we have great content for you today and in addition to that remember that we always have additional resources that are available 24 7 on our Ai and intelligence Community Forum to get there one simple link sn.works slash AI highly recommend that you subscribe to the channel so you get all the latest updates on the contents published before we get started this is our safe Hub honor is and with that I'm gonna hand it over to shamers to start a session today ah wonderful thank you go here so let yeah let me um first going into our goals for today we'll have an overview as we saw we stated uh it'll be about 10 minutes then we'll jump into an exercise to show some live examples of what we're talking about and how to configure our result Improvement rules and then we'll finish off with a q a um again as Lloyd had mentioned the Q a is going to be open during this session so if it's still feel free to ask questions there as you as they come up great thank you Seamus yes uh so hey welcome everybody um great to see some familiar names in the attendee list so thank you all for joining uh we do before we get started this is um we're not covering the intro to AI search in this session but as like mentioned we have lots of those uh resources available to you via either the YouTube channel or as part of the AI and intelligence Community Forum so please do feel free to check those out as a refresher should you need to in order to get started and provide some context for this session specifically we did want to go over some vocabulary so that we're all on the same page so we're going to be covering things like user context result Improvement rules and basically we want to make sure everybody knows what we mean but when we say global companies so from a global company's perspective we're really covering instances where companies have uh multiple locations across different countries as well as a number of languages that are are supported in platform so those are really the the target audience for this and obviously if you're looking if your organization is looking to go multinational then you'll be able to learn a little something here too uh on top of that we're really going to be focusing on user context while the information is generally stored in the sys user table and we'll see a couple of examples of that later the user context is really provided as part of what we would call on AI search side the search time or query time activities so it may be of interest to you to know that you know the language can be updated as part of your session profile right in Portal so just interesting information to keep in mind there and last but not least we will be talking about these result Improvement rules um we like to say that it's really an opportunity to put your thumb on the scale of what happens either out of the box from a relevancy perspective that ranked order uh or as part of the ongoing learning um for for AI search relevancy so you still have some influence over what results are being returned uh and in what order and that is by these three three actions boost block or promote and and with that I know that Sheamus is going to talk through some of our core pillars and how result Improvement rules fit there we're using this wonderful thank you Gerard I appreciate that and thank you all again for joining us today I'm certainly excited to be talking to you about our result Improvement rules and as Gerard heads uh mentioned that uh we um this slide actually may look familiar to some of our returning attendees um but with respect to AI search we identified generally three pillars for success and those pillars are essentially to keep the experience natural where we look to answer any questions across multiple languages in various formats and then smart where we we actually leverage our AI capabilities to provide those you know relevant search results and finally easy which is obviously we're all looking to make our configuration uh of the of the platform easy as possible and how essentially result Improvement rules will work to Aid in achieving those goals uh first we'll look at you know natural where we have these rules that are with configurable triggers and conditions for Boost block and promotes and also AI search supports these results Improvement rules in all of our various languages so as as Gerard had alluded to this is a mechanism for our as admins to put the proverbial thumb on that scale to improve relevancy um uh with obviously supporting these multiple languages and um the perspective of being smart and bursting search results for uh document matching elements from the user context so what this will allow us to do is to to gather the information about the user context and and dynamically apply that to these these particular rules and the uh these these actions um will actually uh add a supplemental waiting to the results uh the search result order uh produced by AI search uh and the Machine learning relevancy features um there is one note that the rules for uh block and promote will actually override the default set of search uh results um by actually blocking them or promoting one or more particular results and finally easy uh we want to have an intuitive management experience for our search AI search admins and each Improvement rule uh as there there can be many if if desired each rule will allow you to perform perform a particular or a single action essentially each rule can only have one type of action it can have multiple instances of that action but it will only allow for uh one one particular type so with that being said I'll jump over to our exercise what we'll want to look at first is to start by jumping into the actual um I search um user profile and this is where we'll start search profiles I'm sorry actually I already had that up and we'll I'm for exercise purposes we're going to be looking at just our default search service portal so this will help us walk through the actual configuration here of our result Improvement rules we can see from the search profile we have the result Improvement rules here and I've pre-configured a language and Country rule I'll jump into that and a few things that we can notice to start with here as it renders we've configured a start and end date we've pushed this end date out sufficiently far so as it's not going to come into play for our particular purposes and we're applying this to all languages and also we're activating it on all queries and as you can see in this particular configuration where we're boosting for language and Country for our our Global country our global companies we have three Improvement rules I'm sorry for three Improvement actions and we'll start with language and as a note if you're using or if you're in Tokyo or or Beyond and you have the bcp-47 with language and Country standard setup you may not need to necessarily go past the use of this language boosting but for all others uh this will each of these boost actions will be relevant and useful so if we start by looking at our language configuration some may be familiar with this but we'll work uh walk through each of the configuration items here and when we said we're alluded to these actions of being smart and it's connected to the user context what we can see here is we're boosting by user context and we're looking at the knowledge table initially for the content and when we're looking at this particular action we're looking to boost the language and as it matches our search users language and we have a boost weight of a thousand and essentially that will double the weight of that particular piece of content so as it's the weight is identified through our the search results this will apply an additional um boost on top of that obviously to Bubble this content up and similarly for our country boost again this is a another dynamic action where we're looking at the knowledge table and we have um a country field this happens to be a list field configured on the