Transcript:
Let me start with introducing myself first. My name is Obaid Ahmed, and I'm the founder and CEO of Botmock. Botmock is a design prototype and collaboration tool set for four teams were building conversational agents all the way from your voice chat. Bots to your text based chatbots in between so, we cover a whole lot of that and today with me. I'm have honor of having gotten law from Center. He leads the global next-gen customer care practice at Accenture and I'll let him introduce. Reduced further. Hey, thanks so much for having me. so, quick introduction part of the Consulting work force within Accenture. Look at all the opportunities on the next gen customer care side. We have a huge, huge initiative on the EI powered care initiative within Accenture. So, very much part of that initiative and chatbots is definitely one of the key things which we are all working on. Perfect. so, today's session we want to talk about digital transformation and conversational AI how that is working how Enterprises are tackling that how can you know your team start thinking about approaching those projects from ground up all the way to deployment and beyond that. Alright, so, to start it off just have a quick question around what are the some of the challenges that you've seen Enterprise customers face when thinking about digital transformation sure. So, I think Colonel put it Very well when was walking through the whole journey that Walmart went through. I think the key thing which we have seen in the last I would say 1 year is Enterprises have matured from the POC thinking to actual deployment of these Bots and what that has made the conversation changed within the Enterprise has is it's no longer just an ideal it discussion it is now aging to your business value let discussion so,, you know, you have to go in and really talk about what's the value that you are talking about implementing these kind of projects in so, so, one of the one of the key things that we Al so, noticed was as soon as this conversation changes from an IT conversation to a business conversation the business owner now needs something more tangible. So, what we what we went through the you know in the earlier part of the sessions today with guided flows, someone even pointed out that isn't the flow that you made very similar to an ivr flow something like a glory be glorified. I we are flow. Why do these kinds of designs come into the market? Why do these things happen? It's because the business owner is really looking at something tangible that he or she wants to put out in the market try to try selling business owner that don't worry the Bots will learn on its own and respond on its own. It's not an easy sell. so, what you need to really, you know assure the business owner is that this is something which will work which will respond in a certain way and your brand will not Be harmed by Rolling this out. So, that's a journey which you have to walk through with the business owners. That's a big change. It's got in a lot of different conversations going. so, so, that's where a lot of our Focus now is and in terms of going on the journey, you know finding those use cases, you know it you start from you know, I've seen companies start for Mike to use cases or one use case or 10. You know, how should you approach that problem? think this is one of those questions which every business owner has we as practitioners, you know, all of us started with two use cases, which we did as part of POC. Everybody clapped what I call them as the hello world use cases, which as soon as our automated response came and everybody called Victory, but that's not the real world. so, one of the key exercises that needs to be done is really intend analysis. Where do you get the intense from you have a wealth of information already within the organization which you can leverage a lot of people actually look at CRM dispositions that this is one of the best ways to start your thinking, but if you if you just step back, All of these organizations have a very, very mature contact center, which is running contact center has voice calls recorded email transcripts that are available and they also, have chat transcripts available. so, as part of your journey, did you mind that data? Did you mind that data to really understand the reason on the intent for which people are calling in? That is the most important thing to do.
If you want to really have a good journey going forward you not only do that exercise to understand the intent, but al so, do that exercise to understand the utterances related to that intent. so, I heard a lot of conversation here with said that how would you go ahead and train your Bot most effectively and most of the answers came in that you Basically, look at the different Fallout reasons, which the bot really, really goes through and you know, you mind that data and you get insights, but would you not want look at if there's a Fallout which is happened with the bond. How did the agent actually re solve it? What did the agent asked and how did he or she interpret the question to give you the right answer, right? so, why is that not in our conversation? Why is that not in our process? Why is it that we not analyzing that data? It's a wealth of data that is already sitting in your organization. So, what we do with most our customers now is we walk this journey with them. We actually analyze this data and this insight that you get from the speech and the text Analytics. It is awesome to give you a very simple example. There was a lot of question on authentication.
How do you authenticate Wendy authenticate? You know, so, if you take an example of a typical Telco, they'll ask you for an account number or your landline number to authenticate what we realized is when you did speech analytics when people were given the option to ask alternate number, which is their mobile number. Everybody was happier. The sentiment score was much better on the speech calls. so, does that make sense to then have that as a prime way of authentication if the organization feels right the data is there to back you up? Why not? so, those are typical kind of things that you can mine and then really help business owners take those judge those judgment calls on how to go about doing it. so, let's step back on the plc's are things. so, how detailed you think a POC should be and how much effort and time should be spending on it. Now you have identified an intent.
