AiBuilder Review OTO Upsell Login Bonuses By Daniel Adetunji

AiBuilder Review OTO Upsell Login Bonuses By Daniel Adetunji

You can have an opportunity to receive extra bonuses if you finish 2 steps below:

♦ Step 1: Order the product by Clicking here.

ClickBankProfit FE (SEE MORE)

Upgrade #1: Giveaway Module
https://clickbankprofit.biz/upgrade/giveaway.html

Upgrade #2: Leads Module
https://clickbankprofit.biz/upgrade/leads.html

Upgrade #3: High-Ticket Module
https://clickbankprofit.biz/upgrade/highticket.html

Upgrade #4: Traffic Module
https://clickbankprofit.biz/upgrade/traffic.html

Upgrade #5: Reseller Module
https://clickbankprofit.biz/upgrade/reseller.html

♦ Step 2: After your purchase goes through, email your receipt to me at alexhuynh.1110@gmail.com so I can confirm your purchase and email you information about how to access to your bonus material.

https://www.linkedin.com/pulse/ai-builder-oto-review-full-upsell-details-coupon-login-aibuilder/

https://www.linkedin.com/pulse/ai-builder-oto-review-aibuilder-otos-details-bonuses-daniel/

hello everyone this is raza dorani welcome to the r d show hey everyone this is daniel christian and thanks for joining us absolutely and uh today our topic is all about ai and talking about ai uh i know dan spoke at a user group uh this week wherein he was demonstrating about ai so dan would you like to tell us what you did this week this this week has been a very ai-centric week for me um i got invited to speak at the portland ug i think it was tuesday night and i did it with ai and power automate it was a very power automate-centric one and even over there i mean i had to you know cover it at a bigger level um and there’s there’s a good audience over there a lot of good questions so there’s definitely a lot of appetite and demand for this topic and has grown um one of the things i pointed out to them was that you know it’s not something new anymore it’s been there for a year and a half so it’s it’s maturing it’s it’s you know it’s definitely tried and tested um so but such amazing questions came up over there so i actually told them hey i’ll answer the best but join this one because we’ve got you know a really good uh person to speak on this topic so i personally am very excited awesome um ai builder is is a topic that is pretty close to my heart as well actually when um when ai builder came out for the power platform uh you know it it peaked my interest and i was i was one of the few lucky ones in the mvp community who got the opportunity to work with the product team on the initial uh components of it when i was you know brought into uh given private preview access so i had uh you know i had got an opportunity quite early and then i have done a few projects over the years on on ai builder from uh and i’ve used a lot of those components form processing business card scanning uh all of those components and i know it has changed a lot in the last year there have been a lot more additions to it a lot more power to it uh of course i haven’t uh kept up to date and i’m sure you know you have not two so we have someone who who lives eats and breeds ai builder and uh you know dan and i thought about uh who better to call on the show uh than for me mr ai builder himself or joe fernandez from microsoft hello joe how are you hello reza hello daniel super excited to be here likewise it’s an it’s an honor to have you uh on the show today joe nice nice yeah and then it’s it’s indeed like i was i was listening uh uh remembering when we first connected raza on a builder like almost three years ago i guess it’s uh if we look back indeed the product has evolved a lot and then i was saying we’re also seeing a lot of more of more and more interest from users and customers around the air builder so i’m super excited to uh today give an overview for those who you of you who might not know about their builder and for those of you who know uh explain what’s new and there might be even some surprises at the end i like surprises yeah so joe uh thank you once again for joining us on the r d show it it means a lot for you to come here and showcase all about ai builder i know there’s an audience for it and uh it is something that everyone needs to know about so well the floor is all yours uh let’s get started with this uh joe is here to show us about ar builder if you have any questions posted in the chat put a queue before the question so we know you’re trying to ask a question and we’ll try and bring it up now or at the end of the show so make sure you do that and uh let’s get started awesome yeah thanks service uh and Agenda daniel and yes it’s friday it’s a fun show so let’s keep interactive keep those questions coming and i will take us as many as as we can uh so on the agenda today uh what i’m planning to do is to give an overview of of a builder for those of you who are less familiar on what the builder is and where air builder sits within the whole microsoft ecosystem and probably the best way to see what ai builder is capable of is by just using the product so we’ll jump right into our demos and see different capabilities so we’ll keep it interactive and then we’ll have time at the end also for qrnap so to answer any questions that you post uh during during this session we’ll we’ll go ahead and answer them so uh first of all ai builder sits within the poor platform i think many of you know where it was where is the power platform but just in case support platform is microsoft local no code platform and what does that mean it’s it means it’s a set of solutions and tools that you can use to build your own applications automations chatbots dashboards without having to have knowledge on programming so if you know how to use powerpoint if you don’t know how to use excel you should be able to know to learn how to use these tools very very easily and we have four big applications in the platform the first one and probably the most well-known one is power bi where you can build dashboards and and more all data that you have in your in your company the second one is poor apps where you can build mobile desktop and web apps as if you were building a powerpoint basically with with a drag and drop and interface and you can use xo like formulas called poor fx to to build your logic then we have por automate which is all about automating tasks so let’s say that you receive emails with attachments and you want to extract those attachments and store that in an excel file with portament it’s super easy to to do that and then our most recent member of the family is poor vital agents which allows you to build chat bots especially like in support cases we’ve all interacted with with a company support website and there’s usually a chat bot so you can build very easily those with pro vital agents and integrate them across many channels like like teams messenger and your own website and then we have at the bottom what we call these horizontal capabilities of the platform we have the data connectors which allows all these applications of the core platform to connect to more than 600 services both microsoft and microsoft services we have dataverse which is our local data storage solution and then the topic of the day the topic that we’re spending the rest of the session today is ai builder which is all around making easy to add ai capabilities to to these into these applications and automations one question that we get a lot is like hey i’ve heard about azure ai i’ve heard about azure machine learning i’ve heard about the ibuilder like it seems like microsoft has a lot of offering around the eye a lot of investment and the answer is yes indeed microsoft does invest a lot a lot of ai and the the wonderful thing about the or microsoft cloud is that you can start from wherever you feel more comfortable so within dynamics if you’re at dynamic 65 users we do have already built-in echo abilities you can leverage out of the box without having to configure anything if you’re a citizen developer so you use the word platform and you want to use ai for your own business process you can very easy leverage a builder as we will see now in the next slides and demo and then if you’re a professional developer you can go deeper and just go into into azure ai and use the quality services and the hrml platform and air builder is built on top of of azure ai so everything that all the great uh advancements and capabilities of azure eye we make them available in a very simple no code interface within ai builder so this is very uh joe sorry to jump in here this is very important