300K+
documents classified using an AI chatbot
6
microservices developed in phase I
6,200+
L3/L4 support team members empowered across the globe
OWNING CHANGE IN TECHNOLOGY
Our client is a market-leading technology conglomerate providing a diverse range of marketing, design, development, and animation software solutions to over 30 million customers worldwide.
The challenge:Documentation Dilemma
Our client faced significant challenges in managing 300,000+ engineering, finance, technology, compliance, and operational documents – several of which were in unstructured formats. Their L3/L4 support teams, comprising over 6,200 employees across the globe, struggled to locate the right information quickly, leading to inefficiencies, data discrepancies, and increased operational costs.
To address these pressing challenges, our client envisioned building an AI-powered chatbot that could assist support teams by retrieving relevant information from this vast document repository in response to different queries. The system needed to handle diverse document types and deliver accurate, context-aware responses in real time. Amplifying the complexity, the solution had to be developed with minimal use of open-source technologies while ensuring strict information security.
Lacking the required expertise in-house, our client turned to TEKsystems to lead this initiative – transforming aspiration into innovation with a solution that blended intelligence, elegance, and bot brilliance.
Our proposal:Chatbot Champion
To bring our client’s strategic vision to fruition, TEKsystems deployed a multi-skilled, high-calibre development team with expertise in Python, Gen AI, and FAISS. Over the next seven months, our consultants developed a scalable, globally accessible, AI-powered support tool that delivered fast, accurate, and context-aware responses within seconds.
Six microservices were developed and rolled out in Phase I. The first three handled document segregation – classifying over 300,000 documents into structured, semi-structured, and unstructured formats, removing duplicates, and storing the rest in a central repository. A content-based folder structure was then created to organise documents into logical parent and subgroup categories.
The remaining microservices focused on intelligent retrieval. Using L2 search algorithms, natural language understanding, Gen AI, and FAISS Vector DB, the chatbot was programmed to semantically interpret queries and extract precise information from relevant documents.
Text
Text
Real-world Results
TEKsystems completed Phase I of the project in just seven months, delivering a fully operational document classification and retrieval tool built in Python and integrated with an AI-powered chatbot. As a result, our client achieved several positive business outcomes, including:
- Significant efficiency gains: The chatbot enabled L3/L4 support teams to retrieve accurate information in seconds, eliminating the need to manually sift through hundreds of thousands of documents. This has drastically reduced turnaround time for support queries and saved operational costs.
- Easy access to information: With documents now categorised into structured, semi-structured, and unstructured formats, and stored in a logical folder hierarchy, users could easily locate the right content. The chatbot’s natural language interface meant users didn’t need to know exact file names or locations – just their search intent.
- Improved quality and optimised workflows: The integration of FAISS vector search and Gen AI allowed for semantic understanding of queries, ensuring that even loosely phrased questions returned relevant results. This improved the quality and consistency of support responses, minimising incident resolution times and elevating customer experience in leaps and bounds.
- Scalable foundation for future enhancements: Following the successful completion of Phase I, TEKsystems is supporting Phase II of the initiative – expanding the AI chatbot’s capabilities to retrieve information from a database of over 700,000 customer records.