And then yeah, like I said, I think u you this this mindset makes everything you this this mindset makes everything you this this mindset makes everything more structured instead of hoping I'll more structured instead of hoping I'll more structured instead of hoping I'll put a ton of prompt, keep changing, you put a ton of prompt, keep changing, you put a ton of prompt, keep changing, you know, the prompt by a few uh characters know, the prompt by a few uh characters know, the prompt by a few uh characters every time and somehow my prompt every time and somehow my prompt every time and somehow my prompt engineering is going to make LLM much engineering is going to make LLM much engineering is going to make LLM much more instruction tuned and sort of more instruction tuned and sort of more instruction tuned and sort of magically start following some of these magically start following some of these magically start following some of these things. So in fact u like I said right things. So in fact u like I said right things. So in fact u like I said right like we have seen us get to 95 97% like we have seen us get to 95 97% like we have seen us get to 95 97% accuracy without having to fine-tune a accuracy without having to fine-tune a accuracy without having to fine-tune a model right and then and the trick is model right and then and the trick is model right and then and the trick is basically like just breaking down your basically like just breaking down your basically like just breaking down your context of what the agent is doing at context of what the agent is doing at context of what the agent is doing at that point with specific u states of that point with specific u states of that point with specific u states of what the agent is going through. what the agent is going through. what the agent is going through. All right u I'm just going to quickly u All right u I'm just going to quickly u All right u I'm just going to quickly u skip through this from a skip through this from a skip through this from a time standpoint. I just see I got three time standpoint. I just see I got three time standpoint. I just see I got three more minutes. Um hopefully that's a bug more minutes. Um hopefully that's a bug more minutes. Um hopefully that's a bug but but we'll leave it at that. Okay. Um but but we'll leave it at that. Okay. Um but but we'll leave it at that. Okay. Um so so this is the fourth area where we so so this is the fourth area where we so so this is the fourth area where we see issues coming in. Most folks take see issues coming in. Most folks take see issues coming in. Most folks take the LLM output and then we send it to a the LLM output and then we send it to a the LLM output and then we send it to a TTS. Obviously I think there are a lot TTS. Obviously I think there are a lot TTS. Obviously I think there are a lot of good TTS's in the market that take of good TTS's in the market that take of good TTS's in the market that take care of a lot of heavy lifting but a lot care of a lot of heavy lifting but a lot care of a lot of heavy lifting but a lot of times it it messes up. Uh what we of times it it messes up. Uh what we of times it it messes up. Uh what we recommend and what we've seen is you recommend and what we've seen is you recommend and what we've seen is you usually want to have a normalization usually want to have a normalization usually want to have a normalization layer between your LLM and what is fed layer between your LLM and what is fed layer between your LLM and what is fed to a TTS. You don't send your LLM output to a TTS. You don't send your LLM output to a TTS. You don't send your LLM output directly to a TTS, right? And and we'll directly to a TTS, right? And and we'll directly to a TTS, right? And and we'll just walk through some examples. The just walk through some examples. The just walk through some examples. The basics which is strip emojis uh markdown basics which is strip emojis uh markdown basics which is strip emojis uh markdown before before any synthesis into the

before before any synthesis into the before before any synthesis into the TTS. Most orchestration pipelines do TTS. Most orchestration pipelines do TTS. Most orchestration pipelines do this like you know a live kit or a this like you know a live kit or a this like you know a live kit or a pipecat would do that for you if you pipecat would do that for you if you pipecat would do that for you if you just set a few flags. So I but but just just set a few flags. So I but but just just set a few flags. So I but but just make sure if you're not using them or make sure if you're not using them or make sure if you're not using them or buildings from scratch that you've set buildings from scratch that you've set buildings from scratch that you've set this explicitly because you don't want this explicitly because you don't want this explicitly because you don't want an emoji showing up on on on something an emoji showing up on on on something an emoji showing up on on on something read out or you know markdown showing up read out or you know markdown showing up read out or you know markdown showing up there. there. there. Okay. I think I think some more common Okay. I think I think some more common Okay. I think I think some more common ones uh custom uh dictionaries most TTS ones uh custom uh dictionaries most TTS ones uh custom uh dictionaries most TTS engines provide this to you like how to engines provide this to you like how to engines provide this to you like how to pronounce custom words whether it's you pronounce custom words whether it's you pronounce custom words whether it's you know proper nouns brands uh acronyms and know proper nouns brands uh acronyms and know proper nouns brands uh acronyms and so on and so forth. So set those in uh so on and so forth. So set those in uh so on and so forth. So set those in uh when you go from your LLM to your TTS when you go from your LLM to your TTS when you go from your LLM to your TTS output because if you don't, you're output because if you don't, you're output because if you don't, you're going to mess that up. And I I'll I'll going to mess that up. And I I'll I'll going to mess that up. And I I'll I'll show you an example of like how we test show you an example of like how we test show you an example of like how we test that. Uh the the other one is like most that. Uh the the other one is like most that. Uh the the other one is like most engines also give you speed. So if you engines also give you speed. So if you engines also give you speed. So if you know you're pronouncing an entity, slow know you're pronouncing an entity, slow know you're pronouncing an entity, slow down. Have your agent slow down. So at down. Have your agent slow down. So at down. Have your agent slow down. So at point 8x or 7x so that it it's able to point 8x or 7x so that it it's able to point 8x or 7x so that it it's able to like inunciate on that specific entity like inunciate on that specific entity like inunciate on that specific entity and and doesn't mess up how it's and and doesn't mess up how it's and and doesn't mess up how it's pronouncing an email or a phone number pronouncing an email or a phone number pronouncing an email or a phone number or a name letter by letter or a name letter by letter or a name letter by letter and yeah just normalize all the messy and yeah just normalize all the messy and yeah just normalize all the messy stuff right like emails currency dates stuff right like emails currency dates stuff right like emails currency dates don't leave it to the TTS to do it uh don't leave it to the TTS to do it uh don't leave it to the TTS to do it uh most of them do it but don't leave it to most of them do it but don't leave it to most of them do it but don't leave it to the TTS to do it like build your the TTS to do it like build your the TTS to do it like build your normalization layer at your end so that normalization layer at your end so that normalization layer at your end so that tomorrow you think you need to switch tomorrow you think you need to switch tomorrow you think you need to switch TTS or you know for whatever reason the TTS or you know for whatever reason the TTS or you know for whatever reason the first one's down and you want to use first one's down and you want to use first one's down and you want to use another TTS you're able to sort of not another TTS you're able to sort of not another TTS you're able to sort of not rely natively on the TTS's engine but

