just a reminder if you have any questions please feel free to post them in the Q a section I'm here today with Marcel gotel and we're going to cover the AI search Tokyo release part um Marcel good morning good morning all right thanks Nabil for the introduction so our goals for today we're going to go through some of the Tokyo Delta items that are available in this release we're going to look at some store apps we're going to look at some family release features and then we're going to show a few of the capabilities live in an instance and take a look at how those capabilities work what you can do with them and talk through some considerations and then we'll pause throughout with q a but as we mentioned please post questions in the chat window we'll be answering questions throughout and pausing periodically to address your questions so let's get started what is new in Tokyo and I'm just going to put up the Safe Harbor notice in case we discuss any forward-looking information or statements um but I wanted to share first what we're going to cover today is is the AI search component but it fits into a broader landscape of some new AI capabilities some of you've seen who have been around for a while and some that are relatively new as of the Quebec release but this is where we're going to focus our efforts today but wanted to share with you where this fits into some other AI capabilities that can yield tremendous value for your organization so as we talk about what's new in Tokyo AI search came out with the Quebec release and you can see that we've had a robust trajectory through the Rome release the San Diego release and then finally in Tokyo some of our Flagship features include the search and next experience we also included search-based autocomplete Chinese support catalog item genius results Improvement and finally the AI search tuning application we also have some other machine learning updates as well for AI search but these round out the the top items that we wanted to share with you today so AI searched for next experience what is it and why do you need it and this is really a build on a more of a global take on machine learning based search throughout the servicenow platform so you're going to get the same relevancy based results regardless of where you are in servicenow you could be on a portal you could be at a service operations workspace you could be in a CSM workspace all of these results are going to be promoted based on relevancy scores consistently across all of these different uis how this addresses value in your organization is it helps reduce mean time to resolution it improves employee experience or net promoter scores because it's surfacing the most relevant answers so it helps resolve issues faster connect you with information in a more consistent fashion this also accelerates search and findability with Dynamic filtering and sorting so you'll notice some of the search facets that you see on the left hand side of some of the screenshots are dynamic they can be filtered and they do change based on the search table that the the results are being populated from I do want to mention that this is under limited availability currently um so this is available through the store app it's something where we're approving customers to Pilot this give feedback share results and for version two which is expected in Q4 we will include an auto migrate Zing configuration feature to get you over to AI search in a more streamlined fashion so a lot of value with the next experience moving on the search-based autocomplete this is a an enhancement that many customers have suggested so we do listen to your your feedback and your inputs and this basically gives a more google-like experience so when you type something in seeing different result options that you can either execute on the query that you typed or you can select from the autocomplete options and go directly to that knowledge article or that catalog item or window so it's a more direct and variety very based experience based on what you're looking for so if you find something specific you can you can go ahead and navigate to that much faster than going through just a query execution and then finding your specific result and so it's designed to really match on title um with the highly relevant suggestions and promote titles and just help you find information faster essentially AI search provisioning and multilingual search improvements so there's a couple things packed into this capability first and foremost we now support traditional and simplified Chinese within AI search so this is something that has been a big request across customers I'm going to share the language language spread in a minute so you can see all the different languages supported but I do want to mention that the auto provisioning for AI search has been streamlined a bit faster so again this applies to hosted in GCC instances starting with the Tokyo release so in Prior releases you had to go ahead and request AI search this is essentially Auto provisioning it in new instances and it gives you the ability to search content across all languages so making sure you get that personalized consumer grade experience with search in those languages that your users are requesting and I do want to also call out it mentions stop words and synonyms in in the the support for simplified and traditional Chinese as well so this gives a sense of all of the different languages and capabilities supported through AI search currently so you can see we've added traditional and simplified Chinese what what features support which capabilities but q a genius results are currently supported in English only just make sure you're aware of that but if you're looking for a multilingual conversational AI deployment there's several languages that are currently supported today moving on to q a genius results so we've added enhancements with the Tokyo release so we've added a higher q a coverage without manually annotating the Q a genius results so it's a bit more Dynamic it's context sensitive ranking and the quality improvements are there we also allow you to tune and review some of the quality improvements with the search tuning app which I'll mention in a few slides but this is something where we've added significant improvements and continue developing in the Q a genius results space so the AI search tuning application so many customers often ask well how do I know if search is working how do I know if it's performance how do we compare it Zing versus AI search and the tuning application which is available through the store is something that allows you to gauge and measure and understand how performant AI searches in your organizations whether you're looking to migrate from zinc to AI search or whether you're looking to compare different search profiles across different configurations of