 cheating cheating cheating is only so what I want to show you hi guys I am Kumar Rangarajan from his company Slang Labs we are a startup here in Bangalore we work on building voice argument experiences for mobile applications we provide a platform that allows you to add a Alexa like experience inside your mobile app that's so what we are here to talk about what I am here is this notion of what is this notion of voice argument experiences and why you should do it and does it make sense hopefully at the end of it please get convinced that this is a worthy enough to get tried out so we will talk about the pro cons and what we do sorry my voice is pathetic I don't know if you guys can even hear what I am talking about I am just mumbling so let's start with first talking about like voice like voice is now why voice why voice is now becoming important and what are the advantages that work there is a lot of hype around voice Google is pushing it Amazon is pushing it a lot of hype so forget the hype part of it let's talk about the what what what value does voice fundamentally give to a brand or to an application developer these are some of them the first thing is like as an end user it's some of them for the end user perspective from the company perspective from the end user perspective it allows you to get your job done the fastest you think of something you can get it done the fastest today if you want to be able to get done in any mobile application you think of something then you convert them into a series of click click click click click click and somewhere after some n number of clicks you get a job done that's the current state of that when we have to get things done today so like thinking to getting it done is a long gap of it what are the voice think get them so that's like the big leap of jump today when you're all app developers here it how many of our parents can actually use their own apps anybody's parents can use apps they actually have done did you raise oh that's great one among 50 so this is a common scenario and our parents are and most are not necessarily not not educated or not really they're all educated most people are educated are studying or worked in good places but still they can't use our tech and take their own kids build but it was like people in the use Alexa if you've seen your parents use Alexa it's much more easier it's a much more newer tech they can start using voice much more immediately because it just opens up to a lot more people because the voice and the fact that sometimes languages become a barrier so that it becomes a huge but another thing is it very natural like this is actually a big trigger for us to even start slang when I watch my kids start using Alexa for your old order first time she sees Alexa I just she watches me how to use it it just becomes instinctive action second time on what she is no training in it she just understands how to use it like this is not a unique experience I'm sure all of us who experience watching kids use systems like Alexa and Google Home are amazed by it no training it's so natural to be able to speak it because you're all born with that ability to speak and another interesting thing this is from a developer also possible to be able to understand the true intent it's like typically if you want as a user as a developer you want to understand what the user is actually intending to do you know that he came to the screen and he did something but you are not sure about his intention he clicks this button he clicked that button that's all you have discrete piece of information and you ought to put together the discrete piece of information to collect understand what exactly was he thinking in his mind but the voice you now get to know the true intent whether you can handle it or not is a different story but the fact that you can now collect the true intent of the user much more straightforward is a great value this is now not necessarily from a end user but from a user perspective and cognitive overload you're not blocked by the fact that I have to go through navigation there are small screen scroll I don't know what how to do it's much more easier much more intuitive so these are all some of the advantages that voice fundamentally give you but today when we talk about voice we are talking about an assistant backed voice experience whether it is Alexa Google Home or Siri anything it's an assistant centric voice experience that's why we fundamentally talk about but that they are separate channels typically have to end up separating a building it for them this is a separate parallel channel but most of us already have a native app not necessarily native app or whether it is Flutter app or React Native app we already have apps and that's a fundamentally that's our biggest channel that which customers today use and a lot of advantages I'm sorry can anybody even read this this bad contrast it's okay there's going to be a lot of this contrast the rest of the slides great so so there are a lot of advantage there are a lot of value that apps fundamentally give you let's look at some of the fundamental apps value that still have today in today's world one is this notion of discoverability there's a play store I know how to discover them it's a well-established process the voice systems will also eventually hopefully get there today if you want to discover an Alexa skill it's not so obvious people will figure it out but it's here it's my discovered and what you can do and what you can't do in front of Alexa is not obvious because it's completely voice only so it is non-intuitive sometimes understand what it can do but in an app there are buttons kind of gives you a set of a context for what I can do and what I cannot do but some of them are the counter example is like it's some of the capabilities are so hard to discover in the app because they are so nested type but in voice everything is fine but typically the market is very mature you're not dependent on a parallel ecosystem to grow you don't need more and more Alexa devices and Google home devices to be grown you already have smartphones smart phone is a well-established market you already have apps they have billions of hundreds millions of users using it