AI-powered automation for HR onboarding

Introduction

In today's digital era, automating routine processes is key to improving efficiency and reducing manual workloads. This article explores how UiPath's Document Understanding technology in conjunction with GPT (Generative Pre-trained Transformer) can significantly streamline the HR onboarding process.

The Onboarding Process

Let's consider an example where HR has received multiple documents from a new hire, including a graduation diploma, a W-4 form, and an ID card. The goal is to automatically extract relevant data from these documents to facilitate the onboarding process within backend systems.

Document Extraction and Classification

First, we will address each document type individually. The W-4 form is a new and unfamiliar document type, necessitating the training of a new model from scratch. Traditionally, training involves lengthy data labeling, but by leveraging GPT, the process is drastically simplified. By using the GPT predict function, we can automatically label fields in the document.

In cases where adjustments are necessary, such as the "Claim Dependent Credit" field, we simply fine-tune the labels provided by GPT. This innovative approach transforms a potentially week-long training process into a matter of days or even minutes for simpler forms. Following the labeling process, a new extractor—referred to as the W-4 model—is created using the dataset enhanced by GPT.

Meanwhile, UiPath offers pre-trained models to handle ID cards effectively, which streamlines that part of document extraction. For the graduation diploma, we access the generative extractor, as diplomas often contain unstructured data.

Workflow Automation

A predefined workflow within UiPath Studio simplifies the entire process. This workflow is responsible for downloading the attached documents, classifying them, and extracting all necessary data from each document type. To classify documents efficiently, we simply describe the document types—ID card, W-4 form, and university diploma—enhancing our workflow without the need to write complex rules or create new machine learning models.

For ID card extraction, the pre-trained specialized model is deployed. The W-4 form is processed by the newly trained model, while the generative AI is utilized to answer a variety of questions regarding the graduation diploma, including:

• What is the name of the graduate?
• What is the name of the university?
• What is the name of the degree?
• What is the graduation date?
• What is the specialization?

After processing, the data is synthesized into a response that can be emailed to the new hire, leveraging the integration of GPT with OpenAI.

Execution of the Workflow

Upon running the workflow, each action is executed step-by-step. If human validation is needed for any document, a task is generated in UiPath's Action Center for HR personnel to review. For instance, a task for the W-4 form appears, showcasing the excellent performance of the extractor trained with GPT. Once validated, the data is saved, and tasks for other documents, such as the graduation diploma, are also completed with minimal effort.

As the workflow progresses, the extracted data is instrumental in onboarding processes—enrolling new employees in systems like Active Directory and Workday.

Finally, we examine the email sent to the new hire. GPT has expertly summarized the extracted data, creating a personalized welcome message that enhances the onboarding experience.

Conclusion

By integrating generative and specialized AI models, UiPath is pushing the envelope of automation in HR processes. This collaboration allows for faster, more accurate document processing, providing companies with the tools they need to enhance operational efficiency.

Keywords

• UiPath
• Document Understanding
• GPT
• HR Onboarding
• Automation
• Data Extraction
• Workflow
• Pre-trained Model
• Generative AI
• ID Card
• W-4 Form
• Graduation Diploma

FAQ

Q1: How does GPT help in the document labeling process?
A1: GPT simplifies document labeling by automatically predicting and suggesting labels for various fields within a document, which drastically reduces the time required for data labeling from weeks to days or even minutes.

Q2: What types of documents can UiPath process for HR onboarding?
A2: UiPath can process various document types including W-4 forms, ID cards, and graduation diplomas, utilizing specialized AI models for each category.

Q3: What happens if human validation is needed during document processing?
A3: If human validation is required, a task is generated in UiPath's Action Center for HR personnel to review the extracted data and make any necessary adjustments.

Q4: How does the generative AI model assist with unstructured documents?
A4: The generative AI model can answer specific questions about unstructured documents, such as diplomas, extracting relevant information like names, graduation dates, and degrees.

Q5: What benefits does automating the HR onboarding process provide?
A5: Automating the HR onboarding process improves efficiency, reduces manual workloads, enhances data accuracy, and provides a smoother experience for new hires.