AI search this allows you to tune to categorize or label and understand search performance one of the machine learning updates I want to mention is the relevancy pipeline so this is something that has we continue to enhance and improve over time so what it does is it really gives better auditing and query processing a relevancy model training Behavior so you're going to see better relevancy scores what that means is more relevant articles articles that match a little bit closer to the the query or the search execution that the user is typing in um there is a new capability we call keyword feature so for those of you who have done meta tags and um some different calculations using The Zing search engine we are now bringing in meta and tags on catalog items and knowledge articles Etc into something we call a keyword feature to drive relevancy across the search results that get rendered or that come up based on a search query and then I do want to share we've increased or we've added rather boost weights for AI search Improvement rules and so this is something that's mentioned more explicitly in the knowledge base article I mentioned so if you want to take a minute and jot down that knowledge article it is 1170030 what this does is it allows you to improve search results using boost weights so um with some of the machine learning changes our our algorithms have gotten Smarter on the back end so what we're doing is we're suggesting you use higher values in your Boost rule weights to make sure that you see the same or better behaviors and so we're recommending starting an evaluation of a hundred we now recommend starting an evaluation of twenty thousand so this just means that the machine learning on the back end is getting better and so we're essentially suggesting you adjust your your improvement rule weights up and make sure that you measure and and you know understand the impact of that experience so it is a good thing um it's just something to keep in mind as you're tuning and configuring AI search going forward and then finally AI search analytics so this again continues to receive improvements and enhancements with every family release or rather every store release it is a store app but this allows you to gain visibility into search and performance look at Trends collect insights on search coverage and just really understand user behavior and click-through rates so what populations are leveraging AI surge are they liking AI search what queries are popular are there any gaps in content that you want to identify so the analytics really give a deeper insight into some of the user behavior and how it's being exercised across the servicenow platform so at this point I'm going to go through a live instance of some of the capabilities in the Tokyo Delta I will take a pause Nabil are there any questions we want to address at this time I see there's a couple questions in the Q a in the chat yeah I see them coming right now uh okay so the first one uh how would genius behave in context of military language experience given the only support English um Rahul so yeah so when it comes to the other languages it just brings up the regular results it doesn't show up any of the because there is no genius results to be presented for the other languages it just shows the regular results got it the other question is regarding AI search will be enabled for customers that start in Tokyo or will it also be enabled if we are upgrading from older version um for that yeah probably in Tokyo if the instance is brand new if it's on Tokyo yes uh can be automatically provisioned but if it's uh upgrading from the older version I think we need to be requesting that to activate Asus because already if the as such is not activated we are using Zing then we need to explicitly request for that activation got it does that mean that uh all Tokyo instances gonna have ai search enabled by default or no all the new instances I think they are totally new you'll still need to enable it but the process will be much faster because the infrastructure needed on the back end will be available to you already so the nice part is you know there used to be some lead time could be a few hours it could be a day or two but going forward if you have a brand new instance in Tokyo uh it will be provisioned uh AI search will be provisioned alongside that godhead thanks Gerard so what that means is that um customers still need to opt-in but the provisioning will be way faster if you're on Tokyo correct awesome thank you very much Marcel back to you okay great so we're going to go through a couple exercises um just to demonstrate some of the type ahead search we're going to take a look at Outlook issues and you can see that based on how I'm typing I'm getting some different results that populate so if I just type Outlook I make it a broader scope of results coming back so troubleshooting create case create contact Microsoft Outlook issues but if I just do Outlook issues I'm going to get something a little bit more specific here so I could essentially hit enter go to that query or I could also just click on Microsoft Outlook issues which would take me directly to that result so what I'm going to do is just hit that execute query so you can see how the results display and you can start to look at some different facets here we have on the left so based on catalog items knowledge all um but let's go ahead and take a look at how this would display with a genius result because what we're seeing is just the the results coming up with just a snippet of some information on that item whether it's knowledge catalog case Etc but if I were to type in a query how do I set automatic replies and hit execute query I'm going to get what we call a genius result so this is going to show that google-like result automatic replies out of office I can view the article directly from here but it's going to show up as bigger because the result is relevant now we're still asking if the suggestion is helpful yes or no this does imply the continual feedback loop from user suggestions but this basically shows if I'm in service portal what that experience may look like if I have genius results turned on and what type of of items are rendering within that service portal itself now jumping out of the portal going into some of the different admin screens let's take a look at some of the next experience capabilities so you may be familiar with this landing page it was new um so again this highlights next experience you may see this in different workspaces or portals or UI in your instance and taking a look at some of the search capabilities here so if we were to just type in email and hit that query what I'm going to see is different results on email notifications and nothing particularly specific now what I can do is go to knowledge and what I wanted to share is some of the