already so you can that's a existing advantage engineeringly mature and we're all engineers are mostly people here and then simple things like login if you want to do a login in Alexa it's a pain you don't know rediscover it if you want to be able to know a lot of things that is well solved in an Android and a mobile ecosystem has to be rediscovered we're not trying to solve in a parallel ecosystem like a voice only system and a lot of functionalities are very UI only like if you're trying to order a pizza this is my classic example trying to order a pizza by a voice only system say I want a pizza question is like which pizza do you want I don't know what pizza I want like I you have to give me a choice oh Viji pizza cheese pizza macaroni I've already forgotten the first pizza ever because it's humanly will forget humans can't take more than three to five things in the main at any point of time that's where a lot of things are still very helpful for visuals like when you have a list it's a lot more easy to process that list and last but not the least is a notion of privacy of privacy because when you're having when you're working with an app it's you and your brand the customer and the brand directly there is no middle man there's no middleman but when you have a third party app integration there is a middleman that you go through we have to log it logged in by the person the question is asked by the middleman answer is passed through the middleman so everything is tracked by a middleman and the middleman not really knows about you it also knows about the other parallel brands so this and both from a brand's perspective and also from the end user's perspective my knowledge is not just how allowed with one person but I'm not sharing it with a lot more people or with this one guy who essentially understands a lot more about me now that's a potential concern might be a concern for some might not but this is a big concern for a lot of these are all now this problem doesn't exist in an app because the relationship is very personal you and the brand so now these are all advantages that a mobile application gives you now can we get the best of both we saw the lot of advantages wise gives you now there are a lot of advantages that a mobile application gives you can we combine these two and get the best of both now that's what we call as the backs experience where you augment an existing visual experience by voice and get the best of both ways so that's the whole idea behind this what called backs a lot of words is theoretical words we all like to see in code or an action let's see a demo what does it look like so this is a live this is from one of the apps at the lab so long to be able to get to what you want activate the one day international roaming pack to USA okay one day international roaming pack for USA will be activated please confirm activate international roaming which country are you traveling to United States which pack would you like to buy one day international roaming pack for USA will be activated please confirm this is an example of an experience that you can build with what we call as voice augmenting an existing application which you can continue to navigate but now you can add voice to be able to now skip a lot of problems one is to be able to know the navigational complexity and also the language complexity now it allows you even my mom can now start using this kind of an app that she does not worry about how to do it just say what to and apple now react she misses something the app should be able to prompt prompt back and be able to collect all the missing these are the birds that's the thing about this notion of acts now what is so what what all can you do with this notion what does it now fundamentally enable the max cases like air one we saw it allows you to know it flattens the entire hierarchy different parts of your application can now be reached almost like in a single shot so single voice command a single voice interaction which now allows you to now go deep deep into any application and depending on what you say you can either drop them at step three or step two or step one depending on how complex the user utterances or if there are cases where there's a lot of data to be collected assuming you have a form we will not collect it now ensure the user feeling one by one field by field user can now more natively speak out what he's trying to say which conveys the entire form one classic example is a notion of an address when you want to be able to not take an address you don't want to say like street name this is not how we typically when we ask someone we don't break it down part by part and say we just speak the address in one shot and after the system to know we have done extract information or I think of another case like where like somebody who is now taking a repairman who is now trying to test the car is okay lift typically yes a form where you say okay lift tire is done right tire is done left front tire is done right front tire so each section has to be explicitly mark but in voice you can just say like all tires are done so now you can now and it is up to the app to be able to convert all tires into those are all four fields that it matters because that's a much more net natural way to be able to speak it cuts shots a lot of steps and a lot of times there are a lot of data that is actually exist in your application but people can't discover the end up calling customer care so banks one of them for example the top call that they get this to get the balance and then the balance is pretty easy information that most of us can retrieve from our banking app but for a lot of people it is not obvious we actually did I don't have a video of it but I can show it different like we actually did a small user research where we took ready ones around 35,000 rupees a month as a maid and she collects other things but because the fact she didn't understand how to operate the app natively navigate balance operation which is right there on the front of the app she didn't know she had to click it she has kept clicking all over the place for like 10-5 minutes to be able to not just get that basic balance operation the net result of the end up calling customer