all right we will slowly ease into our announcements while we wait for more folks to join I'm glad everyone was able to find the newly branded AI academy uh as some of you may know this used to be the conversational interfaces Academy time slot but um the world of conversational interfaces and AI have merged over the past uh year with generative AI Etc so we decided to make a lot of Senses to merge oury as well and you know maybe one less Academy slot for you to choose between when uh looking to learn more about service now and our wonderful capabilities so that let's go ahead and get started um as always uh for more answers and information FAQs and guid check out our AI generative Ai and intelligence Community forum and if you're watching this on YouTube click that like And subscribe button that way next time you can get all the uh videos whenever they come out you'll get notified that they've come out all ourmes are recorded and uploaded to YouTube uh you can see all our Academy videos uh past and present in our uh now Community YouTube channel uh Please be aware that we are going to mention a futur looking uh road map so when we do uh please take a look at the Safe Harbor for forward looking statements so Academy Basics this is for you you know we're here uh to give you fresh ideas better understanding practical guidance uh you know we do try to do a bit of slides but then also a bit of uh instance and demo and and uh exercise work as well uh as I mentioned recorded and if you have any questions feel free to use that Q&A function in the in the uh Zoom panel all right um another question you may have is well you know what about um conversational interfaces material that isn't uh AI related for example agent chat advanced work assignment Side Bar hey we're still going to talk about we're still going to talk about those it just won't be in this Academy we are going to have a we do have a uh platform Academy and workflow Academy those are uh existing or newmy and uh you know those items will be uh discussed in those otheris now so don't worry we're still going to talk about those just uh not in this Academy anymore or at least not as not as much all right uh here's the breakdown of our um Academy today what we're going to talk about and the main topic is the generative AI controller with November enhancement so we talked about the uh generative AI controller last year I think I around the October November uh time frame when it first came out well since it came out we've introduced some enhancements that I'd like to talk about today so with that let's go ahead and get started a lot of uh stuff in chat I haven't got a chance to take a look at it yet but I will later hopefully just a bunch of hellos and happy New Years so you know and to you happy New Year as well again if you have any questions about what your scen go ahead and use that Q&A uh feature and uh L or I will take a look at that uh as certain pauses of our Academy Okay so for those of you who are new to the generative AI controller let me uh describe it to you so very quickly um so the generative AI controller allows you to embed generative AI workflows uh in your service now platform without any integration complexity which allows you to achieve faster time to value for generative AI development so as some of you may know we do have now assist as a um Sol solution available to you now assist for it HR or CSM and we do a lot of cool things like summarizing your cases generating resolution notes it's so so on and so forth but what if you have your own um generative air workflows that you're looking to uh Implement right that's a little different than what analysist has perhaps you have your own subscription of azure open AI or open AI that you want to use well the controller allows you to use all that stuff you don't need to write your own API you don't need to to test or you know strength test any of your um own Integrations we provide for you one for you out of box you just you know provide your API key in our secure environment and you can start building flows start building um you know virtually conversations or or or script something if you'd want uh all available to you with your uh connection right we also provide some out of box capabilities uh in terms of prompting we do we do automatically prompt summarization uh you know summarization uh content Generation Um some uh what's it called sentiment analysis right or if you just want to to use a gen generic prompt and ask whatever you want you can do that as well please take note though that in order to use generative controller you do need uh at least Vancouver patch two and onwards and you do need uh Pro or Enterprise Plus licensing so the licensing to use the controller is the same that you needed to use now so let's talk about what's new in November so the November release was Vancouver patch 4 I believe and onwards so these for these new features you do need uh Vancouver patch 4 and upwards to use uh and I believe it came out November mid November I think before Thanksgiving um so the the the first thing that we released is support for Google Cloud generative a integration so September we we enable integration with Azure and with open AI but uh for those of you who are using Google cloud and if you want to use their vertex AI or maker Suite that integration is now available the other um enhancement we made in November was um enhancements to the pii uh masking and unmasking so we do have a sensitive data Handler and in September in September we did released that as a feature the problem was it was when it was masking information as it should it was affecting the response quality right so for example you can mask email addresses but uh if you mess email address then the controller wouldn't know what the email address was and it would start hallucinating a different email address or work um you know excluding it all together so in the November release the pii masking were tokenizing the mask data so the data is still masked still stripped out from uh from the eyes of the large language mod model but it is still retained uh on the platform for the response so you can see in this in the spreadsheet or in the screenshot there you'll still see the email address right in your response even though it is still being masked in the uh in the inference and in the U request response uh between the user and the large language model so as you can see here the address is is still masked but it's still uh provided in response however you know this is not foolproof so if that email address or if that information that you had mased