all right let's get started everybody Welcome to AI Academy my name is loik Sanchez I'm an outbound pm on the platform AI side and today it's very cloudy in Pennsylvania I hope you all have a better weather before we get started um we have a few additional resources to share with you as usual um remember that you can ask all your questions on our Ai and intelligence Community Forum that you can access with a simple Link at send.works AI you also find a quick start guide FAQs and other type of resources like that on the Forum so I recommend checking that out whenever you have a question and the good thing is that it's available all the time so you don't have to wait for an answer from anybody this recording this session is recorded and then posting posted on our YouTube channel and we invite you to subscribe to our service now now Community YouTube channel so you get all the latest news and all the other videos from the other groups some information here that in two weeks the AI Academy session will be again undocumented intelligence and we'll cover some technical tips and troubleshooting so I want to give a heads up that it's going to be a fairly technical question uh it's going to be a fairly technical session sorry but if you are interested in um more deep Dives and into your more on the development or on service now admin side that's very helpful if you want to understand the document intuition product a bit better and we all take a lot of questions and ask you which topics you want to cover so I highly encourage that you join and share that with um your colleagues that are interested in the product and also just uh shout out to the developer Advocate Advocate team who did a document intelligence live coding happy hour last week or two weeks ago I don't remember but it was pretty great the recording is available on YouTube on the developer YouTube channel but um I recommend also checking that out it's pretty pretty fun session as usual before we get started just remember that we when we talk about the product in the sessions we we try to cover available features that you can already use but if there are any questions or if we do get into the roadmap or features that are coming we meant that as informational only and we ask you to check with your account team before making purchasing decisions on information that we provide as usual some of the basics of the academy uh what is an academy it's a Content that we produce for you we want to bring you fresh ideas to Foster a better understanding of the product and give you the Practical guidance to to use the products these sessions are recorded and posted on our YouTube channel so don't yeah the it's better if you want to re-watch the session later if you want to post that's uh the best way to to go about it but if you are here with us today a great opportunity for you to ask questions and for that we ask you if you can please use the Q a panel inside of the zoom box and we'll answer all of your questions today and with that being said let's jump into the topic of today we'll cover what's new in Utah for document intelligence as I said previously my name is Lloyd Sanchez and I'm an outbound product manager on the platform AI side kind of quickly how the session today is going to go we are just going to cover a quick overview of the features and the slide deck that we are seeing right now but really the most of the session will dispense looking at the instance so we'll spend about 20 minutes of the session going through the new features how to configure and how to get started and we'll get all your questions answered along the way and some dedicated time at the end as well so with that let's get started with document intelligence for Utah so you might know that document intelligence is a store app that is available via a page queue that you can get via a licensing model and that's not new so that's uh it's always been like that um but what's really new in Utah is uh making it easier and faster for you to extract information from view documents to drive workflows and we have different ways to to achieve that now so the first thing we introduced in Utah was a better support for extracting information from tables I'll show you that in a second is a new UI that helps extract value from the the data that is organized in a table or in a list in a document and the better Ai and OCR models in the back end that are processing that as well likewise we are introducing support for extracting checkbox information from from document and finally last but not least we have a new application called document intelligence admin that allows you to set up and configure and monitor your document use case within one interface so we'll show that later it's pretty it's pretty neat it allows you to to get guidance along the way and manage all your different use cases in that view which is so a different view than the typical list and forms that we are used to see and just one clarification when we talk about Utah since it's a stop we actually are talking about different versions of the plugins so that can tell for Utah actually we're talking about the version 2.3 and the version 2.4 that came out during the time frame of Utah and then for document intelligence admin that plugin uh same thing we had two different releases in that time frame version 1.0 and version 1.1 so if you want to get all these new features the first thing you have to to do is to go in your plugins page and install the latest version of uh of the plugin all right the first new feature that we mentioned is the document integers admin so it's an admin experience that allows you to set up your documentation's use case um and the great thing is that you get all the analytics about the solution as well as guidance and uh all of that in in one nice interface