dynamic faceting so what we can do is if I wanted to just highlight email I just wanted to see the results there notice the facets change based on my selections if I just type highlight Microsoft I may see something a little bit more specific so what this does is it makes search more relevant and more Dynamic based on what I'm looking for and especially if I'm looking for a specific incident I can come in here and see all the incidents and I may want to filter by critical and see okay which critical incidents contained email um so again it's searching by Machine learning relevancy so notice I'm I'm pulling up the most relevant articles but I can certainly look at the the newest articles added the updated articles added and again this experience extends to different workspaces so whether I'm in the service operations workspace platform analytics workspace any area I have turned on in the platform that search is going to be consistent across all of those different areas now this applies to specific incidents problem requests cases as well so if I were to type in incident I can see an exact incident match or I can see all results for incidents so what this is doing is it's letting me search on very specific areas so if I wanted to pull up this particular incident for inquiry or help I can do that or I can go directly to um you know something that would pull up anything that contained incident of this area so this just shares some of the flexibility in the search whether I have a keyword a phrase or an incident or a case number um this search experience is going to be dynamic and it's going to be a little bit more intuitive than maybe you've seen in past releases so let's say you know I'm pretty happy with search but I also want to maybe have had it live for a little bit and I want to understand what the performance is I want to understand are there ways in which I can make it better or I can see how well it's performing um you know based on some certain results so if we go into our tuning app and from the all menu if you just search AI search and we go to the admin home page under AI tuning application this is something that's new for a store app so it's currently available through Innovation lab today and it's something that lets me see just how perform it searches so if I wanted to for example look at Zing search versus AI search to get started I can do that if I wanted to look at AI search you know in one capacity maybe I want to look at the the service portal profile compared to the virtual agent profile I can certainly do that too so this is something where I would go ahead now I have one in progress if I hadn't set this up I would see a get started here but since I already have something live I can view the setup and it gives you more importantly what to expect so you're going to set up and it's going to tell you what role you need to set that up it's going to rate some of the search results so again this would be results labeled or categorized by viewers that you select to go through that exercise and confirm if the the results coming back match or if they're not relevant and then finally there's going to be a review analytics phase so again in this middle section of rate search results we highly suggest you select subject matter experts since your administrators and personas setting up the um the search tuning may not be very specific with domains if you have those turned on so let's go ahead and take a look at some of the configuration so in the getting started what this is doing is it's giving me a road map of areas I need to set up so you can give it a name you can describe what it is so maybe you want to understand search tuning before I actually tune it and I want to understand if there's any drift in my model I want to understand if there's any new patterns and data that I'm not accounting for here and I can set a date for reviewers to complete so if I want this done by next week I could certainly do that we have our deadline set for today um and then moving forward it asks what Baseline do I want to set for my application now this is a decision point so if I were new to AI search I may want to use Zing as my Baseline since I've been using that for quite some time and understand how does AI search improve results Over The Zing base but in this case I've been using AI search and I really want to understand how does the portal search profile compare to the virtual agent search profile one caveat is you need to have the same search profile because what you're doing is you're trying to compare apples to apples and seeing which profile may be more performant or you may be trying out a new search profile and you want to see if the results are better using this new and improved search profile or if the one you had previously is optimal so moving on to stage two you have two options you can select from search logs this pulls from the search event table um I highly suggest that you use this option if you wanted to do Advanced configuration this allows you to select a table so if you wanted to use open nlu predict log or service portal log you can certainly do that as well but selecting from these logs will give you um you know the best the best and most variety of results unless you have that specific to a table or configure table of your choice so now that we've loaded Search terms from that table I can see if the terms came back so I may want to add new Search terms here because you know things like email or iliama or leave may not be good Search terms at all right because these are so open-ended maybe I'm expecting nothing to match with these particular terms so maybe you want to add terms that are a little bit more specific or a little bit you know matching some content I have so I can label appropriately but notice 74 results came back now I can choose to select all 74 I can choose to select a subset of that and so I'm going to leave it as is and continuing to step four we can assign reviewers now you want to make sure the reviewer has the right roles and permissions to go ahead and review and categorize or label some of the results coming back um so those reviewers would get notified they would have access to go ahead and review that we'll take a look in a minute and then we're going to continue to the submit now since this is already going I have a save option here so I can save this mid um Mid process I could make adjustments to it or change it but in this case what I'm going to do is impersonate the system admin so you can see what that review process looks like and so as a system administrator I'm going to have access to some of the same spaces so what I'm going to do is go in we can look at AI search all over again you'll see we have the same tuning app but now I'm going to navigate over here to the reviewer home page because this is the space where I'm going to be