care and and waste a lot of money for that but we could just if they could have opened it and used it in the app and last but not the least is a notion of breaking barriers now typically when you want to be able to not target more languages than normal model is to be able to localize an application a localizing application is a typically a costly operation because you have to localize every single screen and make it happen and typically even for most of us if you look at ourselves we don't really want many times they are I at least me I don't really read in my native language I want to speak in my native language but I'm not so comfortable reading my native language but I'm okay to now pass info on most of the output is typically visual or in images I can understand the meaning but I can't tell the command to it I can't I can't express what I want to the application typically so that's where a bigger challenge starts coming in so with voice allowing you to now convert and handle the language part of it even if your application is fundamentally still English you can still handle other languages so without you having to know fundamentally change the application to multiplying multilingual you can now allow a multilingual user or a non-English user to start using an application so that's another thing that starts coming in and typically when you build facts there are two ways in which you can build out a max experience one way is called as a one-shot action or a voice guided navigation augmented navigation a one-shot action is where the user speaks something the sentence and based on how complex or how deep thing he says you go deeper into the application and then continue the rest of the journey by touch you don't want to always keep using everything voice the user spoke something you take them deep in and then you continue by or he speaks something then there's something missing then you keep asking you more information keep asking more information until all information that you need has been collected until the end of it and a kind of IVR style model of navigating the user based on what you speak so both things are possible these are all multiple ways in which you can actually add voice experience into your application now if you want to be able to make it happen so what are the fundamental things that you have to be cognizant of so what does it mean because like typically when you're building as an Android developer and iOS developer we are focused on UI elements and there are different style that we think of but moment you think of the notion of voice there's a different kind of a paradigm that you also have to start understanding there are certain terminologies and texts that starts coming in first thing let's start the basics are these four fundamental concepts all intense entities prompts if anybody has worked on dialogue flow another nlp system some of these terms might be familiar but typically when we use the system there they are like for a particular reason for chatbot building that's where you start using so now our proposal is to be able to know say can you know use it primarily in your main application as part of your navigation experience as part of your form filling as part of your direct user experience then you know start using voice now which means that even mainstream app developers also have to start understanding some kind of a these kind of terminologies as to be understood so it's better it's good so intense entities prompts I'll explain what each of these things so intense are the actions that the user wants to perform like this essentially braves down to a top-level action like that user wants to be able to perform examples he wants to book a ticket and then there is a he wants a list of light you want to see the list of flights that is available he wants to filter a particular option or he wants to cancel these are all examples and there is typically a noun that describes the action I'm sorry a verb that describes an action that he wants to be able to perform so this is how you describe you there isn't no start as an intent definition and they are all typically mapped to something that is executable into your application at the end of an Intel something should be able to do in your application if you can't act anything on it then it's not an intent it's just a part of the whole journey and the next thing is entities how to complete that intent what are the data I want if I want my banking application if you want to be able to see my balance because I don't need any data just the fact that I owe the user the express I want to be able to see my balance I can now go ahead and try to show him the balance but if you want to be able to know book a flight for example in this particular case to be able to complete that action to be able to get to something logical next step I need some basic information you have to know collected like for example I need to be able to collect the source city and to know the destination city I need to know the date without these three information I can't even go to the next screen sort of can't even take them to these are all basic information that I need to be able to correct some of them might be optional some of them might be mandatory depending on how much data that you want you'll be able to know the next one is the notion of utterances now utterances are the actual sentences or the commands that the user will speak which will differentiate between various inputs now the users is what are what do you have an intent now we want to be able to express that in a certain way what what ways in which it's going to be explained and so to be able to now understand what are the ways in some way could trigger that now in this particular example now those are all different ways for someone to say a initiate booking a ticket operation that's what they're saying they could see in different ways or these are like different ways in which you're using a saying hey is there a ticket available is there can I book so here it was more like a question sort of like an FAQ question or it could be something like where he's asking is there a ticket available I don't want to actually book a ticket I just want to know if a ticket is available