In a release already filled with groundbreaking AI features, we’ve saved the crown jewel – Machine Learning Model Foundry – for last.
We’ve already covered the dozens of new AI integrations and form lookups, AI Control Towers and enhancements to our wildly popular Generative AI for AppGen offering, but today’s announcement will not only prove special, but takes direct aim at the most challenging aspect of deploying AI that is hyper-tailored to your organization – building a predictive model!
Some time ago we covered how an AI developer can create a Machine Learning model directly in AWS SageMaker and use it in an AgilePoint application. If you haven’t already seen it, you can find it here. This capability represented a huge step forward in the automated personalization of process execution and proactive identification of exceptions, but did require some legwork in AWS SageMaker. We are happy to report that is no longer the case with the advent of Machine Learning Model Foundry.
What Does Machine Learning Model Foundry Do?
By leveraging context that AgilePoint captures automatically, Machine Learning Model Foundry automates the build and continuous retraining of your very own Machine Learning model that is 100% tailored to your organization’s usage patterns. All by simply by clicking through a configuration wizard.
Which Machine Learning Model Types are Supported?
The magic of Machine Learning Model Foundry is that you don’t need to decide this ahead of time, the right type will be automatically chosen for you based on the data you’ve selected.
With that said, the following types are supported.
Binary Prediction: These models will answer a question with a simple Yes or No. More accurately, a probability is provided and a threshold is defined to turn that into a final decision.
Use cases for this type of model include fraud or spam detection, medical screenings or loan approval.
Multiclass Predictions: Multiclass predictions build on the concept described in the binary prediction, with the main difference being the probability is assigned to the most likely outcome of multiple potential options.
Use cases in this model include document classification or sentiment analysis.
Numeric Prediction: Also called regression, in this model we aren’t predicting a category or a simple Yes or No, but a number. This could be a dollar amount, a percentage or count.
Use cases in this model include demand forecast, demand-driven pricing or real estate valuations.
What Are the Prerequisites?
Perhaps the best news is that because you’ve deployed AgilePoint, you’re already most of the way there. The rich organizational context that AgilePoint has been silently accumulating for years serves as the basis for your very own trained Machine Learning model.
Beyond that you’ll need an AWS SageMaker account (other platforms coming soon), an application with Data Entity as the primary data source and a minimum of 500+ rows in Data Entity for the historical data.
How Do I Get Started?
To get started, move into the process builder for your application, and click Machine Learning Model Foundry.
- Decide Your Prediction Field – the model is automatically selected based on your field type.
- Select Dependent Fields – the inputs used to make the prediction.
- [Optional] Record Filter – this removes any anomalous records based on your criterion.
- Define an Export Location for Training Data, Regeneration Schedule and Automated Training triggers.
- Click ‘Build and Deploy’ and the model is automatically generated in AWS SageMaker
- Orchestrate and govern the model through AI Control Tower to drive dynamic, corrective and self-healing actions.
Sounds easy enough? It sure is. Take a look at the video below and see this feature in action.
Moreover, as shown in the video, you can set up your app owner or any other interested parties to be notified using email once the ML Model is deployed and is in active state. The same information is also available in the Manage Center as shown below which can be accessed by navigating to Manage Center -> App Management -> ML Model Foundry.
