all right everyone let's get started before we jump into our main topics today i just wanted to share that we do have our virtual agent and nlu community for content and experts check it out to explore best practices and up-to-date content and feel free to ask questions there and answer questions there to use the experience of others we've got a bitly link up there very easy to to memorize v-a-n-o-u-com for a community if you're watching us on youtube don't forget to don't forget to hit the like or subscribe button and we do this every other almost every other week and you'll get new content not only from us but also from the rest of the surface now community all right today our topic is mining your data with ai for compelling virtual agent topics uh with me today is jason young who is principal product manager from itsm and as always i'm victor chen i'm an outbound pm for conversational interfaces so the goal today we're gonna do a quick overview of our topic followed by an exercise uh led by jason and then at the end we'll do questions and answers if you got questions during the conference during the conversation feel free to type them out in the q a portion not chat q a portion zoom and uh some folks and i will do our best to answer them so the purpose of today's calls is to answer these questions and these these business questions you know can i discover what kinds of issues are driving the most ticket creations in my organization and once i figure that out can the virtual agent help me automatically reserve resolve them and the answer to those questions that we're going to uh go over today is yes and yes and in order to answer these questions we are going to share with you uh two features intent discovery and topic recommendations namely topic recommendations being being the heart of everything right so it is topic recommendations is available in the quebec release and it allows you to use your goldmine of data to help recommend potential virtual agent topics and this will help you these virtual agent topics are will help you deflect incidents and it will help you reduce time to resolution these are two big goals that every organization wants to be able to achieve right and with topic recommendations uh you run you use we use machine learning to practically cluster your incidents into uh different uh topics right whether it's vpn issues reset password maybe your organization has a lot of software uh issues or or printer issues right well we figured all that out and we actually mapped those to out-of-box virtual agent conversation topics uh that uh that the itsm organization has delivered out of box include you know pre-built right you don't have to build anything and it's basically a one-click publish if topic recognition says hey you know you have a lot of incidents uh about vpn connectivity or password reset that could be handled you can simply select those click add to v add to va or add to virtual agent and it'll automatically be added to virtually for you not only that for each of the topics it can also estimate the time of deflection and the time saved so it'll know already how much time these incidents have usually you know caused for you and then be able to uh calculate okay if you if you can automatically deflect these intelligently this is potentially how much time you'll save uh these are the you know channels the incidents came from that could pos potentially uh go down in volume etc so definitely uh if this if this sounds like a cool uh business solution that you want to use for your organization install the topic recommendations and intent discovery apps in the servicenow store today uh in terms of licensing uh not mentioned on the slide but a quick note of course you're gonna need uh the pro license here whether itsm or hr sd pro or csm pro not only does that give you virtual agent as we as has been discussed previously but also all these other cool solutions such as topic recommendations and intent discovery apps if you've got any questions about this again feel free to go into the q a and ask away otherwise i'm going to jump straight to the exercise and i'm going to hand it off to jason jason feel free to take it away great thanks victor good morning good afternoon good evening wherever you guys might be joining really appreciate everyone taking the time today and uh joining us for this session so let me just go ahead and share my screen um many of you as victor mentioned may have heard of topic recommendation intent discovery um you know one of the big motivations for us in building this was oftentimes you know you as a as as a platform leader of servicenow within your organization um you might often know maybe what are the key you know two or three topics or or content that you might need to work with your employees but oftentimes once you get past that initial you know simple cases like passive reset vpn issues then it's it's kind of a kind of hard to understand what other topics might be worth building um to support it for virtual agent so today i'm going to kind of cover up two two main areas one is just you know how do you get started as victor mentioned these applications are in the store today um to get and i'll kind of talk through what are their relationships and then i'll um follow up in terms of just you know how do you interact with these tools and how do you run an analysis on your in on your own instances so you can take advantage of not only out of the box content we built but also identify what other content gaps that you might have within your organization so one of the best places to start if you have not seen this today is we do have we did introduce a conversational interfaces guided setup and as part of this guided setup you can walk through the plugins that need to be activated within your organization so here you know if i go down to the plugins and store apps i can see there's there's a few different tasks that