Name: Transform SAP Master Data Automation with Intelligent Autocomplete in Automate Evolve
Uploaded: 2026-04-28T16:16:11.041Z
Duration: 1 h 8 s
Description: Transform SAP Master Data Automation with Intelligent Autocomplete in Automate Evolve

Transcript for "Transform SAP Master Data Automation with Intelligent Autocomplete in Automate Evolve": Hi. My name is Sigri Kok. I'm a principal SE at Precisely, and I'm super excited to show you how you can set up and use intelligent auto complete for your master data processes in Evolve. Let's get started. A little bit of housekeeping. The webinar today is through your web browser, and the audio was streamed through it. If for some reason you have an issue, please message us or drop and reconnect. We also welcome questions, so please put them in the q and a chat. I'll be happy to answer as many of those as I can at the end, and we'll be providing a recording in the slides as well as answers a few days after this webinar. Let's first take a look at why we even contemplate leveraging intelligent auto complete. What will it buy us? So for anybody that's seen us present about our Evolve process automation, it's that intersection of process and data. So Evolve can address those SAP data intensive processes. Now a designer would need to know which SAP objects are involved, which groups are involved, whether they're contributing data or approving data, you know, the flow. They'll wanna create studio automations, what we call scripts or leverage APIs. They wanna know which fields get defaulted, maybe they're hidden and not even displayed, which fields require values, do they have lookup values, which values go with each other that are related, create views for those different teams, have fields be editable or not editable, and then knowing the business rules that need to be applied to that data to get it first time right. And that applies for both creation and maintenance, And those rules can change over time. Now when you're creating those business rules, and sometimes there are hundreds of them in a solution, it often requires that you contact the SMEs for the each of those areas, and that relies you going to multiple teams. Sometimes the master data team knows all of those rules, but sometimes that data is owned by other teams, and you'll want to track down the SMEs to get that knowledge. Sometimes that's tribal knowledge, and then sometimes those business rules don't cover all of the use cases. So we want to have the ability to ease the burden of finding all those business rules and covering all use cases and making sure then that your data is complete and correct and also provide a good end user business user experience. So when we talk about the master data, we have a set of inputs that can then potentially predict outputs for a number of areas. And most of that data has very deterministic properties that allow something like this to happen. And here, we just have a smattering of those. Maybe based on plant and product type and a few other things, we can predict most of the rest of the values. At a high level, Evolve intelligent autocomplete is precisely the own algorithm for advanced search and autocompletion of output values. There's a field control table or what I call mapping as part of the solution where you would choose what are the input values that are going to predict the output values. And based on some settings, it will either default the values and it will color code it appropriately. It's an extensible framework that'll work across material master, but also other master data domains. In addition, the framework will support other types of models, and so you'll see more coming. So what does that look like from the end user perspective? They're presented with a form. There are a set of input values that you select. And based on the advanced searching and auto completion, it will predict the values for the output fields and color coding them appropriately and basically prepopulate those values without writing any business rules whatsoever. So let's take a look at the setup, and I'll be walking through some slides, but then we'll toss those and go live. So in order to use this, you would want to install a version of Evolve that has the intelligent autocomplete early adopter release as part of it. So there's a there's an installation process, and then the global administrator would set up some system variables. The ML algo allows you then to create libraries for those models, and the ML service is an IS site needed to process those requests. Then we'll add a library that will enable you to create a new intelligent autocomplete instance and create models and train those models. Then, currently, today, we would want you to extract that data from your SAP system. You can leverage query. You would want to include enough tables to include all of the attributes that you want to be part of your model. So this might, for material master, include the Mara table, which has a general data, mock, m a k t, for description, mark, m a r c, for plant data, and whatever else that you want. You can create different models for different parts of the material master or other types of SAP master data objects. So this is up to you to design and include the fields that you think would be part of the process. Note that the column headers in your extract become the fields in the model. Here, we're using table name and the geeky field name, but you could also use descriptions. Then you