to note because when i’ve worked on ai builder i’ve received this question a lot from customers is this a new service is this a new product uh yeah it’s a it’s an encapsulation but an encapsulation from an existing service that has been that has been delivering ai and that’s azure so i love the fact that you put the slide out that clearly shows that ai builder comes from an existing feature set of azure ai is just encapsulated for citizen developers which i’m sure you’re going to demo today so thanks for bringing the slide out yeah that’s uh that’s correct there’s a yes so uh what are those ai abilities that are within a builder so we group them in four pillars the first one is around documents and this is a very popular one that we see a lot of demand among our users and customers i’m going to dip that a little bit more in the next slides the second pillar is around language and text so we all work with a lot of text whether there’s our emails or teams messages and or sms and there’s a lot of knowledge in those in those texts so we can do things like set them analysis so be able to know if i receive an email from a customer is the customer happy or not just by running that sentiment analysis or if i have customers from multiple countries and i only speak english i can use text translations to translate all those emails for for me and another popular case for example is uh category classification so i receive a lot of emails and i want to route them to the uh to the right person within my company i can build an air model to understand the content of the email and route them to the right person then we have appear around images and this is all about being able to detect and count objects and we see that a lot of use cases especially around the retail so let’s say that you have shelves of objects and you want to quickly count the number of objects in those shelves you can very easily build an air model that when you take a photo it’s going to count and identify and detect the objects in those in those shelves and then the the last pillar is what we call structured data so this is when you have uh data in your database like for example you have a sales data of your customer and you want to build an ai model to learn the patterns of that data and predict future outcomes so let’s say that you have that table that database are on sales data and you want to predict whether a customer is going to renew a subscription that they’re going to churn based on the historical data that you have at the company you can with the help of ai you can build a model to detect those patterns and predict uh what might happen and take the appropriate actions the structured data that’s the fairly new one correct that’s that’s what i’ve noticed too is the documents language and images those those were there for a while it’s the structured data which i thought was like the the new thing so it’s it’s a super good point daniel it’s it’s it’s uh it’s the less well known one it’s been already since a few months but it’s uh it it doesn’t get as much attention as the others probably because it requires a little bit more work so you need to have like structured data in your company you need to have access to that data so usually not not everybody has access to their sales data or constructs data so it requires a little bit more sophistication so that’s why it might be a little bit less accessible to regular citizen developers uh but yeah it’s it nonetheless it is a super powerful feature that we have absolutely predictive analysis has been a heavy you know in demand for ages take existing data and then learn to predict the future so i i see big potential for that one yep last one yeah all right so yeah let’s uh let me talk a little bit more about documents and why we see so many customers interested and excited about this and probably many of you watching this uh can also relate on use cases What is OCR so you’ve already heard about ocr which stands for optical character recognition which is basically you take a pdf for an image and there’s this ocr technology that is able to extract all the text from those documents and this has been available since decades it’s been uh we’ve been available since many many decades uh but the challenge is that uh one thing is to get the text out of the document but the challenge is how do you get the exact information you want so here for example we have an invoice that we ocr and we see all the text from the invoice but i just wanted to know i just want to extract the sales tax of the invoice before i had to build like very complex code or rules to be able to to get that data and it was very fragile because if the the structure of the invoice changed then the rules will break so um getting text out of documents has been here but getting the the right information from those documents has been very challenging but now with deep learning and artificial intelligence basically we have powerful air models that can look at those documents understand the structure of the documents and get exactly what we want so do we want a specific field the amount is going to give us that do we want to extract a table the ml is going to understand the table structure and give us the exact information that that that we want so and to add the cherry on top of the cake now we have low code so we’ve taken this very advanced document understanding models and we’ve added a very simple low code ui to be able to build your your your own ai solutions to extract the data you want in a very in a very simple way so we’ll see through the demo but when Models we go to a builder there’s different uh type of models around documents the first one we call text recognizer which is ocr so the australia that we that we’ve known for many years uh microsoft was the leading if not the leader in terms of state-of-the-art ocean technology you um send a pdf for an image and and the text recognizer model will return all the text including handwritten text and in more than 100 languages then we have what we call the standard models or the preview models so these are models amoros that we’ve already trained for common use cases so for example a very common use case is invoice so if you want to extract data from invoices we have a of the shelf model that you can take you send an invoice and we’re gonna give you back all the common fields and tables you can expect in an invoice without having to configure anything i will see that now also in the as well in the in the demo we also have one around receipts uh the one around business cards so so reza was saying that he’s done a few projects around around the business card reader reading and then we also have ids so we we have models that can given a passport we can extract the data from that passport or even a driverless from the us we can also extract those two data so for know your customer type of scenarios we’ll see a lot of interest on those on those type of ide documents and then probably the most well-known uh one model within a builder is called phone pressing so let’s say that uh so text recognizer is very basic for your use case there’s not a standard model available for use case you can basically build your own custom ai model for any type of document you want you’re gonna go through a very simple experience that you’re gonna teach an a model where to find the data you want to extract and the water is gonna learn about your documents and it’s gonna be able to extract and automate that that information Training Experience and we’ll see i’m going to go very quick with this because i think we’re going to see in the name of so this is the what we’re seeing on the right is the the training experience to train a custom form processing model we can do standard documents but we can do complex documents with with checkboxes with tables with all the 70 languages we you have in a single model documents with different layouts and different templates so that’s that’s all available within within for pressing and then the the beauty and the power of of being part of the poor platform is that then we can very easily integrate with or automate the poor apps so uh having a model by itself there’s not much value so okay model but the real value comes where you can put it in an end-to-end process so once you have the model we make it very easily to integrate for example in a poor automatic cluster to be to be able to build that end-to-end integration as well with perhaps if you have a more interactive type of use case like i don’t know i want to take photos of invoices