How is the space evolving, emerging trends, etc.
  • Chatbot Conference Team

How is the space evolving, emerging trends, etc.



Transcript:


Our panel today. We have a great panel. I mean from analytics to design to building bot. So you're going to get the full overview from each specific part of these people that are in the ecosystem. So again, my name is Joseph Holguin.

I'm the co-founder of smart Loop where a bot platform and live chat platform to help you automate your customer service and marketing and I'll go ahead and let each of the panel introduce themselves. Hello. I'm Stephen and from Mercury II. We also about building platform. We are aiming to do a basically very smart conversation part of the things we integrate our own NLP. So I think very similar in a way also very similarities to what Alex talked about in brother. But we are more focused to the business user not so much to the dolo infrastructure developer. Hi everyone.

I made care. I'm co-founder and CEO of personal things. It's a conversational analytics platform for conversational products and we are supporting more than 13 messaging platforms. Yeah, that's it. Hey everyone. My name is Henry. I'm an account executive at - pot and a dashpot. We are an analytics and optimization platform for any conversational interface.


So on the chat side that includes Facebook SMS web and mobile and then we also support voice experiences like on Amazon Alexa and Google home. So not only do we provide insights like identifying your user abandonment your unhandled use case as well as conversion tracking. We also help you with the optimization and improving the future performance of your chat bot. But by providing you the necessary and training data and so we have over 70 billion messages process up to date and I have over 18,000 Boston skills currently on the platform. Hi everybody. I'm Victoria from bot Society where a design tool for a chat and vice basically creative people can come up with the best interaction possible. And then you can hand it over when you're done with your design to your engineering team that can use any technology down the road rasa dialogue flow. I bought some they can they can pick whatever quick question.

How many of you have ever received the email that said we Toria from bot Society in the room because I'm always on them. I see some hands that's amazing. Thank you guys. Hi everyone. My name is sander. I'm co-founder of flow AI flowise another conversation II platform. We really focus on making it really easy to quickly iterate quickly get something live and then collaborate with your team on designing conversational AI Solutions putting them live on different channels training the AI and doing content management within the platform. We're also focusing on letting you broadcast messages.


So make it campaigning main campaigns and doing customer service. Okay, great. So as you can see we have just a wide range from building Bots design bars analytics So based on your experience. I think a lot of us have been in this industry for about three years. Now. Where do you see the ecosystem going and what use cases? Do you see chat Bots winning it? That's a start. Well, yeah, let's talk to the analytics guys first. Yeah, I can started especially if compared to last year.

I'm seeing a lot of Enterprises cases for the last three years Choppers are failed a lot. But what I see on Enterprises, especially with on-premise Solutions is getting hotter and we're mostly giving her an old text to software as a service. So little messenger Google home. I'm a masonic symbols and although Enterprise side. We are giving is on-prem and what we see especially on the bank's telcos and the insurance side. They're really good deals and with help of Russia and other tools, they're building really nice both experiences. And the thing is this failed last two years and they will get a lot of experiences on that. And right now they can put their budgets according to that and they can put their team efforts according to that and they will tear balls in that presence.




I think any so yeah in the past few years NLP has definitely improved we're starting to see more bodybuilders starting to really invest in more robust body experiences, utilizing NLP, utilizing intense and steering a little bit away from the form the rule base decision tree models now, Instead with the use of nou NLP a comes the importance of optimization in order for you to have a successful NLP engine. You need to be constantly feeding IT training data improving your intense and improving those in ten models over time. So we've seen a lot of increase in popularity of using an analytics layer to help supplement that and provide that training data and that being said we've also we've also seen a big explosion in customer service and support use cases across the Enterprise, especially on the chat side and more on The Voice side. We're starting to see more traction in brand engagement as well as e-commerce. Yeah, I think the experience match for us. So we're designed tool. So we see the projects that are going to get out and like six months versus like iridium production. So it's kind of like the other way compared to analytics.


We see a lot of use cases that are more relative to lead generation when I see when I say we see it's because I am It's not like we can tell from like the database or stuff like that. But that's the main thing and then in terms of new designs, there's a lot of interest in multimodal experiences around Vice and again on your website then all your mobile application. So not not a lot of new projects on existing platforms, like slack and messenger but more on your own experience, especially about customer support and Legion. I think one of the other things that is interesting is the further professionalization of the space like Alexander also mentioned. I think what you really need to have as a company is really a team that is dedicated to do the conversational AI development. If you compare it to the evolution of websites, like in the 90s, you would have a webmaster who would build a web page. I think with Bots. It's been the same guy give one guy who just now is bit of a hopping around making a boat and now you're really seeing these the successful use cases have teams copywriters developers designers.