was contextually relevant right if you expected it to do something if you expected L langage model to do something that email address it wouldn't be able to do anything because it's just going to plop it back in it's not going to be able to do any processing on something that's masked internally okay so I hope that Mak I hope that makes sense but you know we'll we'll we'll go over an example in the exercise here so that I'm going to show you two things how to set up Google Cloud connection and then how to set up a sensitive data hand or how to set up sensitive data Handler okay so that let me pause my share to switch screens momentarily and go to here resume share okay so uh this is my instance I'm on Vancouver patch I think well at least four if not higher and let's show you how to uh integrate with Google Cloud before that I'm going to just check the Q&A is it possible to sync this AI with SharePoint where the knowledge is I'll come back to that um let's talk about Google first so the first thing to do is uh very similar to what we do with uh Microsoft Azure which is we have to set up the conect credential uh the connection of credentials right so we're going to go to the connections and credentials page where is it there it is connection and credential aliases and these are all the systems that you can integrate with including uh you know azers here open AI down here and Google vertex AI is here and this is available to you out of box do you do not have to to create this record so we click into that maker Suites there if you need to use that and then the next step is to create is to click create new connection and credential I've already done I've already done one here but I'm I just do this again just to show you so you'll notice that um we just ask you to put in all your information here uh this is all auto populated so this isn't accurate really but you got to put in your Google Cloud project ID and then your credential name you just give it whatever name you want this is my you know Google cred you can see I've done it before already and the allaf name client ID secret and uh the redirect URL which is again already provided to you you don't have to touch you do do not have to touch this so where do you get all this information so let me go back to this the slide here and uh you'll notice these are uh two separate screenshots here let me just put a border on here so you know what I'm talking about um this is what your Google Cloud console looks like and this is where you get all the information the uh client ID name you get here uh you do need to put in the authorized redirect URL here as well so you know this needs to go into here and then you have to click save to add that URI uh your client ID is here your client secret is here and then your project ID when you you know I've called this my first project your project ID is here okay so this is where all the information is at once you fill all that out you click create and get oo token and you'll get this uh connection right record okay this stuff is um is optional although I think a project I think this is all auto populated but uh it's not required uh so then in order to get an all token you go to here Google thread you can see my o refresh token was available and will expire in a couple months but once everything's act um connected you just simply click get a token I'm not going to do it here but when you click get o token you'll get a popup that appears that says you know hey do you want to connect to your Google Cloud you'll get your you you'll get some popup showing your Google Cloud um login and then you click okay yes agreed and then uh this uh your o token will be generated for you uh you just to keep in mind you may want to you know you may want to log into your Google cloud in a separate tab for example beforehand but it'll ask you to do it anyway when you when you click on that okay so that's basically how you how you do it simple so then the next step so you know you've created your connection you've created your o token the next thing you want to do is make sure that your um your generative a capability is going to use Google so in order to do that we're going to go to CIS one extend capability so for those of you who watched um the last previous summarization we did the same thing for Azure so uh these are all the Avail capabilities available to you and if we go to next let's go to summarize and then go to one extended definition configs and you'll notice these are where you decide whether to use open AI Azure or the one we want is here uh summarize using vertex AI Google B chat completions completions is I think their uh lower version so you want to use their higher version uh so let's do here if I open this record change that to default and click update and you're good to go I've done it already here make sure this is active make sure this is default and make sure you turn off the other one or make sure this is the only one that's that that's true everything else should be false if you set two records would be true I don't know what happens you know it's probably look it's probably look at the draw who knows right um so yeah just want to make sure you use that and then let's make Let's test to see whether or not this works so let's go to flow designer here I'm going to use a flow designer example which I which is I which I believe is the example I used in the um in the previous controller video let me get some water while this is while this loads all right uh let's go ahead and summarize work notes which I again is what I think I did last time or what I built last time so as a review very quickly we have a trigger to an incident State changes to resolved and one it does we wanted to summarize the work notes so we're going to we selected action summarize uh we want to summarize this workn notes field and I'd had to do some uh you know string trans information of you know just because it's it's not originally a string format so if you're curious about why and how we did this just go ahead and watch that previous video and then we update the incident record with the summarization response um and we Po in the resolution notes okay so let's go ahead and test this and I'm going to select record that I have in my back pocket here oops come on now click run test yeah this instant record probably has some work notes already so test is finished running let's view the execution details complete complete complete that's good and then if I go into the a summary you'll see that uh my action status was successful which is good my provider was Google bard which is good and