you want to be able to differentiate between these two so then they typically become two different intakes so because depending on what the user speaks you want your application to do different things how if your application can't handle two things then you essentially map them all to the same action but if you can slice them into different and that's where you start prompts are things that you speak back to be able to collect information from the user like now in this particular case the user said Bangler Delhi 21st now but there are a lot of missing information that is still needed I don't know how many tickets to book I don't know if there is a one-way or a return journey or which when should I if it is a return journey what is the date for the return journey those information is missing so now these now these are all concepts that you have done audition now these things typically we do that in in our normal programming model as feels in a form so that's how we think of the best but in the world when you moment you start thinking of the world as a voice system these are all the terminologies that people start using this notion of prompts this notion of entities portion of the last one is a notion of confirmation now when the user says something it typically want to be able to speak back to the user also advice even ideally you also want to visual backing it up but also some kind of a audio prompt also audio confirmation also comes in he says he's gonna like to continue like he says spoke something you show something and then you want to be able to know speak it can be a positive statement where okay showing you flights from Bangladesh and on to the first that's a very positive state or it can be negative statements a there are no flights available to you on that particular day like technically it will be the same list it could be the same list and that list could be empty in one case and the list could be filled in on another case but you want to be able to now speak back because the user spoke you also want an audio confirmation so there was a thing that was done trial that is done with Alexa and Google Home which is a more trusted brand which do you feel more trusted in Alexa and Google so survey overwhelmingly said Alexa was more trustworthy even though Google was able to now respond to most queries much more efficiently than Alexa they were always able to get a correct answer but Alexa was a more trusted brand apart from the fact name is Alexa and the human element for it another reason that was given was the fact that when when you say a booker reminder for 4 p.m. tomorrow Alexa will confirm back saying okay booking a reminder for 4 p.m. now that gave you a confidence that I spoke something the system now did the right thing for me the response was okay done I don't know what you did so now because when you when you engineering you know code it seems like very efficient where does I say okay done right but you don't want to say okay done you want to see what you do so that kind of the tiny things make a huge difference in people's mindset and what you what you do and what you want to so now all those things starts coming to play when you start using now you want to be able to differentiate between different types of confirmations like how do you speak back those are the fundamentals like the intense entities are the fundamental building blocks that anybody has to be able to not so basics are this notion of context and a person is speaking you also want to be able to maintain this motion called context how for example in that continuing the previous conversation the user is now saying hey show me flights to jamae or to jamae now one way is like saying okay now tell me the form city which date you want to travel because everything is restart can think of it as a fresh command and fresh handling everything but a smarter system would say like hey smarter super smart a super smart system or a super confident system overconfident system will just say okay assume he wants to go from the same source city to a different destination city on the same date and then directly show him the list it to the next point but a more midpoint was to be able to now make the assumption but confirm confirm with the user saying hey okay are you saying that you want to get from Bangalore to Chennai on 21st is that what you are not trying to do because now we are trying to change the flow suddenly but I have context of your previous conversation or using that context I want to I don't want to just directly jump but I want to maintain but tell you this is what you intended how do you intend something else now this is something to experiment with now some people might not want this confirmation directly go why are you asking me this question some people might want this confirmation but since it's a new era it's always better to take the confirmation another thing is multi-modality now the disorientation alexa is like it's a purely voice for system or voice only and a voice for system but when you know do wax when you are augmenting in which should existing visual experience by voice you should take advantage of the fact that there are visuals packing you and people are not forced to only use voice or visuals you want to have a very seamless marriage between these two for example in this particular case you say which flight would you like to book and now we are showing the user said book a flight from Bangalore to Chennai for 21st now it took you to a screen showing your list of it and the question that is being asked is which flight would you like to book now you should not force the user to only speak back because if it is only so first you see back then the data that you're showing me on the list then what am I am supposed to read that list convert them to words and then speak back to you instead you should be able to know speak spoke to the list and touch and click on a particular list item of the list that's another way of now say it or you might just be able to know speaks up see something on the fly list and then speak it out both should be possible I actually made a mistake by saying it but yeah so you should as soon as you select a flight now move on to the next screen you're not