You name it? We're really collaborating together with each other to build this bot and also see that my own background is in NLP. So I find NLP very of course for chapels. It's a must-have good and happy but I also see that you can mix it up with other conversational UI elements like buttons or multimedia. And I'm seeing a lot of more smart Solutions in that space that are combining multimedia in smart ways. I think what we see from our customers or customer requests to set things have changed in a way that in the beginning. It was just important to have a bot or if I have something on an ex arugula home and now basically every conversation we have is around Kpis, it's about kpis that are better than the way of doing it currently and then basically again can run the whole whole gamut of kpis that you do have in digital marketing and you can go from customer acquisition cost. You can go to click through rates to completion of forms to to all the way to to maybe if you're going to the customer service side 222 time time to close the ticket or cost to close a ticket, but I think that is a major major point that in the beginning. It was just important to have some kind of butt and be part of that hype and away and today every conversation is around which kpi you want to improve and what time and what are the expectations the ROI.

So I think that case the industry is changing that we are much more under the Typical business kpi decisions. No, definitely. So I think we're coming back to is we're seeing that we're really focusing or the industry or the ecosystem is focusing a lot more on sales marketing customer service. So do you guys have any examples some real life examples where some of these Bots are driving business results for some of your users or clients? I can I can give an example we're working with some banks and one of the bank's they're supporting to chat Bots one is in their website. And the other one is in their mobile phone.


So on the mobile phone side, you can send any kind of money to your friends just in a voice side and on the website, there must is a customer support. They're shifting like 5% of their law to own the customer support on the website to the chat bot is most of the FAQ chatbot, but you can ask the interest rate on those kind of questions to a chatbot and it answers and I think this is a perfect case what I see till today on the bank side and a to chat bot. Perfect.

Well in one is in customer support of what is in the interaction side sending money those kind of interactions here. Yeah, I can add onto that. So some of the verticals that we've seen utilizing the customers or bought are in banking Finance Insurance retail as well as software. So anywhere where there's a high level of communication with the End customer and what were the business typically starts is by automating those simpler use cases occur mentioned FAQ related use cases. I like to call them like level one use cases or typically uses that can be resolved within five to six turns of the conversation. So starting by automating those simpler responses and then being able to drive down that live agent cost reduce the reduce the amount of live agent sessions and then one more like one of the more unexpected benefits that we received was the actual increase in customer satisfaction score by utilizing the chat Bots and this could be due to a number of reasons. It could be perhaps the 24/7 support of the chat Bots or it could be because the agents now get more time to spend on the more complex and harder. Things or maybe it's just because people are more comfortable telling the bought their embarrassing questions.


Yeah, I think from our point of view. We have seen success in the small design teams tackling a very thoughtful use case. So basically the most success that we've seen is when before deciding. Okay, we're going to do this use case or these other use cases spent some time in figuring out is that the best thing to start from and even if it's not like the giant bot that will do everything it will start from a very compelling use case. So an example like a classic example, maybe taking the existing conversations in the company in figuring out. The most burning questions Alton time sensitive Manner and just by addressing that you usually drive way more weigh less requests or you decrease the workload on your side and at the same time NPS score as you were mentioning goes up, which is a great combination. And so we see this across the board on different verticals Healthcare is one of them Financial Services as somebody already mentioned but also Transportation. So those are three areas where the product and services are complex enough that you need to interact with your customers in a meaningful way and for so that's one of the things and then I think the other side that we see is a more longer-term projects where the development and design is going to take more time and it's more dip questions on like can we completely replace? A drug for example with conversation and that's going to take more time to see if it pays off by the proxy metrics that they get even now are very interesting.


So I guess we're seeing a very similar results across the board, but maybe you know, I think the customer service space. So one of our clients is Samsung in Europe there but they've been doing most of their customer service through Facebook Messenger now and I will fly out if they've tried to automate part of that and yeah, as you also mentioned are really focusing on these kpis, like they're not necessarily focused on getting the cost reduction, but they want to have the best experience for their users at any time that they want to communicate but I actually saw that they could reduce their customer service ftes by 25% but they were still getting 20% more messages that they could handle and also similar to what's been said here is that actually the customer satisfaction went up so they're very pleased with this result and they are actually Expanding because like they started with a few processes and now they're trying to automate more processes. So to give you an idea one of the process that they also are going to automate in this chat board is for instance repairs. So if you have a broken phone from the moment that you report the like that it's broken to the moment that you actually pick it up or get sent back by the by the repair service. It's all automated. So this chat bot. So that's what they are really focusing on automating more and more of these processes. Without rephrasing all the answers yet before I think I would just want to add one thing that just using a bot for a certain use cases gives you the opportunity to reach audiences that you would reach.




