Speaker 1: What's up everybody? Welcome back. John Levesque here, your host as always. Today, I have a very exciting show planned. Today, I'm going to show you a couple of awesome guests who are going to bring some new knowledge to teach us about the brand new document automation inside of Power Automate. Who are my guests for today? Let me introduce them first, Geetha Sivasilam and Daniel Laskowitz. How are you doing, my friends?

Speaker 2: Doing great.

Speaker 1: Good. Geetha, why don't you go ahead first, take a second, and introduce yourself to everybody so that they can find out who you are, and then we'll let Laskowitz go last.

Speaker 3: Sure. Thanks, John, for having us. I'm Geetha Sivasilam, and I am a Power Addict, and an enthusiast of all things Office 365 and Azure. I'm also a Microsoft MVP for the business applications domain, and I work for Artist Consulting, and part of their collaboration and app dev practice, and I'm based out of Dallas, Texas.

Speaker 1: Awesome. How are things going down there with all the crazy weather? You're doing all right?

Speaker 3: Yes. So far, so good. It's really, really good to see the sun out, and I just can't wait for some Texas heat already.

Speaker 1: Good. Good to hear. Do you still have power?

Speaker 3: I do, I do.

Speaker 1: That's a great question. She's online on the internet. Well, you're in the power of

Speaker 3: North Texas, this is how cool is I am.

Speaker 1: That's awesome. All right, Laskowitz, go ahead and introduce yourself, my friend.

Speaker 4: My name is Daniel Laskowitz, and I am the Power Platform Lead at SoCity in the Netherlands. I'm also a Microsoft MVP in the business apps domain. I also host the Flow Pro Show, which is a YouTube show together with Sandy Yusha. We just had one yesterday, so if you haven't watched it yet, make sure to check it out. I'm also the co-founder of the Dutch Power Apps and Power Automate user group, but we changed our name and we are now Power Addicts NL. We're trying to get the Power Addicts name as well. Yeah, that fits a lot better for us now. Very cool. We actually didn't rename to Power Automate because we had Dutch Power Apps and Flow user group. Yeah, we didn't really change it. Papa doesn't sound good, right? Papa, that's awesome. Yeah, that's about it. Cool. Let's move into the Document Automation. Document Automation Toolkit has been released by Microsoft a couple of weeks back. It's a solution which you can install in your environment and it exists out of multiple things. You have multiple flows in Power Automate, you have a Power App, you have an AI Builder model. Also, the Power Automate flows will, for instance, look at a service mailbox, so that whenever a supplier or a customer will send you e-mails with documents, it can actually start a process from there. Whenever something arrives in the service mailbox, it will get picked up by Power Automate. After that, it will store the documents, it will go into the AI Builder to process the documents, and at the end, there will be also the review of the process documents in Power Apps. We're going to show this all in detail in a couple of minutes, so bear with me. The installation is really easy. You can just go to a link which we will share at the end of this session and you can just select the Select button in here, and then you can install the solution really easily with a wizard that you have, and then you can step-by-step actually configure that.

Speaker 3: So it's really easy to use. Sorry? I said it actually does a pretty good job of finding out resources that you can click on and it gives you context around step-by-step instructions on how to use AI Builder, or just if you're completely new to some of the components of the solution, then it does a pretty good job, this whole walk-through interface.

Speaker 4: Yeah, I was really surprised by that because it's really easy to use, and I think those kinds of things are really powerful because if you look at the whole solution, it has multiple parts. Of course, it has AI Builder, Power Apps, Power Automate, and I already talked to Justin yesterday from the Power Automate team about how it's less becoming a Power Automate workflow or something, but it's more going to be about the value that it brings. For instance, with the approval app that you also have in Teams, you don't even see that it's Power Automate. In this case, you will see it's Power Automate, of course, but it's more of a package together, a real solution, and I think it's really interesting to see Microsoft move in that direction. That's really something that's really interesting.

Speaker 1: I think my favorite part of that statement is Lasko is saying, he's so surprised that the Power Automate team put out something easy. No.

