 I come from a technology background and as such, whenever we design products, whenever we design solutions, the UX people always complain that we bring in them quite later in the game. So I think having people like Smita with us, thinking of design at a much earlier stage has brought a cultural shift in our whole process and we start doing this right upfront and part of the reason she motivated me to come here and talk, saying that though you are from the technology background, you fit in here because we are talking about product and product design. So one of the things that we are doing lately is about chatbots and I think it's pretty common. It's the way it's getting commoditized. Most of you have already used something by now. But I think the aspect that we want to focus and that's where UX comes into play is how to make it more easy to use and how to not talk about it only from a technology space and I think Smita will cover some of that because what happens is when we talk about chatbots and when technical folks design it, they always design it for themselves. So if you take it to users, they probably hardly cannot use it. How many of you have used a chatbot? Wow. That's cool. And what's your experience using it? Does it get it every time? Mostly it's bad, right? It just doesn't get it. You give it after five tries, two tries. Either it doesn't understand your accent or what you're typing. It doesn't get it. It's pre-programmed for doing something that only a certain way it does. So that's one of the challenges and how UX can help there. That's where we want to go with it. So before we start the talk, I think some things that we wanted to cover, I think all of you understand probably chatbots, but it's very important. What are the components in there? So some of the people who are hearing it for the first time, I think things like natural language processing, NLP, intent, entities, utterances, these are the important things that constitute a chatbot. So when you start interacting with a conversational or doing a conversational interface, talk to a chatbot, things that are important is you specify what you want to do. Like for example, I want to book a conference room, I want to do a certain thing. That particular language, it's very easy when you're talking to each other. However, when it comes to a machine, there are many things being said as a part of that. And it has to infer all of that and respond. So typically the way it happens is there is a natural processing engine inbuilt into these chatbots, which interpret all of those things. Because a machine is no longer a human. So it has to do, it has to classify an intent, which is the conference room booking. However, to satisfy that request, it needs more information, like entities, when, where, and what time. So those things are very important. So it needs to ask, and whenever you say something, it's not apparent that you have specified the whole information. So a machine cannot give you a response unless it figures out the missing pieces. So one of the biggest intelligence that is there in chatbots or conversational interface is to find out what is the exact piece of missing information and ask you for that. So that rest everything gets interpreted, only what's missing is what is asked. And that's where, that's what the beauty of the app or the interface is in terms of looking for that specific information, asking you for what time, and scheduling the meeting. So that's where, and then the loop goes back and forth again in terms of conversing with the bot. So your experience with bot will go up, I mean, improve if it's just asking you for that one piece of missing information. The rest of stuff, it's doing it on the fly. So having taken you through this, this is the only tech part in this whole thing. So I will now leave it to Smita to cover you how we went about with the whole UX aspect of it, how we got UX into it when it was all the technologies in the room trying to solve this puzzle. So over to you, Smita. Thank you, Srikanth. Thank you for helping the terminologies aspect of it. So now I guess like people are know of what is intent okay, and what are the utterances and entities. So moving ahead to begin with, first I would like to show few of the bad examples of bot who triggered, like those are the one which triggered me to think about, can we have some UX intervention in between so that we can build it better? So first bad example, you can see there, user is trying to ask about menu or options, and chat is giving response the same response again and again, okay. The next example is something where user is trying to talk and ask about whether on weekend, but the kind of response is chat giving is irrelevant, okay. Because chat, like bot is not able to understand the term which is used, which is weekend, which might not be programmed at the backend. And the last example where bot have given some option, and the last thing in that option is, you can see few more, but human tend to ignore few things. So maybe user ignored that and user wrote like, can I see more? But the response which bot is giving is, I will get back to you. So bot is not able to understand that particular stuff. So these are the kind of examples which shows how chatbot are not doing well right now. So let us understand why chatbot fails, okay. So it is basically because of the conversation. So take an example of conversation between two humans. So for example, if half an hour back, if I would have asked Shikant, would you like to have coffee Shikant? So Shikant might have said, no, I'm good. Or yeah, we can go and have coffee. But when it comes the same conversation with chatbot, okay, chatbot will ask me when, at what time. So all those detailed chatbot will ask. But why Shikant is not asking me? Because we