and review the data in in a poor app in a canvas app i can also integrate that model very very easily and to make things even more easier we have what we call the document to mission starter kit um so this is a solution uh for you more more more technical ones uh it’s a driver solution that you install in your environment and we give you out of the box an end-to-end solution that basically is going to read attachments from an email it’s going to extract the data using either a form processing model that you train one invoice pricing pre-built model and then it’s going to show the data into a validation station that we built with powerapps that can be fully customizable and then you can decide to push the data using the connectors to any of the supported connectors that we have today or if you want to do things like rpa pods you can also leverage that using automate desktop but i think uh i don’t know what you think daniel already i think enough AI Builder slides let’s perhaps just let’s jump right into a builder and see this in action absolutely because there’s there’s good questions coming on this as well but we’ll wait until your demo is done and then we’ll see that sounds good so um if you want to follow along or if you’re watching the recording letter uh it’s very easy to get to a builder if you go to aka the ms slash try a builder and you sign in with your microsoft work account you’ll be able to see what i’m demoing today so you’ll be able to start a free trial and be able to use everything that i’m demoing now you’re going to be able to use it as well so if you go to this to this url you land into this page you can also access it if you log into proautomate.com opporabs.com you you will see that we have the airbuilder entry point within the left menu and we have this uh this homepage so for those of you who might know ai builder from a few months ago this is a fairly new page so we launched it i believe two or three weeks ago which does a number of things the first thing is it showcases the different uh ai models that we have that we described before so categorizing the in the four pillars so documents text structure data and images then we have tutorials so if you’re new to a builder we have a good set of tutorials that you can go and learn about the ai builder and we also have some more advanced modules if you’re looking for to skill up on your ai builder knowledge we also have some videos on online builder um overview on different topics like how to build a document automation uh solution how to leverage lobe which is a a capability to build image classification models and hey perhaps we could even take the video that we’re doing today then reza and put it here i think it might also be uh be a good a good showcase many times uh uh usually okay this sounds like this sounds super powerful like can you give me names of customers they use a builder so we put here some customer use cases so you can see customers that are currently using ai builder in a in a production uh setting and then there’s there’s always updates in airbuilder we’re always constantly updating a builder so you’ll find here at the bottom all the list updates and and the great thing about this new homepage is that everything that you see here you can try it out visually so let’s say that hey send me analysis it looks interesting to you you can just click on each of these tiles is each of these uh squares and you get this experience where you can basically right here and for free you can try what the product is able to do so let’s say if i were i don’t know the errand the show audience is awesome and i click analyze text it should of course say that it’s positive uh sorry because i could hear the sentence so i can very easily try out these these models without having to set up anything anything complex and joe for testing this out what you just did do i need a license or can i just take a trial license and and just try out these these different models before i go and acquire the license very good question that there is a trial license so everything everything you can try for free before purchasing a license okay and this whatever you tested right now just to analyze the text for example if i have the full version will this be counted against my license or this is just for trial purposes this one is just for trial purposes it’s it’s once you have once you use your model in a flow in an app in the in that moment will start counting against your license capacity you know that’s one of my personal favorites too and i was i’m echoing what uday already said is that you categorize it into those four categories and that that just helps kind of understand and visualize where all of these ais fall and i mean you spoke about that in your in your slide deck which is great but right here you see try out ai capabilities for each data type and you’ve categorized it into the four different categories and that’s just helps so much from a visualization standpoint so that’s one thing and the second thing is i’m in full agreement putting our video over there when it’s it is the best video out there because we have joe obviously that’s all i had to say you you guys are the rock star you guys are the restaurant that that that gives so that that gives so much visibility to what we do so it’s because of you that will put in that video great uh so let’s uh let’s try a few things extra so one popular case is invoices so i think Invoices in many cases we work at companies where we purchase things and then we receive invoices and we need to extract the data from those invoices so instead of doing that manually we can basically go ahead and use an a model so here we see some examples so very similar to the experience that we saw with sentiment analysis i get some some examples uh sometimes uh users might say hey but these examples like they look way too good like uh i want to see it like with my example so it’s super easy you can just click upload new here and you can take any any examples so let me take here one example i found uh earlier today on on the internet uh it’s an old invoice from from fedex i’m gonna see what’s able to do out of the box without having to configure anything what the model is able to detect for this this invoice and we can see that simple to do a it’s able to accept a good number information out of the box so things like we can expect like what is the invoice date who is the provider what is the total amount by when they need to pay it what is the invoice number so everything is is captured out of the box let’s try we also have another fun example to showcase that we also support handwritten so this is a very extreme case where the invoice is totally handwritten uh usually i think that that is no longer the case but you will see that uh the the the email will be able also to extract all the information that is uh that is unwritten uh so if we zoom in here here we see that it’s able to extract everything invoice id invoice date something to mention as well you see that we also provide the confidence score so you can do things like hey if the confidence score is lower than 70 percent i want a human to review it if it’s uh higher then i’m okay for the automation to just push the data into any system that that i have joe this is this is really interesting Invoice Processing so what you’re using right now is a pre-defined model it’s called processing so this is something that microsoft has put out uh i cannot train this model this model is pre-trained so all i have to do is just use it correct correct if if you need to to you to train your own model we have the phone pressing one so uh custom side yeah okay but can i extend this can i take the existing invoice processing model and maybe teach it something more is that a feature or i just have to go to the custom model if i need my own uh elements added to what invoice processing already has in play so that’s a feature in the roadmap so it’s not available today but it’s something that we hear a lot of demand for so it’s something that that we’re planning to add to the product today you can start with the with the with the pre-built with the pre-built will also give you back all the ocr results so you can pull like extra rules to to get the data that you want and you can also combine it if you can if you need extra information you just go ahead and build your own custom model and you can combine both you can say hey maybe for my for my invoice with larger usage or that are more critical to me i’m going to use the custom model where i’m going