i still need completed um in order to take full advantage of the servicenow virtual agent platform and so if i come in here to look at what plugins i'm looking at i can see that topic recommendation is something that's identified that i can configure and activate there's also an itsm virtual agent conversations i won't go into the exact details of of installing it on this instance but i wanted to just highlight this is a good place for you to see you know all the plugins that you might need to to enable to take full advantage of the virtual agent solution so if you go to the topic recommendations section and click configure it will take you to the store um and as part of the store application um you'll see that let me just pull up here there the intent or sorry the topic recommendation and for this topic recommendation application you will see a get if you haven't um activated this or grab permissions from it and what this will do is um it'll white list this application and plug-in so that you can then install it on your instance the the other application that i want to mention that victor alluded to earlier is intent discovery so when you install the virtual agent topic recommendations one of the applications that will be installed as part of it is intent is the intent discovery application and this one you will see here and these should come together as intent discovery is driving the um the actual analysis on your instance and so these will be these will be captured and topic recommendation will leverage intent discovery to drive the the topic mappings so i won't go over the actual full installation but guided setup is a great place to go and grab that and then the itsm virtual agent conversations this will also be in the store and this is content that the itsm team is developing on a monthly basis based on content that we see is important in organizations and based on customer feedback from folks like yourselves on the call so this iets and virtual agent conversations can be found in the store as well it is bundled with a few different applications along with it includes our nou model as well as some conversations around password management and virtual machine management and this will continue to update with content that we have available and that's something that will continue to release monthly now just to kind of jump into the actual configuration of this to show you once you've installed that on an instance through the guided setup what i i will show you is if you go under the topic recommendations section i've gotten something already configured here like victor showed and you'll see that these are this is an analysis that i had previously ran and this analysis will show me topics that are mapped to the itsm content that's been developed and potentially identify any content that that doesn't have a topic that is coming out of the box so this could be topics that you build content on top of or you could leverage existing content so just to show you this is the main window that you'll see for topic recommendation when you run a first initial analysis you will see there's various settings that can be toggled so when you come here right now i'm running an analysis on the instant data and this is triggering it off of the short description and i have a filter based off of what date range i'm looking at so in this case if i go in and look at the filters this can be changed and modified to to meet your organizations and this is kind of what kicks off that analysis so in this case you're going to see i'm looking at incident as the data source today incidents and requests are supported by intent discovery there will be additional capabilities added on other data tables but as our initial launch this is these are the current data sources that are supported you can then select what field you want to do the analysis on for topic recommendation and then here you'll see there's a taxonomy entry and this is behind the scenes taxonomy that servicenow has created and curated and uses it to map to common issues that you see within your organization now lastly you can run filters so you can say you're looking for just specific date range of incidents could be last two years last seven days last 30 days you could do it based on the smaller subset and then you can also add different filter conditions say um based on potentially you know who is the open by or what uh where did this source in terms of where these incidents came from maybe either you only want to pull email you only want to analyze email sources or self-service content and these then can run an analysis on this and then you'll lastly see a frequency so this frequency will allow you to toggle how often this is ran so you can set this up to run every month and every month you can come into topic recommendation and you'll see whether there's new content that's been developed there or sorry um that has been identified as part of the clusters i'm going to configure this and update this i'll pause here for a second just to see if there's any questions on kind of the initial setup i know just in the interest of time i i didn't want to spend the time activating all the plugins but victor are there any questions that people have before i move on uh we had one question that was addressed basically does talking recommendations work with topics that are scoped to specific apps like hr and uh as jason mentioned earlier uh it is currently just the itsm taxonomy and itsm related topics uh hr will be available later this year with a store release so you know stay tuned for that but currently not now uh we have some other questions what is the minimum incident requirements or are there perhaps if i'm trying to if i may parse this question is there a minimum amount of records that can be uh analyzed for topic recommendations yeah generally our our minimum and recommendation for that is to have ten thousand ten thousand