then this was my response and then this was the original text to summarize okay and then if I were to go to my Google Cloud console I'll most likely see an API call being made you know you want to confirm that just you know be aware this is is GNA this is going to um you know charge against your Google Cloud usage so make sure your subscription is all up to date there same same thing for Azure open a Azure open AI or open AI okay so that's how you use and set up the Google B uh the Google Cloud connection I know they say Google bard which isn't technically accurate but that's just the naming convention that we use we may update it later down the road but for now that's what you're going to look out for okay so that is how you uh integrate with Google Cloud so next let's talk about that sensitive data Handler so I go to all I go to sensitive data Handler there it is all right and all this is available to you out of box you could create your own if you want but you know we have a pretty good healthy you know list here whether it's email whether it's um uh credit card social security number all that stuff and in order to activate one of these sensitive data rexes uh which you know no one's a fan of here but you know if it is uh then you click on here click on email make sure it's active I wonder where is the active well I'm gonna I'm going to assume it's oh here it is I was looking for a checkbox status active and categories you want to add Global is always going to be there you want to add generative AI what and once you add it using the you know you use the magnifying glass and you put in uh add generative once you add it and you click update it's going to start working so um this this is uh relevant to the controller now the thing to note this the controller is used for both now llm or now assist and for the and for the custom workflows like the one I just showed you so it's GNA it's going to do it's going to work on both okay so that said let's go ahead and test to see what this looks like now what I expect is that the you know even though I've asked it to mask the email email address right um and it says oh number of characters mask I I you know you can say the the five or however many right um I expect it to still appear in summarization okay and let's go ahead and test that now it's not going to work all the time but who knows we may get lucky uh let's see I want to use this one this incident here which is all about basically requesting a coffee mug um notice that I tried to you know drop email addresses in here as much as possible you know ship it to here ship the label to here put it there put it in the um in the work notes all that stuff so when I click summarize let's see what happens okay that's a overly short summary there let me see if I can generate one again okay so there you see coffee mug request four and the email address is maintained uh in in September what would happen is we would strip it which we would mask it which is good but then you know we we didn't tokenize that data so it would it would have said you know coffee mug request for blank or you'd probably get these you know get these uh start out which isn't the greatest so here you can see we've still maintain it even though we've asked the sensitive data rejects to uh strip it out okay so those are the enhancements coming in November okay again you need at least patch I'll mention it here you know you need at least Vancouver patch four and with that that is all I have for my prepared remarks so I'll flip to this slide and I'll go over any questions you may have so if any if you any questions go ahead and put it in the Q&A and I'll go through them one by one okay so the first question from Anonymous is it possible to sync this AI with SharePoint where the knowledge is so that is on our road map we currently we currently can uh perform AI search with a SharePoint connector but if you're asking about now assist in AI search in order in other words like generative AI search results which I believe we've showed previously uh that isn't available yet but it will be um you know hopefully hopefully soon just don't know exactly when yet uh yeah I hope I'm pronouncing that correctly asks could now assist automatic convert date and time information in journalist to current users format and time zone um so I'm not sure I don't know if now ass cist can do that however um the controller can basically do whatever you want it to do in terms of a generic prompt so if that if this is something if conversion is something that open AI or Azure open AI can do then you can just use the controller to uh create a flow or a script to convert uh date and time zone to whatever you want and then like I did in a flow designer uh you know po post it in a uh in um in a field or write it to somewhere so um you know in terms of in in terms of using the controller you probably could Katie asks C the current lawsuit between the New York Times open AI have an impact on the open AI option um probably probably it would not impact us because we're M we're essentially in the controller at least making a call to open AI so if open AI makes a change we just follow along their change really we just call upon their change uh that lwuit doesn't really affect now assist and now llm because we did not use the New York Times or we didn't we didn't use any periodical when uh training or fine tuning Our model we used uh permiss you know our own service now data as well as permissively licensed data so ultimately doesn't affect us it may affect open Ai and if you're an open AI user it may affect you uh but we simply in terms of the controller we simply make a call to it to enable to sorry to enable Google Bart Anonymous ask do we need to buy this option options separately so uh as I mentioned previously you do need to purchase your pro or Enterprise Plus skew to have access to the controller and then you do need to purchase a subscription to Google Cloud however they are very generous with free credits which is what I'm using so you don't need to um you know do that there but otherwise those are the two things you need to to look at in terms of purchasing uh Ragu asks can we utilize AI on the active incident to display similar closed incidents and generate a draft article based on the conversation from the incident so KB generation is a a feature that is on that is also on our road map for now assist uh it is possible and I've done it before it is possible to use a controller to build a workflow that like like you know does what you say summarize or generate uh a knowledge-based article or some sort of