to train to extract exactly what i want and for new invoices or lower volume invoices i can fall back to the previous so there’s a number of combinations where you can combine both the previous and the and the custom one but today they’re separate but um we have on the roadmap the possibility to make it easier to breach one another got it and joe i’m noticing that it’s of course it’s doing ocr and giving me those independent values back and i’m assuming all these values come back at as text so even though it is date technically when i read it in apps and flows it’s going to come back as plain text that i need to convert into those respective data types is that is that a fair assumption from my point so it’s true it’s coming out as text but we also give you the normalized values so if it’s a date we’re going to keep you in a standard date format if it’s a number we’re going to give you a number format uh so uh we will give you both the the text as as it is written on the on the on the document and you know in a standardized date format okay uh again i’m echoing some of the questions that’s coming in the chat as well uh another one which i think is is quite popular is signatures right you’re gonna have forms or let’s say contract documents where i have actual signatures in there now i understand it’s doing ocr but uh is there an option or is there something planned in future where and if there’s a signature it can actually give me that just the signature portion back as a maybe as an image in base64 format so i could use it for some other purposes is that in the works it’s in the works so we have we have the technology to do that it’s still not in the product today but it’s something that we have on the roadmap as well the ability to extract today we can extract uh fields we can extract tables we can switch checkboxes and next we’re going to add support for for signatures perfect and joe one last question since i saw that you mentioned tables there are going to be scenarios wherein we have larger documents that have multiple tables or maybe nested tables and i’m assuming tables are returned back as separate entities when i actually query this model so i know that this is table a versus table b so the question is to you is what are the limitations there today and what can ai builder do in terms of nested tables or let’s say let’s say multiple tables inside my invoice love it yeah it’s it’s uh we see that we have a lot of passionate customers and users around table so it’s great to see this question also coming here in this in this show so today we have a pre-advanced table recognition engine which can do simple tables like like this one so this one would be a simple table like it’s it’s it’s very straightforward we also support like complex tables things like you have nested cells nested items within a table so even tables that are done for the classical column row pattern uh we can do that and i’ll show an example in later in in the demo and then there’s also a big demand for table that span across multiple pages so for that we now have an experimental feature that allows you to to to do that to to extract a table that span across multiple pages but it’s experimental this is something that we’re we’re working on it’s something that we’re working very closely with or with your advanced microsoft research team to to be able to come up with a good with a nice solution so it’s it’s a little bit more complex to set up than would like to uh but we’re definitely it’s definitely type of mindful for us and and we’ll have a solution uh in a few months to to improve the the multi-page stable extraction thanks joe awesome so Automate Process so as we said like okay the model i mean this is great it’s able to extract the information but it’s only useful if if if it does something end-to-end for me it’s also scenario so let’s do a let’s automate the process of uh i want to copy the data into this excel file so let me go ahead and back in this page you see that we have this using a flow button so i’m going to click on this and this is going to take me to a very simple cloud flow template that i can then customize to my to my needs and the flow you the template you’ll see is very simple it’s basically so every flow starts with what we call a trigger so something has to trigger the automation that’s though in this case it’s just me clicking a button but i could change to anything like okay every time i receive an email with an invoice every time i add an invoice to a sharepoint folder everything we can customize i’m just going to keep it with manual for this first demo to keep it simple and then uh we’re going to call the invoice standard or predefined model to extract the data and then it’s going to extract the data and send that data to me by by email in our case we said that we want the data to be added to excel so very simple i just have to use import automate i just click and add a new step in my automate flow and here i see all the different connectors that we support today in the platform model 600 connectors so there’s for microsoft services like we’ve seen a onedrive planner but of course also non-microsoft one we see some amazon aws here we’ve seen some google connectors in here as well uh in our case let’s let’s uh keep with excel so i’m going to click the excel connector and we want to do is to add a row into a table and it’s going to ask me a few questions like okay where is the actual file in this case i have it on my on my onedrive account so i’m gonna select one drive then it’s gonna ask me for the library in this case onedrive is always onedrive and then to pick the excel file where to put the data so here i have it and then if i have multiple tables it’s going to ask me which table in this case i just have one table in my excel file and then it’s going to show me the different columns within the the excel file and then what i do is when i click on each of the columns for automate shows me everything that is extracted by the by the model uh all the fields as well as the confidence course and to to the question i was asked before around hey is it only returned as text or as standardized values as you can see for example for for dates like service and date uh service start date we return both the text format as well as the date format that is very neat because that’s that’s very important actually many people struggle with that how do i convert it how do i write expressions i love that it’s it’s actually giving you different variations of of that specific field that you’ve you’ve trained the ai model for i like that awesome yeah it’s it’s it’s a very neat neat feature i agree um so let’s to speed up the demo i’m just i’m just going to add three uh three items so let’s do like typical ones right now the most id let’s do the total amount uh which is called invoice total there we have it so i can have three things i have the the total in text format the total in number format and the confidence score so i’m going to add the number format and let’s do i don’t know this is the due date uh another typical one in date format and that’s it so let’s let’s give it a try let me let me save it and let’s try this little automation and this one is very simple i just trigger by clicking a test manually but i could do i could have changed it to for example when i receive an email so let me confirm my connections and now it’s going to ask me okay please give me an invoice file let’s use i know let’s use the fedex one run flow and now the flow is gonna take that that invoice it’s gonna extract the data using the the invoice pre-built model it’s gonna add the data into excel and then it’s going to send me an email with the with the extract data so it started to excel and now let’s wait for the email and there we have it the flow is complete so if we go to excel we see that we’ve added those three fields that we that we said and if i go to my email account so this is not me but it’s my my demo persona it’s uh we see here the email and we can we can we can see uh the information they extracted i can open the document to see it bigger and validate that everything is extracted correctly and then something that we do in the other box template is we also put a link to a to a microsoft learn tutorial so if you want to learn how to make more complex donations or how to do more complex scenarios you can keep learning and see what the possibilities are this is often uh joe actually i used this for one of my customers exactly this now they had invoices coming in and emails uh use the use the