Simplify Sales order approval in IT architecture documentation

How to create outlook signature

Hello, my name is John Evans, I'm the Director of Solutions Architecture here at Impact Networking and today we're gonna be talking about sales order processing automation using business process management and document extraction software. I'm gonna take you through a current implementation that we have to process orders for this particular demo. I've already sent some documents into the system. We take a business process and we connect to either a watched folder or an incoming email address. Then as documents make their way into into the organization, they're automatically sent into the business process. So in this instance my process is looking for sales orders. Specifically the ingestion section has already happened, I sent those documents in, then they hit the transformation service and the transformation service is trying to take that document and turn it into a qualified set of data. It does that first by a classification process. It identifies what kind of document it is. So, once it arrives at a document type, it's able to then send that document to a specific extraction model and it extracts the data off of those documents in a meaningful way. Any time it hits a snag in either of those processes on the transformation side, it's then able to send that out to a user for what we call remediation or validation. So, we'll pick this process up here at the validation step. The document here for this McMasters order has come in from an email address. It's made its way into the system. It's gone through the ingestion and also part of the transformation beats but we have to validate some data off of there, so quickly taking a job from that queue, I'll see that we land on this order validation page. The order validation page is made up of a couple of different pretty, pretty obvious components that the left-hand side is my extracted data and the right-hand side actually shows me the document that was sent into the process. Then each one of these fields here has a bit of data in it that's been pulled off the document that's intended for us to validate. If we look at this PO number, I'm able to mouse over here and it will show me where I've extracted that data from on the document on the right-hand side. So, as I go through these fields will see that they look pretty clear, right? If there was an issue here, I am actually able to pretty easily go over to the document And just by selecting the text, it will insert itself into that text box. Built into the back-end of this system is something called a machine learning model. The machine learning model is going to take any deviations from the first installation that I have, any spots where I've told the system to do one thing but the document that comes in is indicating to do a different thing. It will take the user's input, the user's changes to that model, it will inherit them into the machine learning engine and then ultimately update what we call the knowledge base. So, over time as as users adjust the extraction validation thresholds and they build more confident data sets. The system learns all of those things as it goes through. The next thing that will happen is now that I have a qualified data set off of the documents. It's going to take that data and move it downstream into what we call a review process. And this is essentially a workflow that could be customized to any steps inside your process that would be necessary. Well, let's take one second to talk about this page that I'm on right here, which is basically a work queue page. So, if my job was a customer service representative or I was in order processing or data entry clerk, I might open up this page every morning, find my list of orders that were in there and given some cues from the system; PO numbers, customers, dates that things have happened, different events have occurred, I'm able to decide which of these orders is important for me to pick up and start working on first. Another thing too, we can add what are called SLAs or service level agreements around all of these different jobs. So, that if there are agreements you made with your customers that all their incoming orders are gonna be processed within the first two hours of arriving in the organization, we can start the clock at various points and keep it moving through there. So similar layout, we have our qualified data set is here on the left-hand side, On the right-hand side is our document. The things that are happening on this page are really around how we validate data from the ERP system to the data that's been extracted. So, we can extract data off of a document. But, that doesn't necessarily mean that it's in good enough shape to move off to the ERP to create an order inside your system. We want to validate that data. We want to enhance it and there's really only three places we get data in a process like this. We're going to get it off of the document, we're going to get it from an ERP system or we're gonna get it from a user. So, in this instance we're now combining two of those sources to validate this data, we're taking the data that's been extracted off the document and we're comparing it to data that's come back from the ERP So, each one of these blue buttons on this interface indicates another look up into the ERP where I'm tying together my data from my extracted document with the data that exists inside of the ERP. If I wasn't certain that this was a McMasters order or if I knew that for some reason this is actually an elite order, I can give it some information and if I search through here it's going to hit the ERP and the ERP Is going to via web service, send me all that data back and you'll see now I have a list of customers. Selecting any of those customers will fill out the rest of the information that's valid for this order. Interesting point: there is a mismatch of some of the data between what McMaster should be getting from an item price and what has been extracted off of the document. So, I'm able to point out places where the data doesn't match between where it's in the ERP and where it is in the document so that I can make sure that the customer service rep is posting the most valid data possible. Right? So, you'll see my line items here and then you'll see where this mismatch here is on the on this yellow portion. What's going to happen next? Once we're done with the workflow section is that we're gonna post that back into the ERP. So, if we remember the big beats of the order processing process, we have ingestion, the transformation services, workflow, and then the last thing we want to do is make sure that data posts back. So, once I have a meaningful verified set of data, I can actually post that back into the system. So that's a kind of a back-end process. So, there's not a whole lot I can show you as far as what posts and what doesn't, but I can show you via the example of this back end process where we look at what a workflow actually looks like, how these are actually composed. So, you'll see each one of these nodes inside a process has its own specified function. These functions can be anything from simply changing a variable from one data type to another. All the way to doing integrations into complex systems pulling data back so each one has a discrete function and the data moves through the process. We make decisions about various things through here whether or not the the data is valid; true; false. We work through this path each time, depending on what data is sent to the system. So, each one of these nodes then takes the data and ultimately once it's validated everything is clean. It will post it off to the ERP where then a clerk can make a final review, submit that order and then start processing that customer's order. So, those are kind of the big beats; a high- level look of an order processing solution using BPM software and and document extraction and if you have any questions or you want to reach out to us, you can find us at impactmybiz.com.