all right let's go ahead and get started uh as always please check out our community website for all the latest information on the virtual agent nlu uh feel free to go there to ask questions give answers all of you are here experts uh and you're probably awesome at giving answers already so please continue to do so if you're watching this video on YouTube and you like what you've seen go ahead and click that like button and that subscribe button and then you'll be able to view all our academies including this Academy once we upload it all our academies are recorded so you'll be able to see this Academy on YouTube at least within a week with that let's get on to our main topic so our main topic is how to improve the virtual agent fallback experience for better service so with me today to talk about that is Trisha Cornish a returning Champion she is a platform intelligence SC tradition if you don't mind introducing yourself very briefly and uh take it from here thank you very much Victor hi everyone my name is Trisha Cornish I'm a Solutions consultant I work with our platform intelligence team for the past five years I have focused on the virtual agent natural language understanding and other machine learning capabilities performance analytics and process optimization all the digital transformation technologies that make the platform sing so um Victor if I can grab the screen share thank you great okay um so here are our goals for today in improving the fallback experience I'm going to spend a little more time on the overview and then I will go through live software in the exercise and hopefully that will allow time uh for addressing any questions so when we talk about the fallback experience within the virtual agent conversation it really is helpful to understand the conversation structure if you're a long time user of the servicenow virtual agent you might have seen the progression of our static topics where we had to insert the greeting for every topic at you know in after the start pill and before the end pill and then we migrated to a conversation structure where each component within the conversation became the wrapper for our automation topics so our conversation structure now is topics for our custom greeting our topic Discovery and then we have our purpose topics these are published topics that are either available in the list or enabled through topic Discovery with nlu or keyword search when we do not have any content available this is where we run into the concept of fallback and fallback allows us to drive any content uh if a purpose topic was not found or prompt the user to ask for their query in a different way um so here when we go through let's say this was my delivered available topics for my users to participate we can think of how users would request password reset or checking the status of a ticket but what if they ask what's for dinner and users can ask a variety of queries we want to be able to provide a response now whether it's a legitimate query or not that's not really relevant here we want to be able to search content or present the user with a prompt to ask their question in a different way and of course then when their query is satisfied we want to go through survey and closing so what it's helpful to have that conversation structure in mind when you think about how fallback works so here in this first demo I've done some recordings so we can get through this process quickly what you'll be able to see is where I'm going to ask a question about the knowledge base but the knowledge base content is not available the reason why I did this is because I want to drive only the fallback topic so I asked what are phishing scams even though I can see the knowledge-based article in front of me I can see that the virtual agent which has not been enabled for this particular knowledge based article is saying I'm sorry but I don't understand your request well that's actually what's happening when we hit fallback now enabling the knowledge based content that's more a question of how we've activated our search uh what content we're making available to us but what we have here is an example where the virtual agent has not understanding the query and then it's it's presenting our fallback topic so now let's go ahead and go to successful results we're going to go through a similar scenario where now I've activated uh content with my AI search fallback topic and this is where I'm going to see successful results so I'm going to get my delivered out of the box greeting and I'm going to ask a question can I order from the vending machines that we've enabled for you know it purchases it's going to present a list of knowledge base or catalog articles this is the results from AI search I'm going to select the one that applies I see my knowledge base article I can click on the read more boom I'm done that is a successful search and when you say I'm all set it's going to complete this and say is there anything else I can assist you with that's a successful execution of the conversation with fallback the content surfacing answering the user's question and then moving on okay so then did you know so if we think about these two things together what we have here is what you saw in my just previous example where I asked about the it vending machines what we went through was a conversation structure that used all elements of my chat experience we have all of our setup topics working but the ones that are really relevant for this conversation are these two AI search fallback and fallback and these are delivered out of the box with uh custom experiences oops and I have um what we have here is the defaults so this is my default chat experience in my default AI search fallback that topic calls the fallback topic whatever topic I have right here now out of the box it's fallback topic period And so what happens when I am looking at AI search fallback I'm going to go through setup I'm going to go through AI search and then if there are no results or the user has said the results are not helpful it's going to go down this decision path where we run the fallback topic and this has created some substantial confusion but once you break it down and know that it's happening then it's a bit more straightforward to identify what we want to do with this now I'm going to stop for a moment because I see some active uh chats here and questions um everybody can hear it thanks for confirming Noah I do appreciate it great let's keep going okay so so now that we know that we have a this relationship between the topic AI search fallback and the topic fallback topic now we can see what might be causing confusion for some of our customers that are newer to this experience so here's an example of the default experience with AI search then fallback and so what I'm going to show you is where we're surfacing genius results first and then we're going to identify more assistance needed so then we're going to display the content which is how the AI search fallback topic works okay so here we've got what are phishing scams so now my knowledge base article is servicing I'm going to say show me more on what our phishing scams this I'm going to surface