records for incident analysis um that's generally where we see enough data that's meaningful and and can be clustered accurately um potentially there you can work with our teams to maybe reduce that but that's that's generally our default guidelines that are set up on the system good question thanks uh yeah that's all we've got so far okay sounds good and uh just to kind of show you you can see that there's a section under the virtual agent you'll notice that there's topic recommendations as the main app which you just saw and then there's the recommendation settings and these recommendation settings are kind of where you see what i just kind of toggled and configured so in this case i have an analysis on my incident records and then i have an analysis on my request records and so if i go back into the main topic recommendation app you will see that um i can do an analysis right now on this content and then or sorry on my incidents or i can run it on my request it gives you some basic information so in this case you'll see that vpn connectivity is something that's been identified that's mapped to an out out-of-the-box topic and if you drill down a little bit more you can see that there is you know the sources that they're contacted from and also if there's any specific assignment groups that handle the ticket and if you're a multi-multi-lingual organization you'll see the different language breakdowns potentially if that's seen here as well what you'll see here is i've already ran topic recommendation on my instance so you can see that i've added these topics prior in a prior session one of the things that the adding mechanism does is that it will duplicate the out of the box itsm in this case content and create a second version of that of that actual topic and it'll allow you to modify and create it for these newer ones you'll see that these are identified ones that i still have not added so if i come in here i can select this to be added to the virtual agent i could also look at additional ones that don't have a topic flow today that i could build topics off of and so if i come down here and take a look i see there's a bunch of different issues that have been identified if i want to go in and say for example add this itsm issues right here what you're going to see here is that this reset password selection is gonna as i mentioned duplicate an out of the box topic because this is something that you know we have content for servicenow has content for for the ones that are down here there is no content that has been built on these so this is up to you your organization to build out that topic flow but because this has been identified as an intent that is clustered we will build the um the nlu intent that ties to this and so your organization would go through that process of building out what that conversation flow looks like and so for now i'll go ahead and click add this to virtual agent and you'll see in this case i have a few different options in terms of language if i have multi multiple languages activated on my instance i'll see different models that i can tag to so in this case i've got three different models that i'm using on this instance and i'm going to create this new intent in this va itsm model so i'm going to go ahead and click add to va and this generally will take a couple minutes and so this will when i go back into the designer i'll be able to see that the this reset password topic has been duplicated and then i'll also see something around where there is going to be an additional issue software browser intent um that exists now in the um in my nlu model so just while this is loading and uh putting in the record any any additional questions victor um yeah we got one question from uh anonymous uh where i'm just trying to parse this question again oh where's the recommendation algorithm run is in our product instance or separate machine learning instance yeah so this is uh these are running off of um so the data is going to be housed on your side but we have machine learning servers in the servicenow instance that that the models are ran off of that use some of that engine but the models themselves are going to be housed on your instance but it uses a little bit of combination cool and another question what if what happens if an incident falls in more than one topic like how does it dif how does it get uh organized and proper recommendations yeah so usually the waiting is going to be it's only going to map to an incident is going to only map to one one actual intent and it generally will just look at that based on a confidence threshold um so if there are two you know it's going to take the one that it has a higher match too okay so um these topics have been added um i can then go over to um my nlu just to my nlu model and uh and i can see that here in this va itsm english model that i have created these intents are now added as part of this so here i now have this itsm issue software browser and if i kind of click into this you'll see that it does take the utterances from the um from the short descriptions and it's a good starting point you know obviously you probably will want to look at this and tune it just to see what the performance is like but this gives you a good starting point for it and then from here you can then go to the designer and leverage this intent to build a topic um so if i come in here earlier i mentioned the reset password you will see that reset password this is the out of the box content that is shipped as a template if it is added and duplicated write the recommendation engine will create a reset password recommendation topic and from here you can go in and edit this topic as needed this is an example of where the intent is identified and it matches and out of the box content so most of this should be [Music] pretty good in terms of what needs to be configured there's probably some of the flows that you will need to