I could have spoken or I could have used my screen we would not continue to be able to now do this this is the kind of thing that you want to do another interesting thing is the notion of error handling want to make sure this is super critical especially in voice because there are lots of possibility where it's like it's and it's the equivalent of like people keep clicking a button and nothing responding there we don't know what's happening but in the voice world you sometimes know what's happening so you can respond properly. So the first thing kind of thing is like when no speech detected user clicked and wanted to speak but you did not actually speak what do you do for it now you the way should I continue or should I just silently get away from it that's an error error handling state that you want to be able to do or you detected a speech but nothing was recognized speech is all gibberish I didn't really recognize anything that actually spoke out of the body what do you know and how do you handle that kind of or you recognize speech but I recognize it completely wrong give me curd rice was recognized as give me cold rice so now what do you what do you do in that kind of a scenario what am I supposed to handle that recognized but not handled like I don't know what to handle I got the system but I don't know what you said for example just said good morning I have no intent which handles good morning so now what am I supposed to do this is another example that we did for when we did a user research for some of the apps that we went to sort of some people the first instinct as soon as they clicked they don't click and start speaking they don't give a command they click and say good morning or hello that's their first instinct and they expect the system to speak back to them if it doesn't speak back to them because they spoke they are not going over here and then other thing is like they don't speak like this they speak like this even though it's a visual app because that's how we are we're used to some things like hello like so like then if there is no audio back you can't you're essentially lost that particular transaction you're frustrated with a particular user because this is all new user behaviors you're not trying to experiment with new user gears and it's and marrying with a traditional user behavior with phones the existing model is like phones so how do you now marry these what are the things to be and not handling it at application layer like everything is fine from us NLP and intent perspective everything got done well but then the application cannot handle it like book a flight from which I don't have a flight to Chennai now how do you handle that and the last is like you spoke an error there was some you didn't understand now do you keep asking the same question again and again you don't want to keep asking the same question again because then the user is exactly going to say the exact same thing then and you're going to get into this nothing is more going to move and I want to go to frustrated the user so every time you ask an error the next time we want to be able to say a different kind of error maybe more descriptive error and a more more descriptive like essentially it could be like a first time he says sorry I didn't get it can you try it again because the next time is like he's saying the same thing then you want to probably want to prompt him saying hey by the way you can say things like this then you didn't get it then okay maybe like oh give you one more time or next time just show him a list of possible choices that are available because then you train this user like okay this is the kind of thing that you can speak those are all possible and also handling novice and experienced users like sometimes you don't want descriptive responses back because it's very irritating for people already very used to it but if it is a new user you want to be very descriptive in how we are explaining back the user but if it is an experienced user because typically what happens like when you're now speaking back an error until the error is completed error message is completed I can't retrain that's a limitation that comes with voice because when you're speaking now if you start listening at the same time listening will listen to your own speech and that becomes very intent that's an unfortunate problem so you have to wait till the error message is completely spoken out and then then you can speak but if it's a long error message then you are losing out the user experience users will get frustrated but first time users probably want that so to be able to differentiate between a first time user and an experienced user so these are all some things and then the information of some some brain candy so just like there are UI patterns that exist there are a lot of these kind of wax patterns that you will start see once you start getting it as well you'll start getting a lot of common patterns that you can apply between multiple applications they are not every application is not very unique per se like things like list how do you show a list and how do you expect people to speak out a list and if how do you select from a list if people have used google show the home show you will see that there are some ways in which they allow you to now process a list go go next select the second item select the third item like so there are some ways in which you can now process list processing can be done filters like show me only or do not show me black like the inverse under four conditions to be able to know until now the main thing is handling the inverse on the recognizing the inverse on the true but whether it is black or medium or 200 rupees the way of applying is similar but the fact that these are the kind of filters that can apply is what you want to be able to understand and the way people will ask for help like the way people will say like they are not going to say help that's not the natural way to know say help like yeah i don't know what to do like that could be the way they will explain so now how do you know handling that kind of a scenario right what do i do now right now what do i do now is very different from i don't know what to do like you have to be able to handle