would add a solution for intelligent autocomplete and provide a path to that particular extract file, and you'll see it load. Once it loads then, you can select it and create models. Then you'll see here the model information. So we'll give it a name and a description and the model type intelligent autocomplete. And this is where the framework for this will be very extensible. We'll be adding other options and potentially offering other model types. Then you'll be able to choose, information related to your field. It's gonna find all the fields in that extract, and you'll submit it. Then you can choose that model and train it, and it you'll see that it's queued. And once it's completed, you'll see that it's trained. And now it's ready to use in your Evolve solution. Let's take a look at this live. So here I am on an Evolve site, and we've got the ML Algo set to true. And, again, that's going to let you create libraries associated with intelligent autocomplete. And then the actual service, that's the IIS service that gets set up as part of the installation that processes the requests. Other than that, there's really not much other setup other than the install. Now inside an app, I can come over to manage libraries, add a library, and we'll have a type called Evolve AI. And that library type then would allow you to create an intelligent autocomplete solution as well as models. Let's take a look at one that I've set up. So here's the evolve.ai model. And if I wanted to add a solution, I give it a name, a description, and then that path to the actual extract. And once you enter in that information, you click add, and it will load the data and determine the fields. So if we look at that, you will see that that this was the name and description. I had the path. And now I can take a look at the models. So I created a model. I can choose here to add a model, give it a name, give it a description, and then the type. It'll be intelligent auto complete. In your version, you might see some other things coming, and we'll be happy to do an educational webinar on those. Then once you have that model defined, you can choose it and click to train it. And now this one, is already trained. But let's take a look at the fields that it found. This is the field name. This is whether it's AI derived or not. And it doesn't matter if it's going to be an input or an output in your solution. It would just be true or false. So you'll see everything defaults to true. And in this case, for purchase order unit, we change that to false by clicking edit, and we uncheck the AI derived value and made it a fixed value. But you'll see as I click through here, pretty much everything else is going to become part of the model. Again, it doesn't matter if it's an input or output at this particular point in time, and everything will default AI derived as true. You will have to change it if you don't want it to be part of the model. And, again, you could then not extract that field if you don't want it to be part of the model as well. So it's that simple. And, again, this this AI model is already trained. If it wasn't, you check it, and you would click to train the model. So here I am in query, and we've got a number of tables that are included. We'll set search criteria if we need any runtime criteria. Okay. So now I have a CSV file with the fields that I want. Now let's add a solution. And let's browse to my file. We'll choose add, And we'll go to models and add a model. And there's all the fields that we have. And I'll submit. And you see that it's not trained. So now we'll train the model, and it's queued. You'll see now that it says training, so that's in progress. Now that the model status for training is trained. So we're all good to go, and we're ready to try that in a solution. Now let's take a look at how to use that model in an Evolve solution. There's a new option here in the solution to import a model. Click import and then choose the model. This is where net when the fields get displayed, you'll see all of the field names. You can map those to existing fields if you have them or generate them new from the model. And everything defaults to output, and you would come through and select which fields become input. So we'll choose input for description and that it's a substring and input for plant and that it's a prefix. And those are the two. In my particular solution, your input fields would be different. And these can change over time. This is where you can play with what are the input fields, what are enough input fields to reliably predict with certainty what the output values would be. Then we have a form solution control with buttons. You'll leverage an intelligent autocomplete control, and then there's a set of buttons that you can choose to use. One of them is to run the model. One of them for clear fields is to clear the output fields so that you can rerun with new values. Reevaluate is used in case you change the values away from what was predicted, and you can determine whether that meets the level of certainty or not. You can change the colors as well as change the predict value range. Right now, it's at 90% or higher. We'll set the predictive value. By default, this option here is not checked. So if it's less than 90%, then it won't fill in the default. I select to do that, and then you can check change the colors. So blue, by default, means value not predicted. Between zero to 90, predicted, but not a high level of certainty, and