A prospect asks "is this house far from the LRT?" — how does an AI answer precisely, instead of "quite near"?
Short answer: An AI can answer a distance question precisely only if it has the real address of the listing and it calculates — rather than guessing from words. With the right address, "is this house far from the LRT?" is answered "nearest LRT is Sri Rampai, roughly 4.4 km, a 7-minute drive" — not "quite near". Without an address, an honest AI should admit it is not sure, not invent a figure.
Do property prospects really ask this?
Of the 32 questions prospects typed into the merpati.my chat bubble between 25 May and 2 September 2026, three were about distance and two about listings — together 16%, the second-largest segment after pricing. Their actual words:
"macam mana ai kira kalau prospek tanya jarak rumah dengan tempat kerja dia"
[how does the AI work it out when a prospect asks the distance from the house to their workplace]
"how far from working place… how does the AI solve this"
Why is "quite near" a costly answer?
The prospect asked something specific and received an answer that cannot be compared. They ask another agent. The other agent opens Maps, screenshots, sends — 20 minutes. The unit did not lose; the answer lost.
Agents answer that way because they are driving to another viewing. Answering properly means pulling over, opening Maps, typing the address, screenshotting. So "quite near, 10 minutes" — and sometimes that 10 minutes is 25.
How does an AI calculate distance instead of guessing?
An AI that guesses distance is more dangerous than an agent who says "quite near" — it sounds confident and it is wrong. The right way, and Merpati's way:
- The AI extracts the listing's address from what the agent typed — a real address, not "the Setapak area".
- The AI extracts the place the prospect means — "the LRT", "my office at KLCC", "the kids' school".
- Distance and time are calculated by a map service, not by the AI. The AI only phrases the sentence.
- If the listing has no address or the prospect's place is unclear — the AI admits it is not sure and asks. It does not estimate.
| From listing | To | Distance | Driving time (no traffic) |
|---|---|---|---|
| PV21 Condominium, Setapak | Sri Rampai LRT | 4.4 km | 7 minutes |
| PV21 Condominium, Setapak | Suria KLCC | 9.6 km | 16 minutes |
The figures above were calculated from coordinates on 15 September 2026, not from memory, and the calculation record is kept. Two questions, two places, one listing — without the agent opening Maps.
What can the AI not do?
- The AI does not search for units for the prospect from a database. The agent types the listings; the AI answers from them.
- Distance is calculated without live traffic. "7 minutes" is an empty road — which is why the answer says "roughly".
- A mistyped address gives a wrong answer. This is a calculation on what was provided, not magic.
After answering the distance, what should the AI do?
Ask when they want to view. A prospect asking the distance to the LRT is not doing research — they are filtering. A precise answer followed by "would you like to see it this Saturday?" turns a question into a viewing. When the prospect says "Saturday works", the AI takes a name and a time and adds it to a list for the agent to confirm — instead of the agent negotiating in chat while driving.
Test it with one listing
Sign up at merpati.my/property, enter one listing with a full address, and ask "how far from the LRT?" from your own number. See whether it calculates, or guesses.
Sources
- Prospect questions from the merpati.my chat bubble, 25 May – 2 September 2026 (32 messages). Quoted verbatim, anonymised.
- Distance calculation PV21 Setapak → Sri Rampai LRT and → Suria KLCC, OSRM driving profile, 15 September 2026 — coordinates and calculation URLs kept in Merpati's internal records.