rely natively on the TTS's engine but rely natively on the TTS's engine but you are building this in-house uh for you are building this in-house uh for you are building this in-house uh for for this to be managed for this to be managed for this to be managed and then yeah u I think I don't have my and then yeah u I think I don't have my and then yeah u I think I don't have my batch here but I I don't have my last batch here but I I don't have my last batch here but I I don't have my last name on that so my first test is if it name on that so my first test is if it name on that so my first test is if it cannot pronounce my last name or my cannot pronounce my last name or my cannot pronounce my last name or my company's name it's already dropping the company's name it's already dropping the company's name it's already dropping the ball so my last name is uh Balas ball so my last name is uh Balas ball so my last name is uh Balas Subramanion and if you cannot pronounce Subramanion and if you cannot pronounce Subramanion and if you cannot pronounce that using a voice AI agent uh like that using a voice AI agent uh like that using a voice AI agent uh like that's a check for me. I I know like uh that's a check for me. I I know like uh that's a check for me. I I know like uh you know the agent will mess up a lot of you know the agent will mess up a lot of you know the agent will mess up a lot of words that uh you know need to be words that uh you know need to be words that uh you know need to be spelled out day by day. The second one spelled out day by day. The second one spelled out day by day. The second one is our company name Po. So a lot of is our company name Po. So a lot of is our company name Po. So a lot of engines pronounce pronounce it pivo or engines pronounce pronounce it pivo or engines pronounce pronounce it pivo or uh pleo and and so on and so forth. But uh pleo and and so on and so forth. But uh pleo and and so on and so forth. But but I think specifically being able to but I think specifically being able to but I think specifically being able to control this in your pipeline is super control this in your pipeline is super control this in your pipeline is super critical. And then if you're building a critical. And then if you're building a critical. And then if you're building a if you're building a customerf facing if you're building a customerf facing if you're building a customerf facing product then then um you know sort of product then then um you know sort of product then then um you know sort of give this option to your customers. All give this option to your customers. All give this option to your customers. All right I'm just going to skim through the right I'm just going to skim through the right I'm just going to skim through the the the last two slides. U I'm I'm the the last two slides. U I'm I'm the the last two slides. U I'm I'm running badly over time. Uturn running badly over time. Uturn running badly over time. Uturn detection. I think this is it own detection. I think this is it own detection. I think this is it own separate topic but I'm just going to separate topic but I'm just going to separate topic but I'm just going to quickly pull up all the points so you quickly pull up all the points so you quickly pull up all the points so you guys can skim through that and if if you guys can skim through that and if if you guys can skim through that and if if you need a chat u after this we can we can need a chat u after this we can we can need a chat u after this we can we can talk about this. Right. Uh talk about this. Right. Uh talk about this. Right. Uh I'm just going to leave that for like I'm just going to leave that for like I'm just going to leave that for like five seconds and then and then we can five seconds and then and then we can five seconds and then and then we can chat about this offline. I'm quite over chat about this offline. I'm quite over chat about this offline. I'm quite over time. And then the the the last one is time. And then the the the last one is time. And then the the the last one is uh bargin and and back channeling. I uh bargin and and back channeling. I uh bargin and and back channeling. I think there's a lot of talk around think there's a lot of talk around think there's a lot of talk around speech to speech models that do some of speech to speech models that do some of speech to speech models that do some of this, but we've been able to see how we

this, but we've been able to see how we this, but we've been able to see how we could do all of this in speech to speech could do all of this in speech to speech could do all of this in speech to speech pipelines. You really don't need a pipelines. You really don't need a pipelines. You really don't need a speech to speech model to do all of this speech to speech model to do all of this speech to speech model to do all of this up. Uh again, I'll just I just put put up. Uh again, I'll just I just put put up. Uh again, I'll just I just put put this up on the slide and and sort of this up on the slide and and sort of this up on the slide and and sort of close at that. Um close at that. Um close at that. Um all right I don't think we have time for all right I don't think we have time for all right I don't think we have time for questions we can take them offline if questions we can take them offline if questions we can take them offline if you have any time but uh hopefully this you have any time but uh hopefully this you have any time but uh hopefully this was helpful and gave you some insights was helpful and gave you some insights was helpful and gave you some insights on uh what we are seeing in productions on uh what we are seeing in productions on uh what we are seeing in productions uh with billions of calls at scale. All uh with billions of calls at scale. All uh with billions of calls at scale. All right thanks