In Cambodia's property market, some foreign buyers have paid for property and received nothing in return. With up to 95% of agents operating without a license, the industry lacked a unified standard for accountability. To solve this, Pointer is building a platform to professionalize the industry, creating an ecosystem where all participating agents are required to be licensed.

The success of this mission hinged on making the platform indispensable. Pointer planned to leverage its decade of proprietary data to provide accurate, AI-driven property valuations that agents could rely on. But its team spent weeks manually cleaning inconsistent records for a single AI model test. Instead of innovating, the company's most valuable technical talent was consumed by maintaining internal tools.

Pointer's migration to Google Cloud was driven by a clear architectural goal: to build a reliable, end-to-end data platform. The team needed to automate the manual processes that were consuming their engineers' time and create a single source of truth that could power their entire suite of applications, from internal analytics to the agent-facing AgentHub app.

To solve for the intensive, manual effort required to build and maintain AI models, Pointer selected Vertex AI to serve as the core engine for its entire machine learning pipeline. The platform now handles the complete workflow, from ingesting and cleaning Pointer's decade of raw property data to training, testing, and deploying its sophisticated price prediction models. This automated the work that previously required a dedicated team of data scientists and achieved a 30% increase in valuation accuracy.

Automating the data pipeline with Vertex AI freed 75% of Pointer's AI team. The work that previously required four data scientists is now managed by one, allowing the other three engineers to focus entirely on innovation: They are now building new products, including a valuation tool for the banking sector designed to accelerate loan approvals. The tool allows a loan officer to get an instant property valuation — a process that typically takes days. The app, which leans on Looker for internal reporting, has already drawn a buyout offer from one bank. 

Meanwhile, its new architecture allows Pointer to update its AI models weekly, a process that took nearly a month. By accelerating its data processing and model training times by five times, Pointer is equipped to provide its agents with the most current market data — a distinctive competitive advantage.