Machine Learning Builder, or ML Builder for short, is an OutSystems tool that helps developers without advanced data science knowledge build machine learning models and add them to their apps. It’s available in EAP, and you can apply here. In this post, I explain ML Builder in a little more detail and describe the problems it solves.
If you are you more of a visual person, I invite you view this session that explains how OutSystems enables you to add AI to your apps and introduces ML Builder.
Machine Learning Builder: A Quick Overview
ML Builder is a development accelerator that enables developers who are not professional data scientists to train machine learning models and deploy and integrate them into their apps. With ML Builder, developers can weave and evolve predictive analytics throughout their digital portfolios to power customer experience transformation and personalization.
With ML Builder, your developers or development team can:
- Easily and reliably implement valuable AI use cases that use your company’s data.
- Crush the skill set required to add ML to applications.
- Add AI to applications to improve efficiency and generate revenue opportunities through personalization.
- Reduce the costs of using ML for innovation and differentiation.
- Add predictive capabilities tailored to your data throughout your organization’s digital portfolio.
The end result is apps that use AI and predictive analytic to improve efficiency, reduce costs, and promote business growth.
How Is ML Builder Different from Other AI Tools Currently Available in the Forge?
The tools in the Forge fall into the first two of three use case categories. ML Builder is in the category called Packaged ML with Own Data. The goal of that category is to make it effortless to use your data in OutSystems to add high-value predictive capabilities and automate processes with no code or data science, using state-of-the-art technologies such as AutoML.
The other two categories are:
- Conversational experience: These use cases offer easily integrated components that enable multichannel chat and voice interfaces and in-device voice assistance.
- Packaged cognitive services: The use cases for these services include using AI to analyze, process, and automate their unstructured data, providing supported, well-documented, integrated components for this purpose. Examples are Language Analysis, Azure LUIS, QnA Maker, and Document Processor.
What Challenges Does Machine Learning Builder Address?
Gartner’s Top 10 Strategic Technology Trends for 2020 ranked AI and machine learning as the number 1 trend. Yet, according to the latest Element.AI survey, only 1 in 10 companies are able to take AI into production. In our quest to discover why, we found another Gartner report that lists three major barriers to adopting AI:
- Skills and staff
- Understanding AI use cases
The 2020 OutSystems developer survey had a similar list of hurdles. Plus, a 2020 McKinsey report shares yet another challenge - time and effort. Many organizations have found that 60 to 80 percent of a data scientist’s time is spent preparing the data for modeling.
In a nutshell, the adoption of AI is difficult and challenging. As part of our mission to help every company innovate through software, OutSystems wants to reduce development friction to zero. ML Builder contributes to that mission, making sure that nothing stands in the way of using ML models in applications. Here’s a quick look at how.
No Deep Data Science Skills and Staff, No Problem
ML makes it possible for developers without much data science experience to create machine learning models and add them right in their apps. For example, it can be used for automatic classification of a support ticket, to triage, and to predict priority. With it you can:
- Automate an approval decision, risk analysis, or business outcome.
- Classify tickets for triaging support, assigning to the right team, recommending a solution.
- Predict sales discounts, product demand, customer support demand, and available stock.
Use Your Own Data
ML Builder is one way that OutSystems enables you to use your own data in the platform to add high-value predictive capabilities and process automation with no code or data science; it’s all done automatically, under the hood, or using AutoML. It’s an integrated end-to-end experience for addressing specific use cases that require the automation of decisions and processes, such as categorizing tickets, prioritizing them or assigning the best agent.
ML Builder is just part of a larger group of features and capabilities for adding AI and ML to your apps. You can see how they all work in the global support use case here.
Although ML Builder abstracts away the complex aspects of machine learning, it is helpful to have a basic understanding of model classification. If there is a data scientist in your organization or someone else who knows this, they can serve as an SME. If you don’t have a SME, here is a basic rundown of the classification types:
- Binary classification: Trying to predict a boolean or binary (0/1 or buy/sell) value. Most use cases related to automating an approval decision (approve or decline), risk analysis (fraud or not fraud), or business outcome can be modeled as a binary classification problem.
- Multi-class classification: Trying to predict one out of a set of well-known possible values or categories (e.g. a ticket priority–high, medium or low). Similar to the binary classification problem, it is for predicting more than two values. Here, there are several use cases related to triaging (a multi-category decision) or categorizing transactions or customers.
- Text classification: Field prediction with input that is just text. Common examples of this are classifying customer reviews into certain categories, or prioritizing tickets just based on the ticket description or title.
There is also regression. A regression problem is when the output variable is a real value, such as “salary” or “price”
Fast and Effortless
In just two clicks and a drag-and-drop, ML Builder reduces data preparation and model-building time and effort. All you have to do is:
- Choose the kind of ML model you want to build.
- Choose what you want to predict and which historical data you will use to train your model and click Train.
- Analyze your model’s performance and add it to your OutSystems app with a simple drag-and-drop.
We’ve taken care of all the AI technology under the hood, as well as the infrastructure needed and deployment.
Watch this 2-minute video to see what it looks like.
Now That You Know What ML Builder Can Do...
You might have questions or want more details. Check out my blog post that answers questions about Machine Learning Builder. And if you’re an OutSystems customer and want to try it, sign up for our public EAP.
Not an OutSystems customer yet? No problem. ML Builder is also available with our free edition. Start today!