custom model though because their invoices come from different forms it could be a pdf document it could be a picture that they take so uh i i took those and i took that information used the one drive connector to convert it to pdf gave it to the model which was pre-trained with those different uh invoice types let’s put it that way per customer and uh then yeah pick that information up put it in their data source and had all the data points out very similar to this demo so yeah this is this is really neat very cool very very cool to hear so yeah it’s i think that’s a great segue let’s look at the custom mode because we see a lot of customers using that that custom model so let me give you show you how simple it is to leverage as well and for this case i i don’t have invoices but they have sales orders so we don’t have a pre-built model for sales orders but as you can see and as as the the kids that you were saying raised up for your customer um there are different shapes different layouts so the different they come in different forms like tables and different formats like this one for example we’ll see that we have some nested items so we’ll see how we can address that i would want to do is um we receive these sales orders into into this um into the sharepoint sharepoint [Music] folder and something i forgot to open but i’ll open that real quick we want that every time that we add uh access orders into this uh into this um into this searchable folder we want to copy the data into excel and then we want to enter the data into this legacy website as you can imagine like i received like hundreds of these every every month and it’s a very boring and repetitive work they have to do so i love to be able to automate that so i can focus on doing more more interesting stuff than just coping data from excel into a website and all that so let’s see how we can use a builder and the power platform to automate all that good so we’re back into a builder in Create Custom Model this case since we’re doing sales orders orders and we don’t have a pre-built model for sales orders we’re going to just go ahead and create our own custom model which we call phone processing so i click here again similar experience as i see a description of what the model is able to do i have a link to a microsoft learning module and i see an example here to get an idea of what the model is able to do let’s go ahead and click get started and now we’re going to enter into what we call a user experience which is going to guide me step by step in building uh User Experience that model for for me and with everything with the poor platform this is no code so uh we won’t have to write a single line of code we won’t have to do a any complex data science operation it’s going to be pretty straightforward so the first step is to define what data i want to i want to extract um and today we support we were discussing before we support the fields we support checkboxes we support tables and now uh as experimental we also support tables that span across multiple multiple pages in this case we have the three we have fields tables and checkboxes so let me go ahead and define those so for fields we want to extract of course like the sales order number we want to extract as well uh the vendor name also we need the amount of the of the search order and for checkbox there’s a checkbox that indicates whether that shipment was done in a priority way or not so let’s call it uh priority shipping and then there’s also a table so we want to extract the information so i’m going to click define a table and the table experience we first need to give a name so let’s call it an online items and by the way these names don’t need to correspond to what’s reading on the document these names are just for you to be able to reference them but they don’t need to match what’s written on the on the document itself and then what are the columns we want to extract uh perhaps you only need one column or two columns so it’s okay we don’t need to define all the columns that i want to extract in this case to show you um what we’re able to do i’m going to define all the different columns that we have in the document so one of them is quantity let me write that correctly the other one is unit the third one is the description of the item we also have a part number here we go and then we have the unit price and the total price okay so here uh first step of this wizard is to define the data i want to to extract joe is there a limit on the number of columns you can add is there a limit that is a very good question uh i there must be a limit uh but it’s like very large it’s like i don’t know maybe like 200 columns something like that oh okay yeah i was like 50 that’s all i need yeah you know we haven’t run into any any customer usually but uh it is it it is very generous yes yeah that’s a good i i i need to refresh what the exact limit is on the one but it’s very very large people people love excel spreadsheets so yes i agree yes Collections so let’s move to the second step so here we need to provide the model samples of our documents so the model can learn about those documents and learn how to extract the data and we have this concept of collection so what collection allows you to do is if you have documents with multiple layouts a single model is going to be able to to handle those different those different layouts uh and so when you upload sample documents in order for a model to learn about different layouts you need to upload for each layout five sample documents uh in a collection so in this case since we have two layouts so one here for atom indus industries and the other one from contoso i’m gonna go ahead and create uh two uh collections so let’s start with the first collection let’s add documents so as i was saying we just need five documents to get started and i can upload them either from my local device from a sharepoint site of an azure blob storage account in this case i have it on my my pc so let me go upload here five sample documents and the number of the usually five if if the if the quality of documents are high so in this case i would say that even though they’re scanned they’re pretty high so i can and read pretty well what what what they are like the the resolution is pretty high uh five is is is enough if if on the contrary like documents like the scan quality is very bad there’s different angles it’s like they’re taking with photos in that case we recommend doing more documents so even up to 10 or 20 will will improve the the performance so to upload you had two options you had a sharepoint document library and then you had azure blob storage was that the one correct or your pc yes now even though it says sharepoint can we upload it from our onedrive because i can get the onedrive url is that is that possible it’s it’s limited to sharepoint uh even though i mean adding one drive wouldn’t be too complex indeed because in the infrastructure it’s not very different behind the scenes but uh but in this case it is specific just for sure got it which makes sense yeah uh and then so i uploaded the five ones from for contosos let me add those five other ones for that too and joe just to clarify this is just to upload and train the model it has nothing to do with syncing to sharepoint or onedrive this is just to upload the files just to train the model that is correct that’s great this is just to to to get some sample documents it doesn’t push the data back to sharepoint or azure okay something that i can do as well is give here a friendly name uh it’s like so once when i use this model in a flow for example um i can get the collection name so when i press a document i can okay for which collection it corresponds to so i can get the collection name if i want to do some type of classification or for some reason i need to know which which provider was was processed okay so on the first step we’ve defined what we want to do on the second step we’ve added sample documents and now there’s just one step left which is go ahead and actually tell the model about about our documents our fields so here we see the documents that we uploaded and on the right the different information want to extract so we’re going to go ahead and tag in the document the different items so it’s very simple i just with my mouse click and click and drag what i want to do so here i want to extract the sales order number the vendor is here the tutorial i’m on for example let’s say that i make a mistake uh i put this amount i can very easily fix it so as you see it is