Success Stories
RAG Knowledge Engine – Using AI to Accelerate Sales Team Decisions
From Repetitive Questions to Instant Business Insights
When potential clients ask, “Have you built something like this before?” or “How long will it take?”, the answer often depends on how fast the sales team can locate past project data—information that usually exists but is scattered across documents and archives.
To fix this, the client asked us to build a RAG (Retrieval-Augmented Generation) Knowledge Engine: an AI-driven system that instantly pulls up relevant projects, timelines, and cost references from trusted internal sources. The result? Faster, more confident client conversations backed by real data.
The Challenge
The project aimed to reduce the time and dependency involved in gathering information for client discussions.
Key challenges included:
- Sales representatives often needed to ask Project Managers (PMs) repeatedly about similar past projects, delivery timelines, and pricing ranges.
- Project information was scattered across folders and documents, making it hard to retrieve quickly.
- There was no central knowledge base to reference when estimating ballpark costs or discussing technical experience.
The objective was to make company knowledge more accessible, reliable, and instantly retrievable while maintaining accuracy and data security.
The Solution
The RAG Knowledge Engine serves as an internal AI-powered search layer that retrieves information exclusively from trusted company data sources such as past proposals, project details, delivery timelines, and cost references.
It enables sales reps to get contextual answers about company experience and project data without interrupting project managers or engineers.
How It Works
- Data Collection and Preparation
The team gathers and organizes proposal documents, project records, and technical references. - Data Transformation
Files are processed into a structured format. Information is extracted from PDFs, converted into vectors, and indexed for intelligent searching. - Coding and Integration
Using Python and API integrations, the team builds the system to upload, store, and query data through FastAPI, a high-performance Python framework used to process requests and manage data flows, and Microsoft SQL Server (MSSQL), the client’s database platform that stores both structured information and vectorised data for AI retrieval. - AI Retrieval and Response
The AI model retrieves relevant data using vector search and generates contextual answers based only on internal information. - Access Control and Security
Authentication is managed through Microsoft Entra ID, integrated behind Keycloak for additional control and protection.
Initially, the system was deployed for the sales team, enabling faster proposal creation. The same framework can later expand to other teams or departments like HR, Accounting, etc.
Challenges for the Development Team
Building the RAG Knowledge Engine required both technical exploration and problem-solving.
- They handled complex data processing tasks, including extracting information from PDFs, converting it into vector format, and building a robust vector search mechanism.
- They integrated OpenAI’s embedding and conversational models to generate accurate, context-aware responses.
- Managing AI limitations required experimentation and fine-tuning to ensure results stayed relevant and within scope.
These challenges could be overcome by the team’s technical expertise in AI integration, data engineering, and secure system design.
The Results
The Knowledge Engine has already enhanced how the sales team operates
- Sales representatives can now find similar project cases instantly.
- Ballpark timelines and cost ranges can be estimated faster and more confidently.
- Dependence on PMs for repeated queries has decreased.
- Internal knowledge can be reused across multiple projects.
Technologies Used
Authentication and Security
Frontend
Backeng and API Layer
Why It Matters
The RAG Knowledge Engine transforms how teams access company knowledge.
By teaching AI to read and respond only within trusted internal data such as proposals, timelines, and costs, sales representatives can now make faster, data-driven decisions.
It helps the client’s sales team respond more confidently to inquiries, estimate projects more precisely, and start conversations grounded in facts rather than assumptions.
This is more than automation. It is about empowering teams with intelligent, reusable knowledge that accelerates business development and elevates the quality of conversations with potential clients.