Agentic RAG for Document Search

Management Summary

HMS developed a RAG chatbot for role-based document search for a global pharmaceutical company. The solution combines agentic AI, RAG, a vector store, and access control to enable employees to find, access, and work with approved documents in context.

Project summary

#lifesciences

Global pharmaceutical company

Industry: Pharmaceuticals and Healthcare
Project Start: 10/2024
Focus: Generative & Agentic AI
Use Case: Enterprise Search & Knowledge Management

Project Objective

Enable employees to find relevant documents faster, access them in context, and retrieve only approved content.

Key Features

Role-based document search across shared and personal documents
Vector store with access control for different organizational units
LLM-based chat interface for searching, querying, and further processing relevant documents

Tech Stack

python agentic ai langfuse qdrant chainlit ai chatbots & conversational ai embeddings llm observability

Organizational knowledge only creates value when employees can find exactly the information they need for the task at hand—and are authorized to access it.
Luis Wirth
Deputy Head of the AI & BI Competence Center at HMS

The Starting Point

Internal documents were scattered across various organizational units. The existing search function was unable to adequately determine which content was relevant and approved for specific roles, departments, or tasks.

Employees had to manually gather information from multiple sources and then determine for themselves which documents were relevant to their specific context. This created a need for a controlled search system with role-based access and a user-friendly chat feature.

Zwei Projektmitarbeitende besprechen auf einem Laptop die grafische Auswertung der KI-gestützten Pharma-Newsanalyse.

The HMS Solution

Role-Based Document Search with Agentic RAG

HMS developed an application for document-based search and processing. The solution combines a chat interface with a RAG architecture, role-based access control, and an AWS-based infrastructure.

Building a Searchable Knowledge Base

Company documents are converted into text embeddings and stored in a vector database. This enables the application to search content semantically, rather than simply returning keyword matches.

Role-Based Access to Documents

The solution takes into account which organizational unit a document belongs to and which user groups should have access to it. This allows documents to be made available based on roles, responsibilities, and approvals.

Chat Interface for Search and Follow-Up Tasks

Through the LLM-based chat interface, users can find relevant documents and continue working with the content they find. The application thus supports not only retrieval but also follow-up tasks.

Core Application and Infrastructure

HMS developed the core application, including the front end and database, and implemented the infrastructure on AWS. Langfuse supports the observability of the GenAI application within the project context.

Customer Benefits

Faster Access to Relevant Information with Role-Based Search

The chatbot helps employees find relevant documents faster and process them in the right context.

Faster

Finding Relevant Content in the Right Context

Employees can retrieve documents using an AI-powered search instead of manually compiling information from multiple sources.

Controlled Access

Control Access by Role and Organizational Unit

The solution takes into account which content is accessible to specific user groups. This ensures that access to information remains tied to existing permissions.

Integrated Workflow

Work with Documents Directly in the Chat Interface

The chat interface lets users work interactively with the content they find. Users can work with documents without switching between search, review, and follow-up tasks.

Our Strengths, Your Benefits

GenAI Applications for Document Workflows

combines GenAI engineering, application development, and cloud infrastructure in one integrated solution. In this project, the focus was on a usable application, beyond a technical prototype.

What You Can Expect from HMS

When it comes to GenAI applications, HMS not only handles the technical implementation of individual components but also integrates architecture, application, and infrastructure into a production-ready solution.

Development of the core application, including the front end, database, and AWS infrastructure
Implementation of RAG architecture with role-based access control
Integration of document search, chat interface, and downstream document processing

Christoph Bergen

CoE Lead GenAI

Would you like to carry out a similar project?

Talk to HMS about how your data, AI, or software initiative can be assessed from a business perspective, implemented technically, and integrated into your existing system landscape.

Discuss the project

Next Steps and Further Study

Generative & Agentic AI

Learn how generative and agentic AI help companies efficiently leverage knowledge, automate processes, and productively deploy intelligent applications.

Learn More About Generative & Agentic AI