so we'll show that later um but yeah and then uh the the other so fit the new release that we mentioned is the improved table extraction and really what we see here is on the front end so you see the UI of extracting table has been improved to give you more uh space on the screen to to do the extraction um and we also have some improvement as I mentioned in the back end to make it more performant and reliable on that side and we are introducing extraction of checkboxes in that version as well so that's a very 2.4 and we also show that later in the exercise but you can Define group of check boxes and extract them again with a it's a new new piece of the UI and that allows you to to extract checkboxes additional things that were covered in that release is uh support for additional languages so before Utah we only supported English and French and now we introduced um German Dutch Spanish Portuguese and Italian and what does that mean what that means really is that our OCR model is capable of recognizing all the symbols and letters of those languages so that's a slight small variation of symbols between those languages but we do fully support those additional so German Dutch Spanish Portuguese in Italian now and the other thing we want to mention is that uh we in the previous so there is an update in the terminology so previously we were talking about for example a tax task definition which was the configuration of your doc Intel use case well that now is called a use case in starting in V 2.3 um and then a few additional terminology updates so key we used to just to call a key and now it's called field the key group is not a group and um a few other that we see on the screen here but really if you use so if you're new to Doc Intel it's not going to impact you um the documentation site has been updated with that but if you already use the solution in the past and we're looking for that just remember that there's a quick change there if you use the the admin really you don't really see that it's uh or embedded in that experience all right and that's it for the overview of the new features and we're going to switch to the exercise and just to give you some context so we can go quickly and fast together I'm going to use an invoice and this is the invoice we see on the screen that's going to be the invoice I'm going to use for the exercise and we can see that I want to extract date and invoice number and the customer so those are filmed at like the invoice level and then I want to extract a table so to extract tables I'm going to name that different line items and each line at item is going to have an item name the quantity and a line total and then forgot to mention that total the total of the entire invoice is also for the whole document and then we are going to extract the check boxes as a group here that I'm highlighting that on on the screen all right so with that being said let me go into my instance all right so as I mentioned uh before we start we just want to check that our plugins are installed so I do have 2.4 and 2.4 for the version of document intelligence and 1.1 for the the document intelligence admin plugin here so once I um I check that I can now navigate to document intelligence and the menu here as uh slightly changed from before but we see that we have Administration and then we have home and use cases so I'm gonna go to document intelligence admin home and this is the admin experience that we mentioning previously so you get a what how to use dock Intel you get the video here to guide you an additional links to the documentation and other questions to consider which is great when you get started for the first time it provide you guidance here once you're ready to get started and you can click on that button and you can actually create a new use case so I'm going to go with that so I want to create a use case for my invoices and I'm going to use a table that I have to to link between document intelligence and a backend table here so I can save that and now I see on the screen all my use cases and I can open the one I just created okay yeah yeah forgot to mention that you saw that pop-up the extraction mode by default was set to recommendation so it means that um at first we need to extract the fields with the help of the tool and once we are confident that the extraction is it's going smoothly then we can move to autofill and straight through processing it we'll see later but by default now it's recommendation and so I'm on the screen for use case and I can Define my Fields so the fields were the fields that we were looking at when we were looking at that invoice and so if you remember we had a date the date so field type is text and then I can select the Target Field for that which is the invoice date because I'm going to I know I'm going to create a few fields in a row I'm just going to check that box here it says Define multiple fields and that's just gonna let me uh the pop-up is going to show again what sorry it's like add field so I can conveniently create my second field uh so we had a date we had an invoice number that I can map to the invoice number field we had the name of the customer and we had the total value of the amount the total amount of the invoice so I mapped that to the field all right so this is that's it for the fields that belong to uh like the higher level of the document and then I want to extract the fields that were on that table so to do that I need to create a new field Group for the table so I'm going to create click on that plus button here and as I mentioned we want to name that line items and now we see that for the field Group we have different types we have field Group we have checkbox list and we have table so in that case we are extracting a table so the right the correct type is stable and when