Published On: 24 June 2024

Company Overview

Empatia is a leading human resources consulting firm that specializes in talent acquisition and employee development. With a focus on talent acquisition and employee development, Empatia serves a diverse range of clients across various sectors, including technology, finance, and manufacturing. In an industry where attracting and retaining top talent is crucial, Empatia prides itself on its rigorous interview process and commitment to hiring best candidates for its clients for over 500 candidates annually.

Business Challenge

Empatia’s interview process involved manual summarization of notes taken during candidate interviews. This process was time-consuming and prone to errors, as HR professionals had to carefully review and summarize each interview, ensuring accurate representation of the candidate’s responses and qualifications. Additionally, Empatia’s HR team spent significant time checking for proper grammar and wording in their notes, further adding to the workload.


Technology Challenge

To streamline the interview process and reduce the manual effort required, Empatia sought to leverage generative AI technology to automate the summarization of interview notes and provide real-time assistance with grammar and wording. However, building a custom generative AI solution from scratch was a daunting task, requiring significant resources and expertise in areas such as natural language processing, machine learning, and model training.


ICS Compute, an AWS Partner, proposed a solution leveraging Amazon Bedrock to build a virtual assistant powered by generative AI. The solution involved fine-tuning the Claude v3 Haiku and Sonnet models from Anthropic, using Empatia’s interview template as the knowledge base.


The virtual assistant was designed to provide two key functionalities:

  1. Knowledge Base Summarization: The virtual assistant could automatically summarize interview notes based on Empatia’s predefined knowledge base, which consisted of their interview template and best practices.
    • Users access web application using Frontend Application, the chat interface
    • Application check User Authentication
    • Backend application process after get success for user authentication
    • Transaction save into Database for BotTable and Conversation History
    • Storage Document is for put document ingestion
    • Process document ingestion & embedding from Amazon Bedrock (Cohere Multilingual Embedding)
    • Result Ingestion document and embedding save into Database Aurora Postgre (pgvector)
    • Amazon Bedrock process chatbot using LLM and get content data pgvector from Database
    • Transaction for saving conversation log into storage
    • Athena get conversation log from storage to analyze usage
  2. Chat-based Summarization: HR professionals could engage in a conversational interface with the virtual assistant, allowing them to receive real-time summarization and feedback on grammar and wording directly through the chat interface.
    • Application check User Authentication
    • Backend application process after get success for user authentication
    • Chat transaction save into Database for Conversation History
    • Amazon Bedrock processing for user chat
    • Transaction for saving conversation log into storage
    • Athena get conversation log from storage to analyze usage

The GenAI HR Assistant solution was adopted by Empatia’s entire HR team, consisting of 3 recruiters and 1 HR manager. This solution streamlines the Empatia’s interview process for over 500 candidates annually across Empatia’s operations in Indonesia and helps Empatia identify top talents more efficiently.

The solution leveraged a few-shot prompting by providing the Claude v3 Haiku and Sonnet model from Anthropic with a small number of examples from Empatia’s interview templates and processes. This allowed the model to quickly understand the context and generate relevant responses. ICS Compute chose Anthropic for its Responsible Scaling Policy. 

Additionally, an advanced Retrieval Augmented Generation (RAG) implementation was employed, which enabled the model to retrieve and incorporate relevant information from Empatia’s knowledge base during the summarization process.

Cohere’s Multilingual Embedding Model v3 was used to generate vector representations of Empatia’s knowledge base, which included both Bahasa Indonesia and English content. These vector representations were then stored in the Amazon Aurora PostgreSQL database with the pgvector plugin, enabling efficient retrieval of relevant information during the summarization process.

On Responsible AI

In line with the Fairness principle, this use case ensures that the AI assistant does not directly participate in candidate scoring or influence hiring and termination decisions. The AI’s role is limited to enhancing written communication through summarization, grammar improvements, wording, tone, and style refinements, without altering the user’s original context or intent. This approach mitigates the risk of perpetuating biases in high-stakes employment decisions.

Additionally, adhering to the Human-Centricity principle, the AI assistant is designed to augment and streamline communication processes while preserving human decision-making authority. This human-centered approach considers the broader societal impacts of AI in employment decisions, striking a balance between leveraging technological advancements and upholding the primacy of human judgment in consequential hiring and termination decisions.

Outcome and Benefits

  1. Increased Efficiency: The virtual assistant reduced the manual effort required for summarizing interview notes and checking grammar and wording, saving an average of 1 hour per day per HR team member.
  2. Cost Savings: With the increased efficiency and productivity of the HR team, Empatia estimated an annual cost savings of $27,000.
  3. Scalability: The solution built on AWS services allowed Empatia to easily scale the virtual assistant to handle increasing interview volumes without compromising performance.

According to Wenny Halim, Empatia’s HR Manager, ‘The GenAI HR Assistant has been a game-changer for our team. Not only has it streamlined our interview process, but it has also helped us identify top talent more efficiently, giving us a competitive edge in the market.

Re:Invent Go:Beyond

Recently case study

  • June 24, 2024

    GenAI Virtual Assistant for Contact Center Application Provider

  • June 24, 2024

    GenAI HR Assistant for Empatia Talenta Indonesia on AWS