essentially going through the exercise of making sure that um you know I'm labeling the the results appropriately you'll see we have a compare a completed exercise here so maybe this was the exercise that we previously ran comparing AI search to Zing to First launch AI search so again we completed this process now we're in the process of tuning different AI search profiles so I'm going to have the option to resume reviewing here and what it's going to do on the left is pull back all those Search terms that you saw populate in that initial configuration window so you'll see about 20 Search terms at a time and if I select any one of these I'm going to see the the currently reviewing um search query and then below that I have all of the items that are coming back from that search query so what this is doing is it's essentially moving the spreadsheet exercise into the platform so I have the the two one zero so if something's an exact match it's useful or it's irrelevant or shouldn't have showed up in the first place I have the option to select any one of those and if I'm not happy with the results coming back I can certainly click add and select additional results from the list here as well so whether that's catalog item knowledge article Etc so I have flexibility to do that as well you'll see most of these have been completed we're going to go through and take a look at create email signature and what this entails is I have about five or in this case eight results that I want to categorize so if I'm looking at create email signature I might say item designer category request well you know I'm not sure what that is so I'm going to put irrelevant a new email account maybe that's useful since I'm asking about an email signature automatic replies maybe that's useful create an edit a contact group yeah that could be useful depending on what I'm looking for um troubleshooting you know I don't know if it's troubleshooting but you know it does contain email um it's certainly not phishing and okay this one says create email signature I'm going to say that's an exact match that's something I would hope to show up first and then a service category request well you know I'm going to say that's irrelevant so notice as we go through that exercise I've completed eight of eight and if we switch back over to our analytics screen so I'm ending the impersonation as the reviewer so assuming the reviewer is a subject matter expert in that area what we're going to do is go back and pull up our AI search tuning now we're going to look at the reviewer or the admin home page rather and we're going to take a look at you'll see this analytics it's grayed out but if I hover over this I can pull up the analytics now I have some information on home screen but this is really where I want to focus my attention so sometimes this takes a minute to pull up but this will show overall progress and this will also show search quality so I can understand my profile on the left is search portal profile on the right is virtual agent so I can see exact results in top five and I have inline guided support here as well so if I wanted to compare search by search terms I can see those terms side by side someone that renders I can see whether service portal was correct or whether virtual agent was correct what the quality was and just get a sense of at a high level how those stack up against each other but maybe I'm more interested in specific performance maybe you want to understand you know with the recall bringing back all relevant articles here and so I have more information about you know item genius results down here so I can see you know for service portal default application what is my correctness versus missed and it looks like service portal is slightly better than virtual agent maybe I want to take a look at some virtual agent scores but I again have guidance on on what that calculation is so if we click in and review results we can see some of the assessments so how to troubleshoot Outlook was a Miss uh create email signature was amiss so that's where some of that data labeling helps understand are the results correct is it performing well and how do you see that stacked against some of the other profiles that you're you're possibly evaluating for tuning so I'll pause there I know we have about a minute left Nabil did we want to go ahead and take some questions yeah so I think um I'm just gonna ask Gerard because he had answered some question here uh if you would be able to explain that so regarding the Arcane and faucets uh category to subcategory would you mind uh answering that oh sure yeah the answer is is very quick so the question was is it possible to have a hierarchy in facets today basically having a category or subcategory and we have something similar for topics and subtopics um the the short answer is it's not available today it's something that we're looking into in the short term we are doing a better job with um our our topic you know pulling in topics and having better rendering of those um and then in a future even future release will have a better higher they'll support it so awesome thanks Gerard uh the other question was for Rahul and uh Rahul if you don't mind reading out the question it's it is regarding the search bar between and the difference between Zhang and AI search yeah shot um so generally if we when Asus is not already been used we do use the contextual search of The Zing in the photo and most of the times within the portal if we change the widget or make any changes after the upgrade they don't get revert back to the out of the box widgets so it does happen in such a way that we need to make sure whatever the widgets they move back to the latest or they use the latest code and in that way we could expect some changes in there be it in the UI or whatever the results being come up because even for the a search to be used within the portal we need to make sure those widgets are updated so I think I shared one of the link where we can reclassify the widgets because most of the widgets will be cloned so that's why we need to make sure they would be having the latest code from the widgets of as such great thanks so just to make sure that uh you need to validate that after turning AI search um Marcel I think these are the questions that we have um and just a reminder if you have any questions please feel free to post them in the Q a if you still have questions um while you watch this recording please visit our community and post the questions there we'll make our best to answer it as soon as possible great thank you Nabil thank you Gerard thank you Rahul and this concludes AI search you can see that we will extend future sessions toward other capabilities that work well with AI search and thank you um appreciate your time and attendance today look forward to seeing you in two weeks thanks everyone