I select table I can actually select the my target table because we remember um if you watched previously but essentially I'm going to create new records here in that additional table and map that to the invoice task so it's a related relationship in between the two tables I can add the field Group and once I added the field Group it's available in that menu there so I can select it and then create the fields for my table so if we remembered from the invoice what I wanted here is the item name each of it's for each of the lines in my table I'm going to look for If an item name I'm going to map to the item name field on my table that I just selected foreign just going to go faster because it's not the most interesting part sorry so quantity and the line total area all right so I created the fields for my table and now I want to set up um the fields I want to extract for my checkboxes so to do that I'm going to create a new field Group again so in that case I'll click on the plus button again and here on the type I select checks checkbox list and then the the name of that group was a whether an approval is required so I'm going to name that approval required and then I can create a value for each of the checkboxes on my on my group so first I select my approval required field Group I see the field type here just change to check box and I have one called um level one and then I have a level to 2 and whether we want to escalate that case here we go all right I created all the fields that I needed so I'm ready to process my first task so I'm going to start the processing because if you you know it takes a few minutes for that to be processed by a machine learning server so I'm just going to trigger that so we have time to look at something else and then come back to it so to do that I just upload my invoice give that task a name and I can click on ADD extraction to start the processing and so while this is processing let's look at some of the other features of that admin home here so I can access the settings I have settings that are general for document intelligence for all my Solutions so for example whether we use an exact match option when typing Fields there whether we show the candidate score the confidence level here I can slightly tweak the way my side panels are showing on my screen as well as the width and the width of the extraction panel and the thumbnail here as well as the the way the document fit into my the doc Intel workspace so those are the general settings for uh doc Intel so they apply to all my use cases and then I have settings that are specific to my use cases so the invoice one that I just created I can select the extraction mode so as I was mentioning by default it's created as a recommendation meaning that I'm going to have to start typing in order to see a value here if I'm confident that I've done that with a few documents and um I've good results in the confidence level of the predictions then I can move to autofill which is a mod in which we set up the threshold and then if the prediction matched the threshold then the value is directly displayed in the field and to do that you just have to change the mode here and you set up your thresholds and you're good to go for now you're going to stick stick to recommendation and then finally um by default the agent will have to verify the extractive fields but once I have run a few hundreds documents I can enable straight through processing which basically does the extraction without validation and again we add a threshold that needs to be made for for that to happen now these are the settings for my use case let's go back to the task that we created to see if it's ready so we can open it now and we see that we have that button here opening document intelligence and it's gonna show me that um the the space in which we do the extraction that is we are familiar with and here so we have a the filter we set up so we have the date and we can start the extraction by selecting the date the invoice number and again here the predictions of a percent of zero because it's the first time I'm using this use case and this document so it's expected here my customer is Physio and the total of my invoice is that number here's four choose six four and I want that value here okay so this is for the regular text field on the document and now I have my checkboxes so you see that we introduced a new type of UI to extract checkboxes and here I want to extract so I look at my document and I see level two here is checked so I'm going to select level two and then through the UI I'm going to select which of those checkboxes I am looking at so it's that level to approval required here so that's for the checkbox and then we're mentioning we have a new UI for tables so I set up my line items here to extract the table that I see here on the screen and to do that so now I'm going to open this table here and proceed with the extraction so I see the the data in a row now which is closely much more closely the value that I see on the screen and I can stop with my extraction so maintenance contract is my first item name and it's a quantity of one and a total of one three three six and I do have a second row in my table um and here you can see it gets way smaller because it knows it's the next row so I can save some time on my extraction and the value is 2928. all right once I am done with that I'm ready to submit my document task um the two of the checkboxes that were not checked so I I'm aware of that and I can just submit anyway and if I refresh here I see that my task was processed and it's um it's done now and the extraction was done and so the last thing that we could that has changed in the admin experience now is also the way we do integration so how we integrate dock Intel with your workflow so um that was