green would be 90% or more. So these are all configurable and easy to drop into your web form. So let's take a look at the setup. So I have one already built. We'll take a look. This is where the AI model option comes in. So I can click import, and any models that I have can be selected. I currently only have one on this system. And then in here, I can change the mapping. This is where you can determine maybe I chose description and plant. Maybe I also wanna include material type or something else, so I can scroll to find it and then choose it as input. If I chose it as input, then it would be a prefix or substring for the search. Then I can come here to a web form the web form, drag and drop the fields for the inputs and outputs. There's the intel intelligent autocomplete control, so I simply drop some form buttons on here and then map them for the run, the clear output values, and reevaluate the predicted, whether it's, you know, high or low range. 90%, I left as a default, but I checked this box because I wanted it to set the predicted highest predictive value even if it was below 90%. And then I have the color scheme here. You can choose to change that. It's simply choosing this and basically changing the colors. Super simple to use, and you'll see all of these have sub labels. Now I like bigger sub labels, so I came over here to the default and changed my sub labels to be a bit bigger. Let's create a new one from scratch, and I'll come here to import the model. There's all my fields. And, again, I can come in here and choose any of the ones I want. So I'm gonna for input, and it defaults to prefix, which is fine for plant, but I want a substring for description. Again, I can include other fields as well and click okay. I like new groups. So what I wanna do is I'll just throw all the fields here on the output. So this will be output. But what I wanna do then is find description and plant, and I'll drag that up to my inputs. So this is gonna be inputs. If I can spell. And then what I wanna do is come over to the form elements and go to the advanced elements and find intelligent auto complete. So we'll add that. And here, we'll point it to a model. Now remember, for each part of material master or customer or supplier or whatever master data object you're using this on, you could have multiple models. So maybe I have one for general data. May I have a separate one for MRP data, separate one for sales data. Completely up to you. And then what I wanna do is add some buttons. So I'll add three buttons. And for anybody who knows me, I like formatting things nicely. So I'm gonna come in here and put in a spacer. There we go. So we got all the buttons, and I wanna call this run. I wanna call this clear, so clear output. And this one is going to I'll call it check check data after changes. You you can you can basically call it whatever you want, but this is that reevaluate. So now that I have buttons, I can click that I wanna see a run button and a clear button. You don't have to have a reevaluate if you don't want to. So that's gonna be run. I'm mapping it to my field names here. So clear output is the clear. And then reevaluate. It's basically going to check the data after changes. This is where the range is. I could lower that or higher it. I like setting the predicted highest value. You don't have to do that. Some people don't want to do that. Because it's not 90% or higher, you might wanna leave this empty. The field won't be predicted, and that requires a set of eyes to look up values or enter values. I'm going to set that the highest probable value, even though it's 90%, it's still selected. And then I can choose the color scheme. So green green tends to be you know, this is 90% or higher. Yellow, warning. And then the not predicted is blue. You could make that, I don't know, purple if you wanted to. So you'll just click apply. I'll go ahead and change my columns a little bit, and there we go. So we have a solution now pointing to the model with the fields from that model and buttons to actually run the intelligent autocomplete. Let's go ahead and deploy it. And before I run through that, let's go through the demonstration. So the demonstration is going to show based on my input fields what my output fields are that came from the model. And by default, it's using this color coding. You could also use other types of formats. We have different kinds of style sheets that can display the predictions a little differently. And then there's also a setting here. By default, the top three values up to three values will be displayed here. You can change that as a global setting in the workflow settings. So let's go back and take a look at what I just built, and you'll see I have description and plant. And I'm just going to put in part of the description. And you'll see here now green means 90% or higher, so this value is 99.76. The purchasing value key is not it's 100%. This one's not part of the model. It's defaulted. Here, we've got 86%, and it filled it in. Now if I didn't check that box, this yellow one would be empty. And you'll see as I scroll through here, there's quite a few yellow ones. Let's let's see if I add a plant. I'm gonna clear the output so it's not reusing the values that were there, and click run and do it again. And you'll see here we've got a lot more green. In fact, we only have one yellow. Now perhaps the predicted value was not the correct one. So here, I might wanna change this to z