totally uh local it’s very simple to to use checkbox is very similar i just draw a rectangle and i say this is a checkbox and a build area text this is a check or selected checkbox and then for tables i just draw an area around the table and it’s gonna i’m gonna be able to draw uh the separations between rows and columns to define the the table and this is something that we launched very recently i would say like four months or so ago uh so for some of you this might be this may be new and now what i do is i basically assign uh each column so this is the this is actually this one i didn’t define so i don’t need to uh i need to assign it this one is the part number this one is a description uh this one is the i i didn’t define it so i can skip it quantity the unit the unit price and total price i’m good i’ve defined i can even do this so i can preview the table to see how the table is going to be extracted everything is looking looking good and once i’m done with the first document i’ll do the same process with with the remaining document let me show you just also uh for the second collection for the electron collection which has a little more complex table so for the fields it’s still the same as i just need to uh draw uh around each of the fields so the sales order the vendor as you can see this is an image but we were able to text the text within the animation while ago uh total amount i can add it here and then uh the change with this table is as you can see i need to extract the description i need to extract the the part number and they’re in the same column so let’s see how we can how we can handle that so again i’m going to draw the area around the table and in this case i can indeed go and use this grid mode to tag the different the different items so this one is possible the unit price is here the unit is here the total price is here but then the challenge that description a part number and the same column so what we can do is we have this advanced mode where you can basically tag exactly how you want the table to be so i’m gonna go and say the description is just this part here and the part number is just this one here so that way i have the flexibility to even if i have a complex table which is not doesn’t fit into a grid or has nested items i can teach the more that hey uh this is how i want the table to be to be extracted you’re getting a lot of positive feedback by that you know all the clicking and selecting that you’re doing i mean that the ui it cannot get any better than that man thank you yeah that there’s amazing feedback coming on that so there’s a question i have though is there any preference on which browser you should use all the popular ones of course we prefer if you use edge but but we support chrome we support firefox safari so i mean all the most popular ones are supported i think we’ve we’ve finally dropped support for internet explorer so that one we don’t support anymore but it was it was long due but uh but yeah the popular ones the big ones that should work pretty seamlessly and then someone left at the table i can also go ahead and see a preview and i see that this is how it’s going to be extracted the description is going to be extracted as a column and the part number is going to be extracted as a separate column and then i would just go and uh and i discussed about something last to mention is that for example here priority shipping is a checkbox in this case for the alatome industries there’s no checkbox for priority shipping so i can just go ahead and say hey for this collection uh just don’t look for that uh for that field you won’t find it so in that case i don’t know how to target so now i just go and tag the remaining of the documents since since we want to leave time for q a i’m going to go quick and i retrain the model so i have the same model with the same documents so once i take documents i train the model and i land it to this page and here what i can do is i can do a quick test so i can basically take a document which i didn’t use for training and validate that is able to extract everything so this is this is the value to validate the quality of the model if i get everything with a high confidence course so here i see i’m getting like the the sales order number let’s see the table like the table everything is extracted correctly the total amount um if everything is looking fine i can go ahead and and use the model um if uh if not i could i can always click on edit i’ll go back to the training wizard and it’s gonna guide me to uh to to to to i get to to which are where i can add more more documents and improve improve To the model and then to use it it’s very simple i just click use model i can use it either in flows i can use it in an app important in poor apps i can also use it in our document automation story kit that i was talking about in the in the presentation so if i click in new flows uh i’m here presented with some popular cases like hey my documents are coming by email my documents are coming by by sharepoint so in this case like i click too fast but in this case since all use cases uh documents are added into sharepoint i just click on on the sharepoint one and it’s gonna take me to um to a template wow that already does that for me uh so i just have to specify what is the the the sharepoint folder and then uh the the air builder action is already configured with the model and the sharepoint parameters and then similar to the to the to the invoice one we send an email but here i can customize it to uh so in our case we want to add it to excel and push it to a legacy website so i know uh it’s being i know we’re reaching out at the end so uh maybe what i can do is i can finish uh the couple of slides that i had and and then we go to q a and if if we have time we can i can go back and finish the demo uh but that in the demo is very similar to what we saw with invoice uh i would just add a file to sharepoint it’s going to run this automation and it’s going to add the data into excel and then to push the data into the legacy website we use a desktop flow but we can see that later if we have time New Logo so uh i did said the beginning that that’s some surprises before we finish uh the the before we go to q a uh so one one more thing uh so for those of you who have been the beginning like like razer with us uh you probably know our logo so he’s being asked with us between the since the beginning since uh two years ago uh super excited that to unveil exclusively in this show in the r d show so it’s the first time we show it publicly starting from now from today we have a new logo so this is our new ebitda oh that looks nice so you’ll start seeing it in in the product very very very soon uh all the big applications poor house proto-made we’ve refreshed the local recently a nice builder um i like that that’s ai and the brain working together okay there you go there you go that’s uh that’s uh that’s the thing and then a second very quick uh second surprise Document Processing word exclusive that we’re doing uh today is uh so we’ve seen the um the phone pressing model um and in one month from now we’re gonna change the name of form processing to document processing and and why is that so if we go here uh so this is an internal version that i’m showing here a test version uh so up until now phone processing supported what we call structure or semi-structured documents so things like invoices tax form well there’s certain structures so there’s key values there’s tables in in in a certain location but what we’re seeing more and more customers trying to do is hey i have things like like contracts i have seminar works i have like documents with a paragraph of text and i want to extract uh information like that is within the the paragraph or i need to extract like a whole paragraph um we’re adding support of that with unstructured documents and the way it’s going to work is you’re going to go ahead and create a document processing model as you do to the reform pressing but we’re adding a new a new step so the first step instead of asking you the fields you want to extract we’re going to ask you whether your documents are structured or semi-structured so this is what we support today and now new so in a month from now we’re going to add support for uh foreign structure documents so first time we show it as well and we’ll soon start a preview on that so um i’ll share some details of how you can join the preview if you’re interested if you have cases like i want to extract data from contracts so so on uh this is now