available in the list view before but now it's even simpler that's to the way to create the flows to match back-end tables and new document extraction so you can set up your Integrations you have two types the process task and the extract values when you set up the process task it really it creates a new document intelligence task whenever you create a record in your Source table but you could add a condition there and it you create a flow associated with that so the call that create a task I can save that and then um once we create tasks and we process the document we want to copy the values from our document task to the backend table using the field mapping that we've been doing in the configuration and that's done via the extract values flow here so I can select extract values foreign flow is enabled and I can clicks on Save so that's the way I create my two Integrations and uh also remember when we create those we have to then open them quickly in flow designer to activate them so by doing to do that you click on on the flow link here it's gonna load flow designer in that screen foreign button all right if there are any questions don't hesitate to put that in the chat or in the Q a sorry all right this is taking a little bit of time for designer foreign just opening the flow in flow designer and it's nothing out of the ordinary but we do need to click on activate so that it works by default it's not activated so that you know you keep control on that all right I don't think it's gonna load but that's that's okay it's just really it's just the flowing flow designer it's uh nothing that you have not seen before just need to activate that um but yes uh other than that that that's it there was an overview of uh document intelligence in Utah so the document intelligence admin as well as the new features uh table extractions checkbox extraction and uh we didn't showcase the new languages but uh with the with uh yeah there's a question so if your target table is a custom table um the question is your target table a custom table yes it's a custom table if so others that work with licensing uh so the use case um I'm presenting is um it's kind of like a Creator workflow type of use case so it means I basically I'm building an entire new flow for my for my for my for the sake of the demo here I'm building an entire new workflow based on a document so that's why I need document intelligence in in when you purchase document intelligence you need to to buy a package and it's uh available via automation engine so it's very likely that um you'll get some some tables available in that as well but yes you do have to check licensing making sure that you build your app in uh in compliance with your licensing uh and then we have a question is there any kind of OCR technology of or feature used for this demo yes so um so when I was showcasing how to extract the values so really uh I'm going to show that again but when I try to extract the date and start to I'm start typing it and you see as I hover over here uh on the screen in the middle I see that the date is highlighted is a purple square around it and that's the OCR so that's the OCR showing me on the document where my value is um and and so it's actually this is where the OCR comes in play and then as I select my value I'm also saving that for my AI model to be trained on that so this is really the OCR piece and aips working together but this is where we see that yeah so I hope that answer your question all right and there's a question if you watch the previous Academy we were showcasing how to how it behaves after you trained a few documents I didn't want to show that again today but essentially after a few documents the the predictions are uh like when you move to autofill your those values are already pre-populated uh because they are predicted so that's where OCR and AI works together to to accelerate document value extraction but go ahead if you have another question go for it yeah yes and that's uh again something we covered previously but really the the strength and and the value of document intelligence is that it can work with different invoice formats so if you have invoices from different companies as long as they have the same fields which is very likely right you want to extract um the invoice number or the invoice total for all of those they might have different formats meaning that invoice number might be at a different location on your document or uh or the total and things like that and as you train your model on this specific formats then the the model will learn and so you don't have to create a separate model you have to create it you don't you don't have to create a separate table this is uh all um available via that OCR and AI model no worries no approach is needed I didn't record that today but we uh at the two sessions last year and so um trying to be um cognizant on everybody level and not repeat things that I said in the past but it's a great product but yes all right if there is uh if there is no more question today I think we're going to stop the session here and I wish everybody a good rest of their day and remember this is going to be available on the YouTube channel if you want to watch the recording later and again in two weeks we'll cover more of the technical details I do have one more question before we go will this work with inbound actions yes it works with inbound actions actually so again if your use case is CSM you should look at the CSM product um CSM with with task intelligence actually uses documentary to process attachment but if you your use case is different than CSM then you can build your flow to fetch the attachments and process the process them with document intelligence that's one of the use cases yes foreign thank you for joining and see you in two weeks bye