Otherwise, I think that is one thing we learned with two clients of ours and what are the use cases? I would present later on so I wouldn't want to take it too deep now, but in the other case, we have a thing that having conversations with your audience as a brand happens in a lot of channels, but for some channels the bot would be just the one where you would reach younger audiences best. So then whatever you do with your audience that is the place where you can do certain things in a way that engages them, right? So I think that is also one thing that we learned. Okay, so we talked about where they go system is now let's talk about where it's going. So I think Stefan mentioned maybe in his keynote three years ago. We kind of thought well that's going to take over apps. There's not going to be apps we're just going to have all these Bots that's maybe not the case now that we're seeing these business use cases come out. So, where do you guys? See pots going and how do you see this kind of changing inside of a business when it comes from getting it approved getting a developed and getting a deployed inside a business think that is this is probably the most Grace change except for energy over the past three years. I think we started out as a kind of a playing Ground for all the Innovation departments and the large companies and you could basically just do a lot and everyone was happy.


And now as a category, I think this is holds for the building platforms for the analytics and all the tools involved is that you now have to answer all these typical questions that I T buying Apartments was bother you with and then you would have to yeah. Well basically you are maturing your as a category where we're more up becoming the same like content Management Systems Erp systems are all these so yeah, I think it's a good sign in terms of proof that there is enough business impact to it. But at the same time we have to have to deal with stuff we didn't have to deal with three years ago. Yeah. I think we will see their cases in our lives more are rigid way like last year their world. 20,000 skills in Amazon like so right now six six two thousand and in a real case if you look at like in BMW 1 series or merge this a series you can talk the voice assistant and you can say hey hit my seed up and navigate me to their or call my mom and it's a real life example some riding in a two to three years. We will see this these kind of cases a lot. Yeah, I think the use cases that are really going to stick around in these next coming years are going to be the use cases that have a measurable Roi so in customer service, you can measure escalation containment understand what what is the level of conversations that you're able to automate and tie some type of savings or value to that Bots or that use case for sales for marketing.


You can look at the number of leads or number of sales. Have you created and get a type of measurable Roi off of that and that's typically with the measurable Roi that's when the team and that product can actually get the funding to push it into a business unit. And those are the use cases that I see really staying with us in the next couple of years. Yeah, what so I think what were actually seeing is companies switching from thinking about a chat pod as one channel to think about their company agent as an abstract idea that Dan can be delivered through watch out on the website and voice and like the Google system to relax or the mobile app etcetera. And so you basically end up with an virtual brand agent that has his own personality and his own I would say principles and then the organization declines that in different use cases and different platforms and we're really seeing them some of the most Innovative companies having this issue were like they want to tackle a lot of different use cases, but they need kind of like a centralized idea of how the brand is going to be a Persona on some level and that It's very exciting from our point of view given that the devices and the ways that you can you know, talk with folks are actually multiplying and the rate at which people use this new interfaces is going up. So like especially for voice only devices or like I like so the Google Assistant like there was an adobe survey where like 70% of the people who have one use them daily or more. And then at the same time we've seen customer expectations in that going up. So like if I've been speaking with a company yesterday via email and then I go on the website.


I expect to have the same consistent Persona and experience in the in the in the Eastern Chad. So we're going to see this this trend where like we go from thinking about a single chat Bots like a single activity to a more overall overarching kind of like brand Persona. That's what we see. Yeah. So to just to add a little oil to the previous comments. I think the truth the two main factors that are going to drive chat pod adoption area, I think like business process automation. I think that that's one of the things especially in customer service this what kind of processes can be automated using for instance chat Bots and the other one is just like conversion if you look at marketing, of course, if you if you compare chatbots to email campaigns or even text Messages as they were sent out earlier you see that conversion is way higher. So that's going to drive I think a lot of adoption in that area and I think one of the big unknowns and that's something I also don't really have an idea but maybe you guys have is the what voice is going to bring like what is going to happen in the void space? Because for it to be to be fair, I don't really see a lot of very interesting use cases in the void space yet, and I'm very curious what's going to happen there.


Okay, great. So, yeah, so now you've got a real good overview of where we were where we're at now and where we're going. So we're going to go ahead and open it up to some questions and definitely feel free. These guys have a lot of ranging experience from again designed to analytics to actually building it. So looking forward to your questions preguntas, but primero, let's clap not just a normal clap but a clap times one, two, three, four, five six and then like a half for Steph on bringing together for that. That was good. That was good. Here we go.