Speaker 5: I'm impressed.

Speaker 6: Come on. Sorry, guys, I didn't mean it like that. Thanks, John. I'll hand it over to Gita to do a demo because otherwise I'm going to say some wrong things. Gita, take it away.

Speaker 3: What we're going to do here is going to walk through a demo just to show you what happens from the time you get an e-mail with the customer invoice, for example. Then you have Power Automate based on the visual that Daniel just showed you that does all of the orchestration behind the scenes. It really is the backbone of the solution. It picks up your e-mail, it extracts your attachment, pushes it back to Dataverse, uploads the attachment, and then invokes your AI Builder model that you have to configure as part of your installation to extract all the metadata that you really want, and then it processes it. Then what it does is stores all of that extracted metadata in Dataverse as well. Then you have a Power App that you can use to track all of your documents. It's listed by status, and then you can choose. The Power App, really cool thing about it is, let's say you have a new model, it gives you the interface within the app to go configure the model that you want to apply for the documents. Then you can also choose to go in and make updates. Let's say you don't like how the prediction came out to be, or maybe there's a missing field. You have the ability to go in and make changes to it, and then validate the documents. You can totally have a gating mechanism for you based on business rules that you would need for you to consider your document as validated to be considered a completely processed document. But we'll take a look at how all of that works. Now, what we're seeing here is a sample e-mail that I just kicked off, just to save us some time. I have an NYPDF that was sent to me this morning. We're going to let Power Automate do all of the fun orchestration behind the scenes and do all its magic to extract the metadata and process it, and then push it out to the app. Now, what I'm going to do is going to show you the end state. That's how the user experience is going to be. Once they receive the e-mail, they're going to go to the app. What I'm going to do is I'm going to refresh this. If you notice in the app, I have a bunch of documents that I uploaded yesterday and you can see the different statuses. Some are validated and this one tells me that it's waiting for a manual review in the sense of a processor or an approver really needs to take a look at it and kick it off to a completed state. Then here's one that's an invalid document, which means I have the ability to have business rules within Power Automate to identify what type of documents I should really be processing. In this case, I have a text document as an invalid extension type that I don't want to be processing. It actually marks that as an invalid document. Then now here, there's also another concept of role-based processing or roles or personas that you can really have within your solution. There's one which is like the process owner. You can do everything in the sense you can come look at documents, you can configure what AI builder model that you want to use. Then this other role, which is like your validator, your document approver who can't really do any configuration, but they can come to this app, pick up a document, own it, and then figure out if they want to pass it through the validation state, mark it completed, or if they want to append any metadata or make any changes. Before we go on to the configuration step, what I'll do is I'll open up this one. For example, this is the one that we just processed. That shows up and it's waiting in my queue. The moment I pick this document, I become the owner on it, which also means that nobody else can actually go in and make a change to it, which is pretty cool. All of that is already built in for you. Here is the sample invoice. Then you can see that the model that I use to train these documents or the process to pick up and extract metadata has done its job really well. It's picked out the invoice number, the due date, the total amount, and my customer ID. Now, I do have the ability to go in and maybe make a few changes. Let's say I feel like, oh, you know what? This number doesn't look right, or maybe it should be something else, or I want to add append to what's already been predicted. I could do that if needed and then actually mark validate. But then before we do that, let's go to a table section too. With AI Builder, you have the ability to mark tables. Right now, it picked up all the necessary metadata that you really want to be storing as part of this process. What I'm going to do is just go ahead and click validate and then you'll notice that it will eventually change the status to be validated. Now, this is my list of documents that I can maintain and track. Now, let's see. Now, I'm logged in as the process owner. You have