additional content that might be helpful and then I'm going to say I need more help now at this point because I know AI search fallback calls fall back when I need more help now you can see that I'm getting I'm sorry I don't understand your request because that's the default fallback topic period they um that is the default uh delivered topic that is in the fallback conversational experience and so you can see why it starts to make a little bit more sense but it could be very confusing for an end user so what we want to do is we want to look at this relationship and think about what to change so that we can make this better so what we have is the original fallback topic repeats and it says sorry and topic search but is that really helpful um I have a better way to do this I think and so what I want to do is put this back into understanding a more useful conversation structure so if we're looking at our topics as delivered as a specific elements and we know our default chat experience our placeholders for these custom setup topics what I present is that we can use fallback as a two-step approach we're going to use the AI search fallback exactly as it's intended we're going to say that is going to be my content search and I'm going to make sure that the Search application configuration that I'm calling from my AI search hyphen fallback that is going to have the search mapping and the configuration that I need I actually am not going to change AI search fallback at all but I am going to change the topic that exists in my fallback topic area my second one I'm gonna I'm going to present to my users if purpose topics don't help and that's what I consider these guys up here these automation topics that are specific for Integrations or troubleshooting and resolution if we don't find any purpose Topics by way of a natural language understanding and we don't find any content in our knowledge base or in my catalog as delivered through AI search what do I want to know next I want to allow the user to create a ticket based on their search query I want to route to the live agent topic perhaps I want to present the topic Discovery so that they can search using using an alternate method or you know if they asked what's for dinner and they're just kind of punking the virtual region give them the option to end the conversation and so now what we have is a redesigned fallback topic and in my redesigned fallback topic what I have is is a structure where I'm confirming my fallback I'm capturing my user search and then I'm presenting three choices where I am going to excuse me I am going to open an IT ticket and I'm going to call a topic block for creating an incident I might route to a live agent and I'm going to use a topic block or the delivered scripting to connect to an agent same with topic Discovery work just close the conversation now that's the first step I'm going to redesign my fallback topic my next step is and I can change this in the default chat experience but because I might have different use cases for my different portals I'm actually going to create a unique chat experience specific to one of my portals and I'm going to swap out my fallback topic for this setup topic demo fallback actions where I've identified that my options are ticket agent or discovery and so now what you're going to see happen is we're going to launch the virtual agent with my custom greeting I am going to type in my question and it's going to be something that absolutely there is no automation well the Denver Nuggets win the NBA finals and now I can see my my custom response it might not be my day I can't find any topics or content so now I want to create an IT ticket now in mine I've used a topic block where I can gather additional information I'm going to identify the best way to contact this user for this particular ticket I'm prompting them for additional information the reason why I'm bringing this up here in this demo is because what I'm using is just a topic block to call my it ticket I'm not tied to it and in the ticket that I'm creating I can actually do like I do and I in create it ticket which is append my short description so you can see that what I'm doing in this fallback topic is actually just enabling a better experience for the user even though we have no content available so that's the custom fallback topic when the user asks something outside of our content let's see it when the user asks for something where we do present results to them so here going again through my conversational structure my greeting my topic Discovery I'm going to ask again can I order from the it vending machines and so now I am going to surface a knowledge article it's not quite a genius result but close enough here in my knowledge articles for my vending machines now available here is my knowledge base where I can read more about it if I go back to my list and select another option because this might not have asked what I needed I get the I need more help so now I'm again using that decision path to go down the path of now calls fallback and so now I'm going to go through my fault my custom fallback topic again to create an IT ticket um Linda thanks for asking him uh this is in the Utah version but this has actually been available for a couple of releases um so here I've created my ticket even though I surfaced some uh some content uh David still has questions but I have my ticket created so our goal is really to um to think about the user experience when we have no content available or when we have some content available so here's where we're going to resolve in the exercise today I'm going to jump into the live software and bring up my custom demo topic that I'm going to make sure is a setup topic so that it's not available from the list I'm not going to see it in topic Discovery and I'm going to swap it out for my fallback I recommend naming it something that you can identify the functions of this fallback but that's completely up to you that's really the first step but the next step and this is key you want to think about which chat experience you're going to place this fallback Topic in whether it's your default chat experience or whether it's a custom chat experience up to you but that's what makes it active within your conversation structure so let's go ahead and go into to my um my virtual agent conversation so here can I find this here we go so I am in my Utah instance all right and here is my demo fallback actions topic I'm going to give myself a little bit more real estate here and I like to for testing purposes I like to put a bot response a text item so I know where I'm going it's this is especially helpful for AI search fallback and for fallback because this is helpful when I'm troubleshooting I could remove it for the users but you know it might be also conversational you know appealing to my users empathic side must not be my day couldn't find any topics or content so this is what the virtual agent is going to surface to the end user in this scripted user search text this is where we're going to grab the global search text this is in the documentation under the virtual agent scripts I've just grabbed my user search text and put that into