human do a conversation which is known as a contextual conversation. Because we know context. Shikant understood the context, it is now, okay. And when it is now, he will not ask me when and what time. But in case of bot, they are not yet there, okay. So they are still learning. But we know, like we have learned language from our childhood. So for us, it is very easy. So unconsciously, we give all the answers. But for bot, we need to add that intelligence. And how to add that intelligence is we can do few things. So these are the five steps UX, okay. Which will help us to take bot to for the next step. I will not say super intelligent, okay. But at least where it is right now, we can take it to the next step. So very first step is normal as a UX process, requirement gathering. The second step is, as you have seen the diagram where intent is an important part in case of chatbot. So intent finalization is the place where we can definitely, we as a UX can definitely help. The third part is flow design or conversational flow design. There definitely we can help. The next one is identifying or defining IA, okay, and interface elements. And the last part is we can also help chatbot to become smart after analyzing the responses from the user and how the actually flow goes, okay. So these are the five step UX. And we have done, like we have done bot in house and there where we played all these five steps, okay. So as I told you, like first step is understanding requirement, okay. So Srikant, would you like to give a little focus on this because Srikant is a person who triggered it. As far as the business goals is concerned, of course the primary goal is of course the financial side, why you build bots, why you do things is it has to be, it's relevant in the market, how quickly you can make a lot of money. So that's primarily thing, but you cannot state it like that. So the way I stated it is firstly, we see this whole trend towards moving towards intelligent applications. So that's what we need to build and provide to our customers ability to convert their current applications to more intelligent apps. And one of the ways that we will do that is by using this conversational bots, which are AI and machine learning enabled. So that was the spec. And if you see in all my spec, there was nothing about UX or anything. So it was purely a group of tech people that I did, our product architects, our engineers and those were the people that focused. But very soon we realized that we needed to bring the UX aspect in it because most of the bad examples that you saw when I showed my first demos after people working, I did for two months was pretty much that it wouldn't do a single thing, right? And all the language used and responses were pretty much all techy stuff. So there was nothing that it really did correctly. But then we brought in Smitha at the right time and then she will tell you how the whole thinking changed from that perspective. So after Srikant told me all this focus, I understood it because end of the day UX need to understand business requirement and business focus as well. So I asked Srikant what kind of bot you want to build. So Srikant said let us take a few like simple example like conference room booking app because our focus was in HR tech domain. So we took two examples or two things like one is conference room booking app and another is time off bot, okay? So definitely my target audience will be employee so that's the reason I collected few things which are employee focus, okay? When I have like discussion with, when I had discussion with Srikant about it, okay? But then I cross question Srikant. We have already app in place. We have already website in place, okay? And the application is doing fantastic when it comes to time off. Then why do you need chatbot? So what we did is we brainstorm on few of the use cases where Srikant convinced me why bot is important, okay? So it's not like that Srikant told me and I just jumped and did. So I did my fair job of UX by questioning him about why exactly he's looking for bot. So one of the use case which I would like to tell you was very relevant. So for example, new joining comes, okay? And he wants to know about what all leave policies or leaves are available for him. Then if you consider mobile app, considering the real estate, that kind of information he might not get on the mobile app. If you consider web application which has a really big real estate, but then a kind of minute level information will be somewhere down the line in the third or fourth level of our information architecture, right? So definitely user will not get that information straight. So I got convinced by Srikant that definitely chatbot will help because here new joining just have to type one single sentence or speak it out a single sentence and then the information will be fetched and reach to him. So I got convinced. So this was the first step which we did. The second step was a very important one which is finalizing intent, okay? So here I will not go by all the bullet points in interest of time. So I will first tell you about why exactly we need to finalize intent, okay? So this is how tech guys or technical people or developers think about intent. So take an example of conference room booking. So developer will think that, okay, book conference room, that's the single intent, okay? But when it comes to real world, I'm sorry, when it comes to real world, okay, if I talk to 10, 20 people, they will have very different intents, right? But if I consider business requirement and feasibility, I might not be able to attend all those intent. So definitely what we have to do as a UX, we need to pick few which are important from both of the aspect, employee aspect and business aspect. So this is why I think intent finalization is