Let's say when you got in the stakeholders to buy into it how deep the POC be in today's world really people have seen these Bots working. I mean like I would be surprised that if you go ahead and show someone that you know, this is the way the bot could work. They will say, oh, wow, you know, I've never seen this before they're educated. They've seen tons of videos vendors and startups or educating the market really well and large Enterprises. Our I didn't that maturity curve with it definitely understand this so, I would not spend too much time working on plc's I would actually go ahead and simulate something off the cuff, which I can just show within, you know, a few minutes a few days. I would not spend more than two days on building anything for a POC. And for that, you know, is it because you want to get the conference and team up in the stakeholders to go through that and you know, once you have that so, simulation going a what sort of tools that we use for that and be just to follow on that is at what point would you start building the team structure around it too so, that you know, so, so, think this is this is very critical. So, the first part of the question is what kind of tools do you use you? Showing you showing business owners. so, there are two philosophies one is show something which is bright and shiny and looks really cool. Nice. so, you develop something which is very custom or you co-create with them. So, you know, you use a tool where you sit with them in a workshop and you say okay. Let's see how this conversation will go and you co-create a conversation along with them now there are tools available. What? Mark is one of those tools which we found pretty useful in doing that where you know, we as consultants could be locked in the customer also, could be logged in and we could go create on the Fly and as soon as you co-create the conversation you can simulate it and see how it flows. So, I think that change the game a lot. We bring in a lot of design thinking Ways of conducting workshops to our customers and this basically facilitated a lot of that, you know co-creation aspect of the Bots Journey for our customers. so, I know you guys have been building other traditional systems for customers for a while. so, the difference between doing those kinds of projects and AI based conversationally, specifically oh, yes, this is where the thinking has to change within our customer base. so, most of our most of our implementations used to be large ERP CRM platforms where you know, you stuck to basically do a lot of structured thinking you do a lot of structured thinking linear process designs and then you implement a solution which pretty much gets you to 90% of what the end product should be. However, you know anybody and everybody in this room will agree that implementing a chatbot virtual assistant of any sort is definitely not in that same ball game. so, you will not get something which is 90% done. so, then the way you got to think of projects is how do you build in the optimization element into your solution. How do you fine-tune your bod going forward now? You need to fine-tune it and you need to build that system up front with good machine learning capabilities built into it? So, one of the things that we spoke about initially was doing a lot of speech analytics. so, what we realized is when you do A lot of speech mining you get a lot of utterances and if you provide your developers with those utterances to really look at free flow conversation, which everybody is scared of that. Hey, you know if I put a free-flow option here and not give a button for the customer to press will the conversation go Haywire. If you're doing a lot of that speech text analytics and Mining the data in you can actually start Machine learning exercises on a front foot and you can really put in a lot of utterances going into the design from day one. All right, and so, before you jump into the team's other things have another follow-up question that is you know, when you have just you know, figured out all the entrances you have intense, you have the data mine now, you know, you talked about design thinking and approaching it from a design perspective that optimization for the conversation at design level. I know that it has helped you guys in instead of getting ahead of it. Yes. so, I think what happens in most of the tools and we heard all the speakers till now speak about you know, AI development is not coding anymore. It is track and drop and it's simple so, which is true with will actually very simple, you know, people can actually build Bots within minutes. It's very important to now think of how is the work that going to flow from the conversation designers to the developers to the testers. And then how do you make it go life. Now the way we've actually worked is When we design these tools using platforms like what Mark what we do is we actually create a workflow so, that the business users are actually signing off on every content block as soon as they sign off on a particular flow within the whole conversation. We are able to then move it to the Developers.
Once the developers build it out on whatever is the chosen EI platform. It's the testers who have different Journeys already mapped in bot mock that they can actually go ahead and test out, so, it's seamless flow and I think it's very critical because otherwise you are moving work from business to it testing and it breaks. All right. So, I mean we can talk about optimization post roll out and how can that play a role in creating that experience and going The Last Mile right? mean you have the power out you ran your plc's you run your prototyping the design looks good. You have the approval. It's now live your work is done. The work starts here because this is where while you're designing your building out your data analytics model as well. A lot of folks earlier today in the group discussion. OST about kpis. What kpi is that you measure? so, you really look at what kind of Bounce rate is happening on those Bots. How much containment is each of the flows really getting you once the bot goes live? I think the work starts right there. You really need to start analyzing where things are going right where things are going wrong the initial 90 days post go live are super critical because chunk of your Happens in that 90 days you will fine tune your kpis you will fine tune. Your audience's you will add in you know, your deviation flows. Your nonlinear flows. You will add all of those things in the 90 days and then post that it's more of you know, what how do you how do you cluster various intense at the back end, you know one in ten flows into the other intent. so, you'll get a lot of understanding of those things. That's how ow, you will then go ahead perfect and just to touch base on this as well the invoice versus text Chat bar platforms. We have Facebook Messenger of your Apple business Channel. We have RCS SMS 11 different that we support. Yep. So, this is a wonderful thing. We have this we have this saying internally within our firm ivr is dead long live. DVR so, what does that really mean? You know, we all hate the traditional DTMF ivrs. We all tried the speech ivrs besides the fact that they don't get what you're trying to say. The fact is that they al so, really helped us do something very specific and that is Route calls. To the agents in the right way to a certain extent now the traditional world of picking up a phone and reaching out to the call center and maybe hearing the ivr may not see any significant change, but I do not see. Why can I not use my Alexa device? Why can I not use Siri to start a conversation? Ian and then seamlessly get connected to an agent so, voice will see maturity in different ways. I do not see why you know my echo dot which is kept in my house. I can't start a conversation with that get whatever I need but then flow into an agent conversation which can happen either through an echo dot or a call. Getting connected to my phone. Well now they have glasses do today. so, they announced that yes. this is a good question. This is a good lead win to the next question, which is how do you see the future of this in next 5 10 years? What kind of use case is what kind of platforms what's your wish list? so, I think we heard today about AI being used for sales service marketing. We had a friend talk about use cases in HR. I think with the way we're looking at it is a high-powered Enterprises. If you think the evolution of websites, it started with customer facing websites, and then we moved to you know, the same, I could of addicts center.com, you know WWX Center.com is a customer facing website. I can then al so, connect to my internal portal can do any and everything through Website, that's the same thing that's going to happen with the chatbots. It's not going to be restricted to sales service marketing. It's going to be moving to a different level with you know, you really looking at order to cash Cycles being murdered managed, you're going to be looking at your core it functions being managed, you know, we in Accenture have started getting notifications on skype that mine or timesheet is due for some mission the last thing that I wanted but you know that started, so, I'm saying that you know, there's no end to this thing which is which is happening here. Everything is up for grabs. It's not one-use case. It's not to use cases. It's any high-powered Enterprises that you're looking at in the next five years. Perfect, but that would leave some peculiar three questions.