modify for the other example creating new content in this case so if we're just to do just kind of a quick test i won't create this topic all the way through but so maybe this is software issues browser you will see that i can select the model that was just created and as part of this model you will see the different intents um that were identified and can be used so this is where you know if it's an intent that's been identified but the content doesn't exist you can go through and uh and build this content here so um you know topic recommendation is meant to help your organization find content that can be developed easily by you and and kind of surface what are some of the spots that you may not have noticed in helping you develop that content um i'll uh one thing that might be just worth understanding the relationship really quick is uh i didn't get a chance to jump over to it but the intent discovery uh is a separate plug-in and it is the engine essentially that drives the clustering analysis on the incidence table and topic recommendation then we'll take that data and map it to content that exists and and help leverage it within a virtual agent context intent discovery is today more focused on virtual agent but could be used more broadly across your organization to identify where users are kind of doing certain things so if i go in here to the intent discovery the running the analysis is very similar to how you would do it in topic recommendation they're linked together essentially so when you trigger that analysis and topic recommendation you're actually kicking it off in intent discovery so if i go in here to look at something that i ran in june this is the the last incident ran across 15 000 records you'll see a little bit more of a drill down in terms of the intent the number of records matched so this should match also what you see in topic recommendation and then you'll see the additional breakdowns here so these tools work hand in hand um and they take an intent discovery essentially will take in this case the itsm taxonomy and build these clusters that are associated with it so i know a lot of times folks that are new to it see the two different tools and the relationship isn't always that clear so hopefully that was helpful just to cover that i know we have five minutes left um just wanted to open up and see if there's any additional questions or anything that might be useful to cover as part of this session uh no i know it's a pretty short amount of time so uh yeah jason we have one question from tyler are these recommended topics being created in the global scope or specific um perhaps in its scope itsm scope yeah so for these recommended so if we go back and just show you really quick so if you look at for example the password reset recommendation this one they they are created today um in whatever scope that uh you're currently set up as but usually this is going to be global currently um but if you see topics that you might need um to change into a specific scope of yours you can certainly always duplicate duplicate that as well but that's a good question um i i need to con i i'll probably double check and maybe that's a follow-up with the topic recommendation team just to see if that is something that can be configured awesome thanks um as a reminder everyone the recording for this will be uploaded to community site and to youtube uh it's the servicenow community channel the virtual agent academy playlist later this week's so if you came in in the middle don't worry you won't miss us you won't miss a thing um okay awesome anything else jason otherwise i'll switch over to i know he has some additional resources for folks and uh and perhaps a slide just to leave up during for more q a yeah i think that that pretty much wraps up most of the things i wanted to cover um maybe just one other thing is if you when they're when this uh when you when topic recommendation supports additional areas you'll see different tabs for those so if you had like hr and csm in the future those will be options here the other thing too is if you have requests that you're running you'll also see those as a separate section here and lastly one thing that's worth mentioning is that the recommendations are going to be driving based off of content that is being shipped by the itsm team and so my recommendation is always just to check for updates and see if there's a newer version that you can install in your instance to get the latest content awesome any updates we should be looking forward to uh either later later this year yeah we are doing uh we're working on a lot of content around software management software access software installations um so that's definitely something that uh we're looking to try to tackle over the next uh couple releases a big focus area of ours very cool um any other questions otherwise jason mind if i grab the share and i can just share some of the additional stuff you had yeah sure cool let me go back to here and uh as i mentioned here we do have the documentation so whether you're looking up more information for topic recommendations or on intent discovery feel free to check out our documentation uh again thank you uh very much if there are any questions we'll or i'll let me just leave it back to this slide uh keep it up in the chat otherwise uh thank you everyone for joining today uh we will uh you know again the video will be uploaded to youtube and virtual age virtual agent and nau community sites soon uh you know please stay tuned we also have uh academies for hr and we academies for the mobile platform as well so once again you know visit the community site visit youtube for all that great content uh we are you know enabled you know enabling you as much as as much as we can because we want you to succeed and uh that's all i've got uh we'll leave the uh recording on for just a few minutes more and then we'll sign off thank everyone