• Project : Prod Fast
  • Company : UK TIC Firm
  • Date : 10 Aug, 2025
  • Duration : 3 month

Project overview

A global TIC leader relied on manual review of complex customer product documents to complete Product Questionnaires. Missing details often led to back-and-forth with technical teams and slowed BOQ preparation in Salesforce.

We delivered an automation that extracts product data with high accuracy, auto-fills questionnaires, cross-checks against historical BOQs in Salesforce, and generates clarifying questions to close gaps—speeding quotes and improving collaboration.

Client Challenges & Requirements

Reduce manual effort, avoid rework from missing info, and accelerate BOQ generation while leveraging past proposals to improve completeness and consistency.

  • Manual document review to complete questionnaires.
  • Missing details triggered back-and-forth with technical teams.
  • Delays creating BOQs in Salesforce.

Project Solution

AI extracts required attributes, auto-fills questionnaires, fetches similar historical BOQs from Salesforce to validate/suggest fields, and asks clarifying questions to close data gaps.

0%

Manual effort

10x

BOQ turnaround

100%

Data quality

10x

Cross-team efficiency

Technical Architecture Overview

The technical architecture of the Product Data Extraction & BOQ Automation Agent is built using Python, CrewAI, Llama, Qdrant, and LangChain, forming a multi-agent, context-aware system for automating document interpretation and BOQ creation. CrewAI manages specialized agents for document parsing, validation, clarification, and BOQ generation, allowing them to collaborate intelligently within a shared workflow. Llama models handle natural language understanding, entity extraction, and reasoning across complex unstructured product documents, while LangChain orchestrates prompt templates, retrieval chains, and memory to ensure contextual consistency across all agents.

Extracted data and historical BOQ embeddings are stored in Qdrant, which serves as the vector database for semantic search and validation. The backend, implemented in Python (FastAPI), provides APIs for document uploads, attribute extraction, and Salesforce synchronization, enabling seamless integration with existing enterprise systems. This cloud-ready architecture enables the end-to-end automation of technical questionnaire completion, validation against historical BOQs, and dynamic generation of accurate, ready-to-submit BOQs—delivering higher accuracy, faster turnaround, and reduced manual intervention.