Case study
Lobang: a WhatsApp property-matching agent built in 2 weeks
Grail built a working first version in 2 weeks. It turns property listings and buyer requirements sent through WhatsApp into relevant matches, while keeping contact details private until both sides agree to connect.
The problem
Property demand and supply already moved through WhatsApp, but matching them by hand was slow and inconsistent. Any automation also had to protect contact details until both sides wanted an introduction.
The system we built
Grail built a WhatsApp agent that reads listings and buyer requirements, checks property-agent details against the Singapore CEA register and ranks the best matches. Contact details stay hidden until both sides agree to connect.
WhatsApp
Temporal
Postgres
TypeScript
SlackThe result
The working agent was built in 2 weeks. Grail is now working with a small group of Singapore property agents to take the product into live use.
Results at a glance
| Primary result | Working WhatsApp agent built in 2 weeks |
|---|---|
| Operational result | Agent details checked against the Singapore CEA register |
| System scope | Contact details shared only after both sides agree |
| Delivery note | Live rollout now being prepared with Singapore property agents |
How we built it
We built and tested the product in five parts: understand the message, check the agent, filter unsafe content, rank relevant matches and get permission from both sides before sharing contact details.
- An agent sends a listing or buyer requirement through WhatsApp.
- The system checks the sender and turns the message into structured property details.
- Rules remove unsafe or irrelevant messages before matching begins.
- The system ranks current listings and requirements against one another.
- Both sides confirm the introduction before either receives contact details.
How people stay in control
The software handles repeatable checks and routine work. People still approve decisions that affect customers, money, privacy or compliance.
- Agent details are checked before a listing enters the matching pool.
- Unsafe or irrelevant messages are stopped before the system replies.
- Contact details remain private until both sides agree to connect.
- The team can inspect the steps behind a match when something looks wrong.
What we delivered
The finished scope included:
- WhatsApp intake for property listings and buyer requirements.
- Property-agent checks against the Singapore CEA register.
- Structured property details extracted from normal chat messages.
- Matching based on location, property type, budget and timing.
- Confirmation from both sides before contact details are shared.
- A private review trail for investigating failed or unusual matches.