supported and and yeah uh very i’m just going to Highlights take one minute and then we jump right into q a um just some highlights sometimes i’ve rolled up so our structure is a big one uh that i’ll just show feedback loop so this is the ability of as you have a model uh and you have the model running the ability to provide feedback to a model to improve it over time so that’s something also that that we’re releasing in in a bit over over a month we’re also adding in the in the last page that we saw where we can test the model we’re adding some telemetry some some uh scores around here how performance your model is uh how you can improve it so i’m going to improve that page to to give you recommendations so how you can improve the model collaborate model authoring so this is today there can only be one owner of a model we’re going to give the possibility to have multiple owners if you have multiple people at the company who want to collaborate on the model we’re going to enable that multi-page so this is something that that indeed as we saw so in this show it is top of mind for many many users so we’re working actively in improving uh extracting tables that span across multiple pages and then all audiences global i’m sure folks training today are coming from different countries so adding support for many more languages as well in different preview models that that we have and before we go to q a so if you want Free Trial to try it uh as we said at the beginning it’s free there’s a free trial so if you go to aka then slash builder you can try what we saw today for free and if you would like to to join some of our preview programs so we’re going to do our structural documents in a couple of weeks as well as feedback loop just go to akdms ms air builder insiders there’s a form we’re going to ask you okay like what you want what do you want to try and you name an email i will contact you back to to give you access to those to those previews and with that i think uh sorry for taking a little bit more time that that that we agreed to raise and danielle but let’s let’s maybe jump into q a oh joe that was that was awesome and i Road Map think uh the road map and what’s coming is extremely interesting and i love that you’re renaming things you’re making it more user-friendly you know that thing around the form processing model i want to post it or i want to leverage it and flow here are your options sharepoint email that’s exactly what folks are looking out for and i’m specifically talking about citizen developers so i really see where this is going and uh you know i think your team is doing an amazing job and uh yeah something that uh i would definitely uh try out in a lot of scenarios and use cases so thank you cool dan any comments before we head to q a Comments no i mean um i’m super stoked about that new um you know logo it made perfect sense to me um and i did like the uh you know some other walkthrough that you did um you know what always blows my mind is that um you know ten years five years ago it was in preview type of a thing but 10 years ago you you can’t do this type of ai work unless and until you were a mad scientist phd you know you know in a university doing that kind of stuff but here in literally when you build that flow you did the training you build the floor and then you got the data that was in five minutes less than five minutes i mean if you didn’t talk it was two minutes so i mean we are at that you know a point in technology where a.i and and just us you know simple people can actually build that that’s that if you said i mean stop and think about it it’s mind-blowing it really is so thanks thanks for demoing that joe my pleasure and a very very well said daniel did like uh this uh even a couple of years ago i would have required like a very complex multi-week multi-month project to to put into and now uh basically in one day a couple of days uh you can be that very sophisticated solution yeah yeah yeah a lot of comments coming in joe a lot of uh compliments on your presentation and Questions how simple it was the simplicity was appreciated a lot so uh a lot of feedback coming in there uh i’ll head over to questions now so there were a lot of folks who asked questions during the session we will take their questions first uh if you have additional questions this is your chance posted in the chat and dan and i will try and populate the screen as many as we can so uh this question is actually quite interesting because i think i have this question as well when i’m leveraging the form processing model what kind of documents can i provide to it does it only have to be pdf or could it be any other office based file is the question pdf and images so today we support pdf we support png’s jpegs and t-files um we do hear office files uh so we are we are something that we’re thinking of adding in in in an upcoming release but today is pdf and images it’s only pete and just to help you out i did this on a project if you’re getting any other format there is a there’s an action in one drive convert to pdf it’s a free action you can use that to convert various types of files into pdf and then you can give it to the model so uh try that out i do that for i did that for for a customer so it works and it’s free okay uh next question is uh AI Models an interesting one again it’s from roshan kumar does these ai models and i think you touched on this joe yeah does the ai models have the learning capabilities as of now whether it can self learn based on the confidence it’s gaining based on uh data points that are being provided so that’s a very good question so today the way it learns is through the training wizard that we went through so you upload sample documents you tag the documents and that is the way that the motor learns so the more documents you provide the better you can get to learning but there is something that that the maker so the creator of the model needs to do needs to go back to the model and add more documents with this upcoming feedback loop uh feature that we’re releasing what we’re enabling is what what the question is saying like as you use the model and as you get low confidence scores and you provide feedback to the model we’re going to be able to improve over time for you so the model will learn over time and improve and improve automatically so yeah that is something that that is coming very soon now nice Starter Kit all right the next question is you mentioned about the starter kit when you have that entire process predefined for you so uh is there a link to the starter kit solution uh i noticed one of the templates uh that you popped up when you went to when once you created your model was the starter kit so i think that’s one entry point but is there a direct link to the the starter kit solution yeah so the easiest way to get your starter kit is if you log into portomate or you go to aka.mslstria builder and there’s the airbuilder menu and you’ll see three options you see explorer you see models and then you see document automation and so document automation is a starter kit so it’s within port automate within the navigation menu you’ll see the document automation link and that is that is the the starter kit perfect here’s the next one um roshan kumar again asked a question he says does or can models be retained by users based on actions validations and suppose if it didn’t detect a few fields um and you know user goes ahead and makes any corrections can that model basically learn from it for the future executions um and i’ll just add a little bit more to roshan’s question over there is that how much of that is us going back and manually retraining and if there is any built-in where through ai it will automatically learn you know based on its experiences so that’s russian and i just added that yeah i think i think i hope i hope russian you sign up for the pride preview feedback because this seems like somebody who can provide like great great feedback to us uh so to what evaluation will enable is uh if you if we process a document with a low confidence core you can give feedback to the documents and say hey the feedback this document was extracted with a low confidence core and this is how you should have fixed it and we’re going to take that feedback and we’re going to retrain the model for you automatically and improve and improve over over them i think i think what what the question is thing is even a little bit more advanced like the model being self somewhere and being able to autocorrect and there is something indeed that like