One, thank you for the conversation. My name is Russian. I have one question. Maybe I can have one more question if I have so, I heard a lot you mentioning about not trying or POC and directly going and trying out is an actual implemented product. When I look at the way conversations at design right now. I kind of see two main patterns one is like a safe way where you have a very structured approach where you walk you see. And now predefined pattern using a list-based approach and all another way is a very open very is open mentioned like you let the user top whatever they want. But most of the time we start with the first one, but I feel like when you design a minim viable product or a safe product to start with and we use this very structured approach how effective it is that how effective such a method is to get the right data from the users because most of the time the data you get is what you said click you won't get user to Say actually what you are user want to say. so, yeah and get it. So, this is exactly when you start working with a business user. so, what will happen is the business user will say this is my process, right? They'll give you a process map. Right and you will mimic the process map into a conversation. I mean like we just saw that in one of the examples that were shown here earlier today. But conversations are never linear. so, when they give you the process map go back and check. How about 20 or 30 conversations around this area have happened with the call center agent on call until on chat and on email and that will give you a lot of insights into you know, what people are really asking give you know once you get To that data and you start bringing in that insight into the conversation and showing it to business that yes. This is the way if you consider all of these utterances and these possible conversation flows, that's when you really get containment ring. Otherwise, trust me the happy path can actually be Google searched and your website can actually give you the happy path. 'The yeah, we think about it from a design perspective. These are the golden pants. That's where it's easier to start with. so, if you were writing a script down your you would by default right the path that would always work and more you can start thinking about the edge cases quickly and show them those and say look what happens if this is not there or what happens if the user does not select one of the buttons doesn't do something that you're expecting them to do. so, you can you can definitely put an associated utterance for every content block that you're creating. You can have thousands of utterances that could be associated with one content block something definitely does that show. Go ahead. Yes. Hello. Yes. When you say that rarely have Business Leaders seen this technology and especially given well frankly Skeptics out there. Wouldn't it be akin to vaporware to show them a simulation as opposed to a working proof of concept with the understanding that there's a print work involved in creating a proof concept when it goes along our way to convince them of value add. Yes. So, yes is if the question is that you know, will the proof-of-concept help give them. The right context in which you want to build out the Bots. The answer is yes provided you can actually get that level of details into your proof of concept bot a lot of times and I won't mince words here a lot of times. It's a vendor who's trying to sell their platform or their service to the customer so, they know only so, much about That particular process so,, you know the use case that you will automate or simulate as part of the POC will get you know, the conversation started, but would you want to spend two weeks building up you see in today's time? I would I would think twice. So, my question is more on the omni-channel contacts management and the flow that is a possibility that the flow DAV designed for one particular Channel. Probably may not be relevant or may not be applicable for the other channel. so, when we are implementing a pure omni-channel solution whenever the cassava car allows the customer to Copper one channel to another Channel, what are the Problems. Actually, we should look up friend and architecture level actually need to cater to that to build a real cute omni-channel true good. That's and that's the perfect question. So, an orchestration engine is critical for any Omni Channel design why I what exactly is the orchestration engine? so, when we look at channels and let's put some of the channels so, you have Facebook Messenger, you'll have your web, you know the chart which is initiated on the web. You will have the mobile app biggest bass chats you will have chance initiated through WhatsApp for business. You will have you know, Apple business chat and RCS as we speak about now. Now, these are all multiple channels. What do you keep a different agent to handle all of them? The answer is no you will have one chat agent who will handle all of these conversations. so, it's very important to have from an architectural perspective of very strong orchestration engine in your ecosystem. This orchestration engine then filters out, you know, what kind of media can be Sent out to which channel so, it's not just what Facebook Messenger can do. It’s what Facebook Messenger integrated with an orchestration engine integrated with your agent desktop Agent app can do that's what you should be solving for. Thank you very much going to give her an Applause.