those things and then conversation markets i so stringing intents like how do you sometimes people want to be able to connect multiple things together right like this is another thing that is here for example in uh when you're ordering food right the classic way is like hey add two uh pizzas to my list or like add three garlic bread to my list that's one way another way is like hey add two pizza three garlic bread for pepsi cans i might kill me if i say this but so like you can keep stringing multiple sentences together in a single form and this kind of a list processing mechanism is quite possible like this is kind of list mechanism will exist in multiple cases like in book my show like it's like i want to book this this ticket for two people and maybe like something sorry i'm just throwing out of the bed but there are things that you could potentially do here another thing that can start coming in is personality so that's something that voice can now start giving you because typically when you have an app so now the gay people now start like exactly like the word Alexa like when you use Alexa now Alexa brand as a personality that there's a notion that she's smart and there's a but that's the way they're trying to build because that's the centralized personality where brands are sitting behind that personality so you having individual personality becomes but now when you're now inserting voice experience into your app to your brand you can now start giving a personality hey uh smarty you're like whatever like you can keep a name like the way people want to be able to know refer to it it's not just a command that they'll give you might prefix it in some sentences and and you know want to give an error back it dude sorry man i didn't understand what you're talking about right like whether you want to be saying that versus like sorry i didn't understand what you're talking about right now there's a very different way in which you want to be able to now respond back and the style in which you want to be able to respond but that gives a brand a personality which are sometimes harder visually to be able to give a personality today's ways of visually giving personalities by colors and the layouts so with voice now the notion of personality starts coming out with like the way in which you the errors error message what how do you know handle an error message what are the sentences that you know uh hello what is the name that you choose for your assistant all of them starts becoming a part of your your uh your uh things that how how smart can you know handle it all and another thing with personality is like you can we were now we're working with a gaming uh company for example uh mobile cricket gaming example now when you want to talk to it one way is like just keep speaking to the app but what if we are as if we are talking to an umpire in the app like you are now you are immersing with your cricket game then you can say like hey umps uh declare this match like or like something that essentially you want to be able to now make it feel it's very native experience when you want to be able to now talk to the person rather than just a boring command particularly and another thing which you can potentially do is like version of personalization like where you now start having personal commands for person he says like a book with the usual right i don't want to now go through the like in a food you know book with the usual kind of figure it out uh very easily from here like it's a command which maybe is basically for me i want to be able to do uh or if you say like alu one yes maybe it's not a food item that typically exists like my classic uh volume i don't know if it matters you guys see i the only northing and dish which i like is paneer butter masala so every restaurant i go i keep asking paneer butter masala now that guy says this restaurant does send out paneer butter masala but there is just something else paneer i will give you doesn't matter it has to look yellow it has to look reddish uh color grass top paneer that's my paneer butter masala for me it's like you give it to me no for me that is paneer but i'm always going to call it as paneer butter masala doesn't matter what but now you don't want to centrally change the menu to handle the stupid people like me so like but now with with voice now you can map it to my personalization when i say this this is what this person always means to be able to now understand that person's more individual capabilities and now be able to now handle that record now those are all things that you can potentially do the last thing that you have to super worry when you know handle the notion of privacy so moment you had voiced do anything voiced today as there's a lot of coolness factor it has a lot of scariness factor like this people are scared and may start it at the same time there are a lot of stories that we keep hearing uh of alexa people hacking in and by buying hacking in the sense like it's not even like a criminal hacking it's a very uh it just happened and then people make stupid mistakes and people end up having uh to think and it is also possible to do uh criminal hacking and to be able to know this stuff so there are concerns that starts coming in when you now exhibit this notion of voice voice medium voice as a medium to interact with your brand starts creating these are extra problems potentially now you need to be cognizant of the fact you can't say this now things like what we as a platform for example tiny things that we do it's like we don't allow hot word deduction a hot word deduction is very cool like where you can just say hey okay google or say something and then do it but the problem is like it triggers it creates a lot of false positives and your brand could get you can start capturing in sentences with the user never meant to capture now every single company will have now google and google others can probably can manage it but like every single company we have to now start worrying about its privacy issues becomes a problem so where you want to make sure like the intent of the user to speak and only then you do it don't allow inadvertent uh intents then we now capture these voice things don't keep don't