like we love together but today is the way we work at least for the first iteration is that those documents that are flagged as uh as low confidence anybody can go ahead and review them and provide feedback and that feedback will be incorporated to improve the model so that they still need will need to be a a human to to review and validate that so it won’t be fully automatic they are taking control of of the model perfect uh the next question well joe we predicted this at the start uh and there were a few questions around this initially as well is licensing so uh here’s a question from i like the name explaining power platform all right there are a lot of students who cannot access m365 you know they need that free trial yes there is a free trial we know that there is a free trial for power platform of ai builder you can go and try that out uh but in general i brought this question up because i wanted to ask about licensing oh many people have asked that question is it included in premium licensing uh is this a separate license if you can just give a 30 second overview of what’s required uh to use ai builder and an enterprise grid level definitely right now that’s a super good question so to use a builder so indeed you can try for you can serve for free with the trial but once you go into a production or like a a real usage you will need a license and what you need is air builder credits so you need to purchase so every action you take in a builder consumes what we call air builder crates and there’s a number of ways to purchase a bill upgrades either you purchase the air builder addon which is going to give you 1 million air biller crates per month or now with any poor apps premium license so a per app or per user license you’re going to get some credits as well so if you go to the to those licenses you will see the the per user has like 250 credits the per app has 500 billion credits so that’s that’s one way also to get credits now um at the lower at lower cost than the than the add-on and also importantly the rpa licenses and the unattended item as well include some some apple uh builder credits so you need to have your grades you can either purchase the add-on or you can purchase one of the premium licenses for pour apps and proteomics that will give you some number of crates and if you search you also find that we have an air blog calculator to estimate how many how many add-ons you need to purchase um so yeah that’s uh that’s a quick uh the quick answer gotcha and joe this is all at the tenant level just to clarify this this is not a user level credit it’s a tenant level credit that when a user starts leveraging ai builder it goes against your tenant level credit so yeah is that is that correct yeah no that’s that’s i think i find that yeah i did yes it creates the you purchase for the for the whole tenants of the whole company and then the administrator can decide okay i want to assign to like to this environment only or to other but it’s not that per usually it’s like for the whole company purchase credit yes and then you assign it to an environment is that the least level i can go to or can i go to a user level for example dan and raza can use ai builder in my organization because they know what they’re doing as an example yeah environment is is is is the table but you could do in that case if you could do like a we would put an environment with just you and you too so daniel and reza uh but yeah it’s at the environmental level okay that’s important uh one last question can you see any more i see one lost i’ll just pop this up yeah that was the one i liked that it was it’s a good general question which i liked yeah so the question is uh joe can you talk about similarities and differences between ai builder and azure cognitive services very good question very good question so it’s it’s ai builder is built on top of azure quality services uh so you can expect the same quality the same results the same performance the same scalability or all that goodness the difference here is where you are more comfortable like if you’re you’re more comfortable the poor platform uh you don’t want to create azure resources you don’t want to call api’s air builders the way so you you we’ve seen the demo like it’s very simple interface to video models to put in a flow to put in an application so if if if if you’re a more assistant developer or you’re a professional developer but hey you want to quickly build a project and uh or a solution with a poor platform a builder will be the the way to go if instead you’re a professional developer like you you spend your day in azure you you’re comfortable creating subscriptions and api keys and and calling apis and you build your own custom code solution then quantity services will probably make more sense but it’s going to give you that api flexibility to to call the models uh so it’s a matter of of where you’re going to work if it’s an upper platform air builder is the dis solution if you’re a developer and you’re going to work with code in visual studio or any other any other coding environment then cognitive connections connected services probably makes more sense so pro cord was that’s that’s the the general consensus around this and uh i know the answer joe but for the common audience because i had asked you this in privately before is how long before a new feature gets added and azure cognitive services does that feature flow into ai builder because it has to be in azure first right for it to flow through into ai builder that that’s correct that’s correct we we work so the air builder team and as we were from if uh client resources team we work very closely uh so so even even before it’s public in azure we work with that team to make sure that we’re aligned it it depends on the on the capabilities for example we see a lot of demand that run documents so everything around the document we try to be very aligned with azure so when when something gets for documents it’s called azure from recognizer so when something gets out of phone recognizer we try to have it uh basically at the same time in a builder for other quality services where we see less demand we take a little bit more time uh but yeah it’s it’s uh the goal is to be as much a line as possible but you will see that in some areas with pre-aligned in some other areas we might take a couple of months of delay to to have that future parity between the meeting both perfect all right i think that data wraps up uh all the questions we had that’s a lot of questions um before i close first thing uh i want to call out is a big big uh thank you to joe oh really thank you for taking the time out for joining the call for for being on the show uh it really helps when someone from microsoft someone who knows the product in and out comes here and talks passionately about the product that they’re spending a lot of time on and you give us some great insight as to what this product does what’s the future uh you give us real world scenarios so i think for me this was uh this was a great learning experience so uh thank you joe thanks thanks for joining and on behalf of the r d show you know really appreciate you being here yeah i i echo that so nothing else to add recording later i really really appreciate taking the time to learn about the builder i think a big thank you to you guys to daniel reza i mean uh you guys make i mean one of the special things of the poor platform is the community uh you are a big part of this community so thanks for everything you do and for this spiritual and um and yeah i hope i had a lot of fun so i hope uh as we keep evolving a bit i hope to be here back in the future episode with more more updates i’d love to i love to do that the show is yours the show is open just tell me and we’ll bring you on awesome and of course before i close out let’s not forget the audience thank you thank you everyone we do these shows because of you all uh you joined these shows and that’s the reason why we keep we keep pushing the envelope and uh uh we had joe this time uh next time we’re gonna have something around microsoft teams and power platforms so we have we have a lot of interesting content lined up we’re trying to mix it up this time in this season and i hope you guys are enjoying it so spread the word and we’ll see you in our next session which would be on april 8th uh follow dan follow me on social we will promote the next event as well and once again thank you everyone for joining the show and thank you joe thank you thank you be off bye everyone all right bye