document or text based on similar closed incidents so you know your imagination is your limit there in terms of using controller and then using summarization or you know generic prompt along with scripting or um flow designer to uh build the workflow that you want so you know won't won't say no because anything is possible with the power the now platform right um but otherwise if you're looking for an a more outof box solution we're working on that it's on our road map terms of KB generation uh Adam asks do you have available white papers for implementing the genni controller I'd like to understand their organizational effort involved um yeah so if Adam if you view the November uh AI Academy video and if someone could you know do me a favor and Link it it's on YouTube or their Community Forum um I show you how i' I implemented that in less than half an hour as you can see here um if you want more information on how to implement it uh you can look at our documentation we have a lot of information there as well as our community Forum which I believe L has dropped a link in uh there so you can look there as well we provide a lot of uh you know what's new setup guides best practices fq Etc so I would start there uh yanos again asks can it be integrated into a stream component that displays activities um I'm not an expert and stream component I'm not sure exactly what that is so I'm sorry I can't answer uh that portion there uh Victor asks is there a walkthrough like this for AWS so we ask AWS is not one of our uh out of box providers so unfortunately we we don't have one available like out of box Spokes and connections like I showed you that's included um let me think here it yeah it would have to be a manual implementation so you'd have to probably create your own connections and create your own spokes uh to connect to the apis ETC so unfortunately Victor we don't have one uh planned yet uh if you have any more if you have more questions feel free to post them in the KE a I think we have a couple more only a couple more minutes REM uh remaining um Anonymous asks what securities are in place for now assist so um good question so in terms of the in terms of the controller that's really um you know up to you because it's your subscription and it's your connection to open AI or Azure open AI so I I am aware that or Google cloud and I am aware those providers do have some security in place in terms of how they manage your data but just be aware that you know when you use the controller you are sending your service now data to open AI or aure open AI for their models to infer off of if you're okay with that then that's good you know if not then you may want to use now assist what securities are in place for now assist well we use our own models uh we use our own data centers and Ai and AI um servers so uh your data never leaves the service now ecosystem uh when it when it comes time to you using uh now assist okay um Brent asks I have a customer largely because operating outside our service now scope they like to use our virtual agent as interface for our chat bus we have a list of external chat bu we can surface through virtual agent and how this is best done so yeah a bit off topic but you know lucky for you I'm I'm also the virtual agent product manager so you would have to use the virtual agent API to make connections to other chatbots and the outof box and right the outof boox implementations that are allowed to to connect other chatbots uh would be Microsoft's uh chatbot um IBM Watson and I'm trying to think here I think Google as well uh dialogue flow right uh those are the three that we provide outof bot implementations for uh for any other chat Bots that are out there in the market you would have to it' have to be more manual implementation using again using our virtual agent API a bit higher code there unfortunately but you know it is possible you have to just keep in mind that it's um you're bit on your own there in terms of building that in in integration uh let's see uh chea asks generative summaries of KB articles as part of road map for which version launch uh we don't know yet uh it's targeted for gosh I don't even know I'm I'm thinking I believe sometime this year again Safe Harbor it is it is currently an innovation lab that much I know which is meant which which basically means internal testing for select customers but I don't know when it will GA yet but when I do I'll let I'll let you guys I'll you know I'll let the community know is now assist learning Cross customer so that's a good question and what I assume you mean is is now assist uh being trained on data from from other customers potentially and the answer the technical answer is yes um these customers have in the past have they've opted in we've cleansed their data of any classified information of any pii uh you know we service now we are our own customers so you know technically very much so yes um in terms of other customers when we do um again we've we've we synthes not synthesized we've cleansed all that data uh and we do use that in our training or fine-tuning of our Larn language models because um you know customer data is more domain specific and relevant to other customers so the answer is yes are you able to opt out you're able to opt out as well and that's very easy to do you just click a you know toggle and you can opt out so um technical answer is yes we do but again it's it's it's very much cleansed data when that we used uh can we use AI to automate decision trees I can take that one uh very quick answer because we don't have a lot of time but I would recommend looking at productive intelligence or task intelligence which is a product that used to automate triaging of cases if that's what you're looking for thank you L last two questions I will call it uh Victor asked does now does virtual agent use now assist for customers it does in September and with more enhancements in November which I will share in a future AI Academy so stay tuned there Victor and then last question from Raymond has the masking tokenizing feature been reviewed or approved by federal clients so our federal clients our regulated markets um we do not support with now assist yet u in terms of like all the you know check boxes we need to fill out there so uh it has not been it's not yet Raymond but it is also in progress and in works there all right a lot of good questions a lot of good information thank you everyone for joining uh we'll see you soon again I think it'll be the same time slot just in two weeks from now and uh you all have a good rest of your day thank you everyone