very important, okay? But how we did it? We did it really quick. We designed, like I designed a research plan. We did a quick in-depth interview with 12 people. Then we, I'm seeing the time. So I want to skip this one because only five minutes are left. Okay, so we did in-depth interview, really quick one. We got few intents, okay? From, like, we got a list of intents which are near about 10. We validated it with stakeholders, okay? And whatever the output was, six intent, what we did is with that six intent, I did a survey, okay? And in that survey, we were able to finalize four intent, okay? So this is how we are able to reach two intent. I skip one slide, which was about persona. We, there was output from the research where we were able to get few insights about the users where we are able to get them categorized. So that categorization is something at the end I will connect, okay? That's the reason I just skipped that one. Okay. Then when we know that there are four intents for the conference room, the next step is we need to define a conversational flow for each of those intents. So as I told you, developer and tech people was thinking like booking conference room is a single intent, okay? But when we did in-depth interview, okay? We came across that is not the only intent. People want to book conference room based on their specific need, like I want the conference room in head office at second floor with 15 people capacity. That was not the intent which we will be able to achieve with simple straight forward way of booking conference room, okay? So we are able to find four more intents, okay? Like a specific conference room booking, then weaving the detail of the conference room booking, like who booked it and all. And the last one is requesting someone to cancel their booking so that I can use that, okay? So these are the intents, and I have also showed utterances there. So what is utterances is, intent can have, so intent will have many different ways to say the same thing that those are the utterances. You can click this and you can see the utterances. I will not go in detail in interest of time, but utterances is like based on the human, how they say same thing in different ways, okay? Then this part comes, which is important one, where actually we have to write a script how bot will talk and how human will talk, okay? So this is a happy part, a straight forward one, okay, to begin with. This is something we design, like we have written it for all four intents and we gave it to developers to begin with, okay? So these are the happy part, and when we done with the happy part flow, we started working on information architecture, okay? Information architecture is something like this. So first level is straight forward intent and their happy part. But there will be few scenarios. So let me clarify one thing, bot understand whatever context you set it at one level. So for example, if you set the context of booking conference room, so whatever steps or whatever flow it has, it will follow that only. It will not understand where and when it has to move to the next intent, okay? So what we have designed is connecting parts. So what are those things where if user is in one intent flow, which is like book conference room, what are the things when user ask bot should connect it to next intent? And the next intent is specific conference room. And the next intent is what are the things which will take it to the view details and the last one is request for cancel. So this was the next layer of information architecture. And we also helped bot with getting few terms which are like stop, I'm not getting it. It's not clear to me, please cancel it. So those kind of things, if it is coming from user, then bot has to clear everything because it is based on the context. So it has to go back to the happy part so that it can start fresh. And there are limitations which we all know, like for example, the four intent. So what if user wants to do something with the fifth one, which is not designed yet? So for that thing, we definitely help user to get a human support. So there should be few words which will help bot to understand now I have to give the contact of HR or things like that, okay? So this is the fourth step and then visual design starts. In visual design, again in the interest of time, we need to define bot personality, okay? The tone, the conversation tone, it should match with the way bot will, what all things bot will help with, okay? So then the next theme, yeah. So the end of the fourth step is defining the interface elements, okay? I'm just skipping this particular slide. Here we just have to keep one thing in our mind that if we are designing it for customized mode or customized platform, we designer have free hand to do it. If it is based on some, like we are using Microsoft platform and if we are using some channel like Facebook or Skype or Slack, then we have, there is restriction. So designer needs to understand that restriction as well. So this was it. I don't think we have time to show demo. We were having two demos, one for time off and another, if you are interested, I can show you later on. The last step was, the last step was where you can actually, so understanding the responses, studying those, there are analyzing tools. Based on that, we can, we can improvise the flow which we have already defined. So based on the responses, we can improvise the conversation flows which we have already given to, like we have already given to the development. So that was the analyzing part and future. Future what we can do is, that's where I said the persona stuff. So based on the responses, we can do a chatbot can converse differently based on what type of user he or she is. So that's our future. We are trying to solve this. So thank you very much.