Providing efficient access to knowledge and documentation with Generative AI at HR2day

The Challenge

HR2day was facing a growing challenge as their organization and cloud product expanded – how to manage an overwhelming amount of HR and information resources, which had accumulated for close to a decade. This information, critical for daily operations and compliance with legal standards, was scattered across various formats and storage solutions, primarily Google Drive, without proper organization or easy access. As a result, new employees found it challenging to navigate the maze of information, relying heavily on a few seasoned experts for guidance.

The situation was far from ideal. It slowed down processes and placed undue pressure on key personnel. And so, the growing team’s need for a streamlined way to access and share knowledge was evident. HR2day sought a solution that could organize their vast documentation efficiently, making it readily accessible to all employees and thereby enhancing operational efficiency and team autonomy.

The approach

This solution was first conceptually demonstrated during an Eraneos internal Innovation Day, during which the Data & AI team gets together to work on cutting edge tools and techniques chosen by the team themselves. The demo which was crafted by the Eraneos team and pitched to HR2day showcased the potential of Generative AI to improve knowledge management and information retrieval.

The project kicked off in July 2023 with a one-month development sprint with the goal to tailor the solution to HR2day’s specific needs, resulting in a Proof of Value. The solution integrated a process of document ingestion and conversion into a vector store, creating a dynamic, searchable knowledge base.

The result

To address HR2day’s challenges, Eraneos developed a Generative AI Virtual Assistant. This Virtual Assistant was designed to sift through HR2day’s extensive collection of documents, organizing them into a searchable and structured knowledge base. The solution utilizes Semantic Search functionality to find relevant information instead of the conventional term-based document search.

Eraneos employed Retrieval Augmented Generation (RAG) technology and semantic search to enable the GPT model to access the most relevant content from the knowledge base in response to free text queries. This approach ensured that HR2day’s team could quickly find accurate and relevant information without the need to read through all the documentation.

Another notable feature is how the model can identify when information related to a query is not available and inform the inquirer instead of hallucinating an answer – this helps to prevent confusion that might arise from irrelevant or erroneous answers. This solution not only streamlines access to information, but it also paves the way for more efficient and informed decision-making within the company.

Lastly, next to retrieving required information, the tool is capable of summarizing complex or long texts and creating custom rewrites. Summarizing helps to quickly understand the relevant aspects of complex texts. Rewriting capabilities help create customized versions of documents based on technical descriptions and combine sections from different documents. This speeds up the process of creating end-user documentation for specific situations.

The deployment of the Virtual Assistant marked a significant milestone in HR2day’s journey towards digital transformation. This solution has drastically improved the efficiency of information retrieval within the organization, reducing the dependency on key individuals and democratizing access to knowledge.