

Automating Construction Orders with Generative AI
Transforming email-based orders into automated workflows using Amazon Bedrock and RAG.
cubotoo AG, a dynamic start-up founded by the 125-year-old Swiss cooperative HG Commerciale (HGC), is disrupting the construction industry with an innovative B2B marketplace. Beyond direct trade, cubotoo's vision focuses on providing real added value through AI interfaces that alleviate manpower shortages and automate redundant processes. Partnering with copebit, cubotoo launched an AI innovation project to bridge the gap between traditional communication and digital efficiency.
The core of the solution is a managed RAG (Retrieval Augmented Generation) workflow powered by Amazon Bedrock. This system allows construction companies to send orders simply via email. The AI reads and interprets these requests 24/7, matching them against the product database to generate an automated order for approval within minutes. This eliminates the need for manual product searches and triggers deliveries faster than ever before.
The technical stack utilizes Claude 3.5 Sonnet as the Large Language Model (LLM), chosen for its superior analytical skills and human-like interaction capabilities. Amazon OpenSearch serves as the vector database, while AWS Lambda and Step Functions coordinate the complex logic. By integrating the knowledge base directly into the AI agents, cubotoo ensures precise and relevant answers based on its specific data sources. The result is a Bronze prize winner at the Best of Swiss Web Awards, boasting over 70 registered companies and 13,000+ products on a platform that continuously scales to meet the industry's evolving needs.
"Thanks to the cooperation with copebit AG, we came across ideas and opportunities during the course of the project that we had not previously focused on. Productivity has already been massively increased in the classic tendering and procurement process, and now we want to improve the experience even further."

Download the complete case study with detailed architecture insights and results.