knowledge table that will allow for any particular article to be associated with multiple countries and that will match in this particular action we're looking to match the the logged in users country designation can you maybe remind us what what is the range of number we can use on that boost weight I see it was a thousand um what's the range of fat essentially a thousand would be doubling the weight of that and you certainly can scale that back this is just putting a pretty strong weight on this particular action so anything um above 100 and um between a hundred and a thousand again this is something that may require some um some testing but those are the the Boost weights that we're looking at this point good question thanks Luke we'll bring that up and and just for for additional context for customers that may have configured boost rules prior to quite recently you may see this guidance is significantly lower than what you may have configured and that's due to some really more intuitive back-end changes on Boost weights so basically the step function is 100 basically a weight of 100 is 10 so you can incrementally add a 10 boost essentially per 100 so excellent thanks Jordan and finally we have a potentially optional boost action here where we're gonna um it's a very um it's a slightly different setup with regard to this particular action and it would only hold true for uh uh sort of it's a global country boost and this could be useful um for a global corporate policy documents where regardless of where the individual may be located or identified as their their particular country may want to actually boost some documents for everybody and if we were to do this we could add a value to that that same country you know country ISO field for whether it contains global and that would allow for that particular content to be bubbled up as well following the same sort of strategy with our boost weights and again that would be an optional boost action okay so that takes us through the actions and the rule itself I'll actually start now with looking at some impersonation so we can so we can actually see this in action and full disclosure on the system that we're working with it's a fairly limited set of data but certainly adequate for the purposes of this demonstration um what I'll do is I'll just bring up our knowledge table here so we can get some sort of an idea as to what we're at least the expectations of what we'll see from us with regard to the results uh specifically looking uh the first impersonation I'll do is an individual that has French identified as their language of preference in the in the user assist user table and when we search for something um we should see this French content bubbling up to the to the top here so impersonate particular user and I'm going to be using the service portal for this exercise here if we look at address so if we look back at our knowledge table we're going to see that we have some of our French contents coming up here oops sorry uh apologies I during that impersonation I I lost my knowledge table um but essentially this particular article um was identified as a French language article and we'll see that uh coming up to our top of our search results and similarly for the language I'll end this impersonation and we'll look at a another user where we'll look at the language mm-hmm this particular user their country is Germany I think it's probably worthwhile after I do this to show you again the knowledge table and our German content is bubbling up to the top here um again I'll jump back to our our knowledge table but essentially what we're seeing here is that um the the country um boost is being taken into into consideration and the uh and the content is being boosted up uh within our search results and that sort of concludes our exercise I will um certainly happy to answer any questions that we see yeah did you want to show it the United States in Seattle they were these are in the syllabus French or Germany sort of back here if I I think it was on the second line up there yeah yeah that that's the one you look you were looking for right correct yeah so what I may do is maybe I'll sorry it's a little cumbersome to uh have this up and then do the impersonation so what we'll do is we'll search for spam so essentially we should see um this article uh KB 11 and we should see we should see 29 not necessarily up in the top of the list so we should see 11 as we're looking at a German and we'll look uh for uh zero four so let's do that just as a as a demonstration of them here hopefully that page is there so you can jump back sorry about that we have our user here we'll search for spam and this is okay so this is um hopefully our knowledge table still up yes okay so we can see here we have three results that would uh add a comeback for spam um and what we should see is 11 and our o4 should be at least one and two and that's good yeah and our um our non-sort of country identified piece of content is is last in that list um if I ended this impersonation and rerun this same query my what will happen is the result order will actually change um that may be a good demonstration as well here so we can see as our default user myself without a country identification we can see that the result order did change with regard to the the priority or are waiting um being applied to these particular results this is great um do we want a quick clarification around the change that will be coming or or came already for real the weight is that something that's already there or is that something that we are becoming yeah so so the results um the waiting like you've mentioned that is actually rolling out uh to the the back end um I may actually look to Gerard to identify if that's a rolling window or how that's going to be identified yeah great great question and this is part of the the AI search sort of infrastructure so AI search as ever as everyone May learn going into some of the intros um AI search really consists of three different um uh platform deployment features so the first is the version of servicenow your instance is on so if you have a compatible instance you're great you're good um the next two are any updates that are part of a store app so things like the advanced management tools and so on and then there's also that that back end right that AI search engine um that's that's actually running behind the scenes and so the upgrade is currently in flight at the time of this recording but starting um everyone all customers should be on the upgraded back end by next week so that means that if you're viewing this recording it's very likely that those boost weights that that we talked about the Thousand the 100 and so on will all be in place for you and uh operational so great thank you and so just to recap today we saw how to use the result Improvement rules uh based on Dynamic user context such as users user specific language or specific country to boost uh the results of say in knowledge articles table we saw how the Android the Boost weight of 100 is related to that 10 increase and then a thousand would be doubling so a hundred percent increase of the the result likelihood and then we walk through the example of setting up the result Improvement rules for those knowledge articles for a global company with users uh from different language or country and the process to do that was to set up a result Improvement rules and then to set up three different boost actions and then uh you should actually use impersonation to test the results I think that's a very important step in in the process of implementing those problems all right and if there is no more question that would be it for today thank you so much Sheamus and Gerald for joining us today and uh I hope everybody got great Knowledge from that we'll be looking to publish the video on YouTube and make sure to come back to the community to see more like that thank you thanks everyone thank you thanks all