security roles within Dataverse that you can assign to your users. One thing to note is that I think this process is so flexible enough, the solution lets you either choose to have role-based persona use cases that you would want to configure, or you could choose to not have one in the sense you could make sure that everybody is a process owner and everybody can do everything, or everybody has the role to have the gating mechanism to approve or validate documents. Let's go look at what this configuration tab looks like. Here, I have default configuration that I want to be assigned to this process of documents that are coming in. Then if I actually go in and look at that, I right now just have one model. This is the model that I used to upload a bunch of sample invoices and then train the model to identify the metadata and extract the data as it needs to be. This is the one that I had. Now, if I have a change to it, I can always reload the model and I can click on the model and look at the actual field mapping. You can see here, these are the fields that I actually have within AI Builder. That's a step that you would go through as part of your doc installation kit that Daryl just talked about. But this shows me these other ones. Let's say I want to change something, I can always go and change it, come back, hit refresh, and then that will make that association or refresh that association to your process as part of the kit. This is just all about the app. Now we've walked through this demo of how you send an e-mail, how to automate all of the magic behind the scenes, and then now we have this app that you can use to track and maintain and process documents. Now let's go take a look at what magic that Power Automate is really doing behind the scenes. What does that orchestration look like? If you actually remember the little visual that Daniel was showing us a couple of minutes back, it listed a solution that comes with the Space Kit. Part of the solution, you may want to notice actually about four artifacts that are really important. The first one is the app, which is the Canvas app that we just looked at. Then at the bottom here are three flows. That's actually doing all of that heavy lifting for you, and it's all packaged for you and already built out for you. This first one, which is the e-mail importer, this is a flow that will process documents as they're sent in as an e-mail. Now you have the ability to go in and update that e-mail to maybe have some business rules around only picking up attachments with certain subject line or maybe coming from a certain e-mail address. You have to be able to do that. Right now you have a placeholder and it's just picking up any e-mail that comes in with an attachment. Then what this flow really does is it actually creates a processing request in Dataverse. Then whenever a new processing request is created in Dataverse, you have the second flow that gets kicked off, which is the document automation processor. What this does is actually trigger the AI Builder model that you've defined in the app or associated to your process. Then that'll go through the prediction process and then it'll extract the data and then store it for you in Dataverse tables. Then the very last one over here, which is the document automation validator, that's more of a placeholder in case you want to have a gating mechanism to include any business rules that you would want to put in place to define how to validate a document, or maybe you don't want some documents to go through a manual review. Some documents may be based on your business rules that you depend on logic and branching. You could choose to change the status on this as validated. Maybe there is some need to be manually reviewed. Right now this is a very simple placeholder and you have a flow basically anything. By default, any document that comes through this process, it puts it in a manual review mode. But you have the ability to go and change it and add any business rules that you would want, and branch it out based on how you want to divide and rule this whole document processing within your organization. That's about how this whole process works. It's really amazing how this toolkit has been packaged in a way that it gives you the flexibility to extend and work based on how your use cases are within your enterprise. But then you have all of your work done already. It's really a fully working solution that you have, but Power Automate doing all of the backbone work for you, and then Dataverse being your backend. What that really means is that you can scale the solution. You have a really good advantage of scaling it to a level that you would want to and it can work at an enterprise level solution. Then you also have the ability to use any personas. Security is pretty good with how Dataverse allows you to have granular level of security. This is a pretty cool solution. I've been pretty impressed with it.

