What is AI Fabric? Pairing Artificial Intelligence with RPA

What is AI Fabric?

Robotic process automation (RPA) relies on the scaled intelligence of AI to learn new skills, such as visual understanding, process discovery and natural language processing (NLP). On the other hand, AI needs RPA to act as the delivery mechanism to solve real world business problems. It’s the combination of both of these technologies and their capabilities that encompass AI Fabric.

AI Fabric, the brain-child of UiPath, is an end-to-end intelligent skill creation and deployment tool for businesses looking to automate processes with minimal effort. AI Fabric connects seamlessly with UiPath studio and Orchestrator to create a unified experience within all UiPath Platforms. Through this tool, users developing automations can better orchestrate all functionalities of AI; deploying, consuming, managing and improving machine learning models.

AI Fabric is essentially the first step of deploying hyperautomation throughout your entire automation model. This tool allows users to rapidly examine all possible business processes with an evolving set of AI technologies, and identify what end-to-end processes are hyperautomation candidates. Through AI Fabric, both the human workforce and intelligent RPA work together to scale the automation capabilities in the organization.

Breaking Down the Machine Learning Model

Essentially within the AI Fabric platform, RPA bots are responsible for bringing Machine Learning (ML) into an organizations business processes. The Machine Learning model in AI Fabric is comprised of 3 parts – ML Packages, ML Skills and ML Logs. Each of these parts represent a section of tasks to be done before being able to implement your model within the AI Fabric network.

ML Packages

An ML Package contains all of the code needed to run your Machine Learning model. These packages are somewhat similar to those that exist within your RPA workflows in Orchestrator, and the UiPath Studio. Packages can be managed within Orchestrator via the ML Packages page, which allows you to view the packages available versions, statuses, and changing logs. You can also upload new package versions in this page, delete unused or undeployed packages, or learn more information about each one.

• PRO TIP: It is highly recommended that users acting as Data Scientists handle these ML Packages.

ML Skills

An ML Skill is a deployed package ready to be consumed by RPA workflows. When deploying a ML Package, the container exposes a REST API endpoint so that it can be called within a RPA workflow using the ML Skill activity.

Skills can be managed within the Orchestrator tab in the ML Skills page, where you can see all the models deployed on the tenant (both ML and OS). Also within this page, users can view the statuses of packages that were created through ML Skills, what version they are, predictions and descriptions, and whether or not a GPU is required.

• PRO TIP: It is highly recommended that users acting as Process Controllers handle model deployments.

ML Logs

AI Fabric deploys skills that are created within the ML Skills tool. This deployment entails optimization and security checks, installing various dependencies, setting up a network within the tenant namespace, and finally checking the overall performance of the ML Skill.

ML Logs are essentially a consolidated view of all these ML-related events, such as package validation, deployment and prediction errors. Each event illustrates the start and finish times along with an any errors that may occur. Logs can be managed via the ML Logs page within Orchestrator.

• PRO TIP: It is highly recommended that RPA Developers handle these specific workflows.

An AI Fabric Use Case

Given all this information, how and where can AI Fabric be used? Email classification is just one example of how AI and RPA can work together to accomplish a manual task that would have otherwise taken countless minutes, or even hours.

Through a simple UiPath workflow, AI Fabric can augment an unassisted bot to scan bulk emails (all with unique subjects and body copy) using ML Skill. These emails would then be sorted accordingly, classified by whether a message needs human intervention, or just an automatic response. Likewise, a similar use case can also be used for text classification across various industries.

Through the intelligent and scalable capabilities provided by AI Fabric, users can govern their own automations, save minutes or even hours on daily tasks, and focus on innovative drivers within their organization.

Keep Reading: What is Hyperautomation? How to Automate End-to-End Processes