Sorry, Stefano me half for you. Any question, okay. She's got a laptop and everything. This is serious. We've heard some of the good stories. I'm curious if anyone could talk about unintended consequences that you guys have observed seen in your experience dealing with bought so voice interfaces. So if I understood the question is about unintended consequences bad Bots. Yep.


Okay at bots so I can share something. So I think on the for example and non-device side, but also on the Shad one of the things that we see is that the way that you design the conversation turns up personality is going to have a lot of consequences for the user experience. And so if you don't control the Persona properly two things may happen. So the first one is that you set the wrong expectations and I can make an example if your career is about that and then the second one is that you're conveying a brand especially Brand values which may be off compared to the rest of the experience. So for example, if you're building a personality that is to I would say friendly to apfel people will start to ask more personal questions and then may be fun, but in may not be what you want. And so when you start off, you're always like oh, you know, we need to be extremely friendly, but it may not exactly be what you want and most of the times you don't in the same way that say when you go and Buy a plane ticket the counter you don't want the person on the other side to go. Like hey, dude, I was going like that would be kind of weird even though it's friendly right. It's kind of like the same thing.


So the personages the first is has this aspect and the second one is expectations. So if I have a Persona that is makes me feel like it can answer more questions or about topics that I don't foresee. Then you may have problems. So for example, one of the experiments that we have run back in the days was to give like a strong accent to one of our Bots and that was very fun and people had a lot of fun with that but then they start any was about to that would tell you prices of goods like you could ask like how much is a smartphone like a secondhand smartphone and but the problem with the accent is that people were assuming he was local from a certain area so there Start to also to ask recommendation for restaurants and we're like What like why it is a like there's a prize pot and just based on that thing. So that that's our experience. I just wanted to add a note to that expectation piece. I totally agree. So nobody is ever perfect on launch when you launch the body will quickly realize that your customers are way more creative than you are in your developers and will dream of a million different ways to break your boss.


So that being said, yeah, it's really important at the early stages to manage your customer expectations or user expectations and some of the best practices that we've seen our to, you know to make it clear to the user what the batch can handle at the early stage of the conversation that can either come in the form of a menu or just really just identify that like this bot is meant for this certain use case and also making it clear to the user that they are talking to a bar and not a human. And I would have something and in our case, is there a lot of unintended things going on and there will be so in that sense. You need to iterate your conversation flow. So you release your experience in virtual assistant and you scale up and you need to get the data and analytics part and unity iterate. Your conversation flow is kind of cycle you have to do this and you will face a lot of unintended things and then you will reduce them. Here you go. Thanks for comments this morning. My name is journals in panic from Microsoft and I was curious.


I love the reference to kpis especially at the Enterprise level. Could you share some of your experiences integrating chatbots with process tools like Salesforce sap or even Dynamics in order to get some of these kpis in terms of their thinking that's a good question right there. It's a very good question because I think this is some of the things that always come comes along. If you do a Chapel process that you need to integrate because if you don't integrate your chat with any system, it usually doesn't have a lot of value. So I think this is a process that you need to manage very well and you need to make sure that within the organization. It's very clear that there is like a product owner and that you have all the people lined up correctly that can do this in my experience is a lot of communication between for instance a company like Salesforce or another company the business itself and then the provider of the of the solution and then it's figuring out how actually the this tool or this this system fits in a journey of the of the of how the customer uses the chat bot. And I'm making sure everything connects. Well, so yeah, it's an interesting experience trying to make this work.


But usually you can figure it out. But of course it's like it's also a very broad question because there's lots of different tools lots of different functionality and how is it being used within the company and how do you want to automate parts of that? So those are things that need to be well understood before you can even start to automate those processes. I would add maybe that we with our integration with Salesforce. We learned that we can make steps into the reaction what we are earlier from Alex in terms of personalization because a lot of data about the actual user the individual user lies in those systems so we can if we managed to sink them in a smart way. They can really in the beginning without the user telling us a lot of information about themselves. We can really personalize experience at least on a on a factual level that we know things about the user that the company as a whole nose, right? So I think the kind of stuff you would expect from a good call center or service center would be easier to do with Bots because I mean you can really integrate them on the data level. So I think that would be something I mean in terms of kpis that was probably what drive time to resolve a ticket because you have all these data available. But other than that, I mean just because it's just a data availability.