Speaker 1: Yeah. I love how modular and customizable it is. They've done a really good job, even just thinking about how they do the flows. One flow creating one thing where it's going to check off the processor, the processor fires up and hands it off, and just the way that they're thinking about this solution is really nice. I'm so surprised that they put out something so easy.

Speaker 5: Thanks, John. That's my cue, I'm out of here.

Speaker 3: Yeah, I think it's pretty cool. There's lots of ways to extend it. You can look into integration within Teams, maybe embed your app within Teams, and you're not limited to things. This is just a very base template for you, just to show you how you can automate content or document processing at this point. It's pretty generic. It can be invoices, it can be purchase orders. It's generic to what you tell your model to extract and based on the type of documents that you're training the model with. But then I'm looking forward to the other two solutions as well, which are more geared towards receipts, and the other one, I don't remember it. What was the other? You had a little slide.

Speaker 4: We'll come to that in a minute. I forgot about it. Luckily, I have a slide about that. I really love this whole thing where you have a whole platform solution instead of just one thing, because that's what used to be, where you have a template in Power Automate, for instance, and it's really about Power Automate, or you have also a template in Power Apps, for instance, which you can use, and which is really about Power Apps. Now, it's all integrated, and it's like more of a Power Platform thing. It would be really interesting to see as well if Power BI will be included in those kinds of things. But if you can have that whole end-to-end scenario inside of the Power Platform, that would be really, really interesting.

Speaker 3: Yeah, I can totally see that happening. Having a dashboard, more like a holistic view of actionable insights that you would want, maybe showing you insights over the number of documents that are being processed, just metrics around business process improvements.

Speaker 1: Even that main screen in the app could become a report that showed how many need validation, how many have been processed successfully, what types of attachments came through, who did what, what time it took. There's a million insights you could gain there.

Speaker 3: With all the gap analysis and things like that, yeah.

Speaker 1: Yeah. Well, and one thing I want to call out too, I did notice when you were showing how to build that model, it looks like you build the initial model by importing one of the documents and working right off of that. That's, I think, one of the key parts of this that really makes it easy for people to step into, where they can just say, oh, this field, this field, this field, this field, and are able to build a model without having to do any code or anything like that.

Speaker 3: Yeah. The GUI for that is really intuitive. It requires you to have about five documents, and you can also create multiple collections within the same model, and then about five documents, and then you just train it by marking and tagging the metadata that you would want, and it does a pretty good job of walking you through it step-by-step.

Speaker 1: Beautiful. Very cool. I think Laskowitz is going to give us a couple of places to go and get more information.

Speaker 4: Yeah. There are some resources that we wanted to share with you, so we will put them in the video link. We will have the Document Automation Toolkit, Microsoft Docs page, where you can see all the details about the solution. There's also a Microsoft Learn module where you can see all about it, so that's really cool. The other link is the direct link to the Document Automation Toolkit where you can directly go to the AI Builder part in the Power Automate website where you can easily select install and go ahead from there. We will share those links in the description, so they will be available there. What's next? This was the first solution. As Gita already said, the receipts is one of them and also invoices. This right now is the custom documents one, so you can train your own model. They will have some specially for invoices and specially for receipts as well. Those will come soon as I might imagine because the button says coming soon, so I hope so. Let's see what's going to happen there. Maybe they will also include some other templates in there, not just document automation, but other things. I would love to see some more really easily deployable solutions in the Power Platform. I think this is a great example of how easy it can be. Yeah. With that, I think it's a really nice solution, and I hope there's more to come.

Speaker 1: Yeah. I think it's interesting that the receipt one will be fun. I've built flows before to manage receipts, and using custom vision, and building a model, and doing all this stuff is pretty wild. To have it all just happen in a few clicks where I can walk through a wizard experience, I think that's going to open up a lot of scenarios for people.

Speaker 3: I think this whole process, like starting from the install until you have it working, takes less than an hour. Less than an hour, you have a solution that can scale, and just work with your trained model, which is pretty amazing.

Speaker 1: That's amazing. I think it would have been fast to build the flows in comparison to having to deal with that process, but then now how they've done it in a solution package takes it even to that next level, where Laskowitz could probably build this thing, but it would take him a week. But now to be able to deploy it in an hour, like, man, it's epic, it's super cool. I'm so surprised. I had to get it in one last time. Awesome. All right. Well, let me just say thank you so much to both of you for coming and hanging out on the show and showing everybody these new features. Really appreciate you being here.

Speaker 3: Thanks for having us. Thanks for having us. We had fun.

Speaker 1: And you guys at home, go ahead and leave some comments and questions for Geetha and Laskowitz down in the comments. Go ahead and check the description for all those links so that you can go ahead and get started. Click that like button and get subscribed so that you don't miss another video. Much love from us. We'll see you in the next one.

Speaker 4: Bye. Bye. Bye.