personalize their information too much like they're storing they're saying you something if you use third party uh system now the third party system will store analytics about what they spoke now that could give you personally identifiable information now as a personally identifiable information not being saved or being saved now how do you now uh make sure now those are all things that starts coming in so worrying about what data that you capture what and i'm like are you always listening to me uh what if i make so those things will all start coming in so we're now descending out to be cognizant of the fact that there is an issue and then eliminate that thing very clearly saying voice is only to help you to help you navigate faster do things faster and only when you tell me i will do it otherwise i'm not gonna you don't even have to listen to me i'm not listening to you at all make it very very clear and very very obvious so this is all there now uh base in which today you can do it is by especially nandor and ios systems are two ways one is what is called as an outside in integration so there is an assistant that exists either at a city or google assistant inside an app you can now in the i'm not in google is a city you can create what is called city shortcuts and deep links can be exposed inside your application through city people can talk to city and then go it but that typically is like a one-shot thing like you speak and you get into the application after that you can't do anything with voice typically or you can't keep having back and having a conversation it is typically one way and even same with what uh google assistant you can essentially trigger and launch your application but after that you are left with your touch base system or you can do a much more in-app integration side note marketing so this is our product so uh so this is like where you can now you can have a customized uh integration inside your application so where you can now uh have a voice experience right in and have that similar experience i have a back and forth do a bunch of stuff inside this and where you can focus and use systems like us to be able to do it later and the way you typically do it will arson any system you do will end up being something similar the first thing is to be able to design your user journeys just like how you design when you start you don't have to start writing code you figure out like okay where is the various screens that's gonna happen here are the various flows that's gonna happen and i start coding in similarly even for voice you have to figure out what are the user journeys that person is going to take and what are the intents that's going to come in what are the ways in which somebody could speak some samples and then what are the data i have to be able to now collect and if this data is spoken if this data is not spoken what should i do all that has to be and how do i map it to my visuals you have to be able to do some kind of a high-level uh design journey then you typically configure them in a system where you say you describe all your intents you describe your intents you describe your sentences which will trigger that intents the data that you need to be able to collect as part of the intent all of them you describe visually in some dashboard uh kind of system the third thing that you typically do is to test out this thing before you write code see like is then i'm is the intent deduction working correct not working correct what when i'm speaking something am i expecting so that before i write any code is my system even working correctly you want to be able to now uh test that the fourth step is the notion of code you then integrate a library into a system or you directly use uh google or whichever will use our slavery so use uh system to be able to now use just pop in uh your library and then the system will typically take care of everything that needs to be done you only focus on the actions because the system like us can only handle the question that the user can only handle the voice part we can automatically collect the voice do everything else that is handled with respect to the voice part but we don't know how to actually do anything with that command when the intent is satisfied what to do that's a business logic that the app has to take care of so we have the systems like we just focus on your app actions and leave everything else to a framework to be able to now get it uh get things done and that gives you some of the things that we spoke about to be able to now handle error escalation and everything automatically so without you having to worry about like how to handle all those things and the last but not the reason the notion of analytics because you can't get this done in one shot it's like i build something i release it oh magic is going to happen it's never going to happen your most first experience is going to be bad because you will not imagine how people will speak while it's not just about what they will speak what will be missing there will be all the things that will happen so you want to be able to now track analytics of what people are speaking and what is now what is your most popular intent and more importantly what is not working now what is not working is not like oh this new utterances which mapped on it are like completely expect patients now as soon as the user clicks the button on the if your message says hey what can i do for you today you're just opening up the entire world for it with respect to your app now you can just speak whatever and whatever that the imagines is possible with your app right in that particular case so you want to be able to now tailor the expectations correctly by the message that you are now giving to the user saying like hey would you like to order some food is one way to now curtail and say that that's like what would you like to do if you can only handle order food so these are all things that you can almost lastly so this is the things you can see i said i'll stop thank you great talk so we have run out of time but next we have break so those of you who like to proceed for lunch can do but we can just continue into the break for questions you're not trying to embarrass me great thank you guys