You could drive a lot of different other kpi's. I mean also in terms of marketing efficiency you could do a lot of things if you can infer data from a from a CRM solution into a conversation flow, then that would give you a lot of efficiency towards the marketing end as well. He sees there's one back there. Okay, I'm going all right, we have three one, two, three, whoever raises their hand the highest gets it. There you go. Hi. I'm sure on from City one of you mentioned earlier a chatbot that was deployed on the website and on the mobile app, and I was curious because it sounded like these are two distinct pots. I understand that the intense were different in each of them.


Yes. So are they distinct and were they trained separately or is this the same kind of parent pot? Yeah, its separate balls and their teams are separate. So why is that but they're connected in the bi site all their bi teams are same but the both experiences are different because the one project started like two years ago, and another one is started like a year ago. So their teams are different. It's going to this like most of the cases we don't see this kind of thing, but their teams are different. So actually most of the time the bank says but if you talk about the bank side all the chat bot owners are the same so they'll eat all the chat Bots. It can be like web part or it can be either in the mobile world. I think on that case is different because when it started like in size, but the other one is in the Enterprise on on-prem side.


So we set up an on-prem and then I don't know maybe later they all come in on Prem, but they are separate right now. So if I can add something about this, that's one of the ways that you can set it up. I think another way is also to so from a design point of view. So it's a different point of view. It's you design it once in an abstract way either for mobile or web and then depending on what you want to do inside the mobile app where you have maybe more freedom or you Different context of the interaction because it's inside and experience in the mobile app versus just somebody arriving on your website for the first time then you may want to adjust the experience. And so that's how you arrive to the point where you have two different experiences. So for example, like a logged in experience or a logged out experience, right? And so what we see is like, okay, I make a general design and then I create some variation for the platform that may translate into a different intent structure or it may not if you have the same intent structure, then you can train it once to answer your question. So you have one NLP Engine That Feeds both variations or you can have a different intent structure but you will discover that while defining the user experience.

It's kind of hard to do it the other way around it's like starting to say, okay. It's going to be the same or different before you actually figure out the overall map and blueprint of the interaction. Right? So like you design it and then you will figure it out. Okay, we have two different actually intend structure because in the logged in experience, we have three more intense or whatever or maybe you have the same you'd like answered. All right. So in the sake of time instead of just moving on and I feel bad for the two ladies, it didn't raise your hand high enough. We are going to do this question and the other question but the challenge to you is to answer the question in the form of a tweet. So you have 280 characters to do so, so this would be a little lightning round and I'd rather get lightning short answers that no dancers at all.


Here we go. First tweet question. Beyond goals and funnels have you guys seen any novel approaches to getting user feedback on an intent by intent basis? Open up the Twitter app in your brain. Can you repeat that last part of the question? Yeah on an intent by intent basis. So like when they're going through a specific conversation trying to address a specific task. We're trying to evaluate what they thought of that. I don't know if you can answer that in a tweet, but give it a shot anything can be answered in tweet. By the way.

I just want to get that out there. That would be a very short we have multi-step conversations and in our platform and come back to the table give you a demo. Hi, my name is Karina. I'm from Alba. We're in the energy industry helping utilities better communicate with their customers as well as improve the customer experience for the utility customer. So obviously we're trying to save them some money on that side as well as they raised a lot of them raised a lot, you know interest and a pain point in the field of participation in their pre-existing programs. So this is a question of the user actually taking action.


For instance.

It doesn't have to be you know, reduce your own service or challenging for the end-user just sort of setting up smart home devices or things like that. But also just taking out action in terms of peak load reduction to for example on pluck your pool pump from 2 to 7 a.m. Wait, so have you experience in that field? Not only using it to serve the customer better, but also make the customer to take action. Yeah, I mean, I guess just to kick it off. If you're looking to increase Drive higher engagement with the user. Right and I guess one quick comment that I made here is to use a platform. That's that does support some type of push notification. What like Facebook Messenger is a great example of that is available where you are or if your customers are using that platform, but with Facebook you at least have the ability to send push notifications if the user based on events, right? Like if the user it has like leaving it is they're leaving is leaving their lights on you tend to send those users like a notification that hey you should snap your lights off.


Maybe give them a fun fact about what it may what may impact the environment based on their actions and give them a call to action to actually perform that that specific action. So we have some experience with that. We had a an insurance company who run it ran a campaign also true Facebook messenger and it was actually a digital detox campaign as they call it to switch off your mobile device more often and to in order to get more sleep and I actually did this you could sign up for the for the challenge and then they measured actually if people would increase the or improve their behavior, and actually they found that there was a massive Improvement in behavior and that response rates were rehired an email. So I think you can do it in a very successful way. All right, ladies and gentlemen, seven and a half times the Applause to these people.

© 2017 by Chatbots Conferene TM.