four, which is one of the other values. And now if I check this after making changes, this is just telling me, hey. This is not above 90%. Perhaps it's a correct value. That's what the reevaluate does. And now if this was part of my solution, I could, you know, again, check all of this out. And when it's ready, move it along in the process. That's how simple this is to include in a solution and use. And you'll see in my solution, I have no rules. I don't have to maintain a matrix of fields that says if I choose this value, these other values are the potential values. I don't have to write any rules or set defaults. It just works. And I'm not a data scientist nor am I an expert in large language models. The goal of this was to make it super simple for business people who understand the data to leverage, to set up, and use in an evolved solution. So the style sheets include by default, again, it's this solid color. You can also use outlines, and you could use this little gradient display where where it's clear and has a little bit of color on the side. And there are places in the documentation where you can go and download the different style sheets. So I hope you can see the benefits of Evolv Intelligent Autocomplete. This gives an improved end user experience because much of the data is predicted with a high level of certainty. This provides you improved data quality for those solution designers, much faster solution development and maintenance because you're not tracking all those rules or data grids that feed the solution. There's reduced complexity to building the solution. There's less reliance on those subject matter experts or IT to determine all solutions, it's much simpler than most solutions out there. Don't require data scientists. Don't require those large language models. You just point to the data, build the model, pull that into the solution, determine your inputs and outputs, and you're off and running. Again, use cases for these scenarios for intelligent auto and complete include material master creation and updates, but there could be other use cases for other types of SAP master data. And it can also be used for identifying issues and data cleansing, which our customers who have been early adopters have found. Again, we're initially targeting the MetroMaster. So for next steps, we have an early adopter program. You can reach out to your account team. You'll work closely with our account team, product management, and engineering to ensure that this supports your use case. You'll be able to provide feedback, and there's no licensing cost for that early adopter program. Please reach out to your automate team to learn more. We'd we'd love to talk to you about it. So let's go through some of the questions that I received, and some of these may have been answered during the webinar, but a little bit of reinforcement can be good. So the first question had to do with where this data gets stored. So currently, you would extract that data from SAP as I showed you earlier, and then we would load it into a SQL database and basically train the model. So all of it is behind your firewall on your servers, whether they be cloud or on prem or that we host it. So your data's secure. There's no large language models necessary. You don't have to bother IT, nor any kind of data scientist to set this up. So one of the other questions had to do with, what kind of master data this supports. Now we showed you material master because we find based on our vast experience that a small set of inputs could predict what the outputs would be that to set those defaults. And it works really well for material master, and we had a number of early adopters prove this out. You could also use it for other types of master data like customer or supplier. One of the other questions had to do with, the inputs in order to predict the outputs, and that's something that you can play around with. One of the ways I did that was initially just simply in Excel and using filters, to to be able to determine does this narrow it down enough. But you can simply import that as I showed you into a solution and then change the way the mapping works. Choose your search fields, whether it's a prefix or contains, and then have it determine what the outputs are. And you can change that over time. Now we've had a number of customers use things like a description and plant, which we showed you in the demonstration. You could also use material group, product hierarchy. Some of our customers wanted to use class characteristics. So that's completely up to you. And, again, you can also decide whether those fields are included in the training. Some fields would always need a set of eyes and potential of selecting the values, and you can also default instead of using a predicted value. So those are things that are really easy to change and deploy. Again, our customers that have tried this have verified that indeed it's quite accurate. And, yeah, and then we had a question which I always have is when is this coming out in GA, and the plan for that is later this year. I don't have an exact time frame, and, of course, that could potentially slip. So please reach out to your team, and we'll keep you updated on when it's available. Now I didn't get to all the questions, so I will be following up. We'll also share a recording of this as well as the slides. And, again, please reach out to your account team. We're happy to talk to you about this. Thanks for attending.