Sansiri adopts GenAI on AWS to speed responses, automate documents, and upgrade Thailand’s home-buying experience.
Sansiri has launched Thailand’s first large-scale GenAI program in real estate using AWS technologies. The initiative addresses rising operational complexity as the company serves thousands of customers across 500 projects. The challenge was clear: improve customer experience while reducing manual workloads across sales, service, and back-office operations. GenAI now supports both goals by accelerating responses and reducing friction in key decision moments.
At the center of the transformation is ChatX Assistant, a GenAI concierge built on Amazon Bedrock with Anthropic’s Claude models. ChatX delivers instant answers on financing, zoning rules, construction progress, and project timelines. This reduces service wait times and gives buyers cleaner, faster insights during critical steps. The system has cut customer response times by 30% and boosted satisfaction scores by 25%. It also lowers the support burden for Sansiri’s sales teams by handling routine questions with consistent, high-quality outputs.
Sansiri also uses Amazon Bedrock to automate internal document processing. A GenAI-enhanced OCR pipeline now handles more than 50,000 invoices each month with over 90% accuracy. This automation has cut processing time in half and freed employees to focus on customer care and project delivery. Sansiri is exploring agentic AI to extend this workflow. The goal is to create self-directed invoice processes that validate data, route tasks, and trigger approvals with minimal manual checks.
The initiative signals a significant shift for Thailand’s property sector, which is adopting AI at a rapid pace. GenAI gives Sansiri a strategic advantage by enabling faster decisions, more accurate operations, and consistent service quality across its large footprint. With AWS infrastructure and scalable foundation models, the company plans to broaden its GenAI portfolio and redefine how Thai families search, buy, and manage homes.