a variable because if the user chooses to open a ticket or perhaps if I'm passing information to the live agent this is going to be helpful information my last chance okay how else can the virtual agent help this user where we weren't able to surface content for this particular chat experience I chose to see self-service topics create an IT ticket speak to a human or end the conversation that this does not have to be the solution for you perhaps you're not surfacing conversations to live agents no problem you just would take option two out I know when Victor and I worked on the knowledge 23 chat bot we didn't have anyone to help a user if the user asked a question to the chat bot so we directed users to go find a Goodwill Ambassador to answer their questions so this option right here is completely up to you but for my it tickets I know I'm going to enable virtual live agent I know I want to create an I.T ticket perhaps I want to surface topic Discovery or I want to give the user the option to end the conversation as for the rest of my fallback actions I've also included deflection points so that as I'm testing within this instance I can also capture deflection metrics this is also up to you uh deflections covered in the documentation but these are not required here is where I'm handing off to a custom topic block for creating my it incident where I'm connecting to an agent and this is really where I'm just handing off to my live agent topic or I'm executing topic Discovery so these scripts for connecting to an agent or topic Discovery these are standard scripts that are available in the default fallback topic that I just copied and moved over or you can find them in the documentation this topic block I created to accommodate a couple of different things I want a little bit more when I create an I.T ticket I want to actually append my short description to identify where in the virtual agent this query failed so that from a process perspective my teammates can identify where we need additional comments so that's why in my examples you saw fallback colon and the user search criteria and then of course my topic block which are the elements of the it ticket are surfaced so that's really the demo fallback topic if I go to my conversational interface conversation and I go to my home setting and on my conversational interfaces console I'm going to go to my settings and look for my custom greetings and setup and so here are my custom greetings and setup here are all my different uh variations I can change the default chat experience so that is always an option if you're only using one experience of course you can go into your default chat experience and swap out your fallback topic for um the original so here I'm going to make sure I'm in my Global scope I have all nine of my setup topics and the this is the catch all the default chat experience is the catch-all for our topics when it's not defined by exception in other chat experiences here is my fallback topic what I could do as my petrol is I could swap this out for my demo fallback actions okay so that's that's how you would do it in the default chat experience the other option in your custom greetings and setup which I think is very helpful to know that you can see that I have a lot of different use cases within my custom greetings instead of so my order is relevant But Here in My Demo portal which we just saw I might Define my fallback by exception so instead of thinking about my default chat experience I could say you know when it's my ESC portal only I want this fallback to apply and then the other conversational experiences in the default chat experience will apply but for this one you'll see my demo fallback actions and so you swap that out there you update and then when we go into our virtual agent and we refresh we will see that here so I'm going to wait for my greeting and can I purchase from the it vending machine and in my previous example this had been a bit confusing but now it's going to surface my relevant knowledge based content I'm going to say you know what I need more help and so in here I can see my confirmation so now I know as Trisha the tester that this is what I put into my demo fallback actions topic must not be my day what do I want to do I can create an IT ticket speak to a human see self-service topics or say I'm done and when I say I'm done now I'm going to go through my survey if my parameters are set and then my closing chat okay so that is how we can improve the fallback experiences Victor has been all over the Q a thank you Victor um are there any outstanding questions that you would like me to bring up or Victor would you like to jump in and how we doing I did see Eric you um you ask about form knowledge catalog items showing directly in the chat windows instead of redirecting to a link there are a couple different ways to do that that might be a great topic uh for follow-up but real quickly that would be URL navigation and pop-up locations I see Victor typing that um so are there any plans to uh are there any plans to wrap this fallback topic into our out of the box delivered fallback topic so we don't have to write our own uh great question you know I'm just highlighting one way to use a fallback I know actually multiple customers have created different um fallback options within their default chat experience some have not turned on AI search so they have used the search fallback topic here or um just the fallback topic like I said for knowledge Victor and I created a custom ending so I think this is just a way to highlight all the different ways you can use these two topics together I don't know of any plans to wrap my suggested solution uh into the delivered content they don't take a lot of my ideas I hope that answered your question um but that's just one way to create your own um questions other thoughts Victor would you like to jump in with any questions that we might want on the recording can I pull how uh can you show how you're pulling the user search to the short description one more time absolutely um and is that to the incident or the interaction okay if incident how you great okay so let's go back to here so what I did um so this is the question Cameron thank you for asking one of the things I want to ensure that I do is capture the user search text into a system variable here uh so for the recording it's going to be captured here when I hand off to my topic Block in my short description I'm act I added it so you could see how you could add additional content but I'm putting it into a variable just that gets search text and so that becomes the input into my short description I'm just instead of adding it as an option from the elbow wrench I'm just taking a quick script to do this and so that would be how I'm I'm handing it into the short description we could also do that same for creating the ticket and then that becomes the basis for the interaction so I hope that answered your question Cameron thank you for asking that um and I think we're at the top of our time so any other questions please feel free to post them on the community otherwise we will see you for the next virtual agent Academy thank you very much everyone