WhatsApp AI bot

Nursery Hut sells fruit and garden plants and sets up whole orchards on empty farmland. Its Meta ads bring hundreds of WhatsApp messages a day. We built the AI agent that answers each lead, asks the right questions, scores it and puts the serious buyers on the manager's call list.

WhatsApp AI agent
Ad leads
Lead dashboard

the Nursery Hut lead dashboard, with 809 total leads, a call queue of hot leads and where people stop

The problem

The ads worked. WhatsApp filled up with hundreds of messages a day, about 25 new leads, and staff answered each one by hand. The typical first reply took 51 minutes. Every chat meant asking the same questions again, and the serious orchard buyers sat in the same pile as people just looking.

How it works, end to end

Ad click
What we built
When someone taps a Meta ad, the ad they came from is saved on the lead, and the dashboard shows which ads bring the leads. Orchard ads are spotted from the ad's own wording.
What it changed: The nursery knows which ads work, and orchard projects are marked hot from the first message.
WhatsApp greeting
What we built
The agent replies on the nursery's own number, on WhatsApp's official business platform. Staff can still reply from the phone app, and the agent steps back for 6 hours when they do.
What it changed: Every lead gets a reply in about 2 seconds, day or night.
Questions, in their language
What we built
One question at a time, in the customer's own language and script: Hindi, Hinglish or English. State, city, purpose, land, plants, size, timing and delivery. Voice notes are understood, nothing already said is asked again, and if the AI ever fails, a fixed question flow with buttons takes over.
What it changed: Nobody types the same questions into every chat, and no lead goes unanswered.
Scoring
What we built
Each lead is scored from what it says: how big, how soon, what kind of buyer. Hot, warm or cool, plus tags like orchard project and from an ad.
What it changed: The serious buyers rise to the top on their own.
Handover
What we built
It sends the 58-page catalogue and the nursery's map pin when asked. When a lead is ready, or wants a person or a site visit, it passes them to the manager with their answers written down. It never makes up a price.
What it changed: Hot leads reach the manager ready to buy, and they turn into orders.
Call list and dashboard
What we built
A call list sorted by score, a chart of where people stop answering, top plants and places, and every chat with its photos and voice notes. The team replies from the dashboard or the phone, sets a status and notes on each lead, exports to CSV, and signs in with a code sent on WhatsApp.
What it changed: The manager starts each day from a ranked list, not a scroll through WhatsApp.

The result

typical reply, down from 51 minutes by hand: 2.1s
WhatsApp leads in one dashboard: 809
leads finished the questions or reached the manager: 258
hot leads handed to the manager in the first 5 days: 38

The agent went live on 2 September 2026. In its first five days it handled 152 chats and 1,002 messages with 0 failures. By 28 October the dashboard held 809 leads, and 258 of them had finished the questions or reached the manager.

Replies in about 2 seconds, down from a typical 51 minutes by hand
Hot leads reach the manager ready to buy, and they turn into orders
38 hot leads handed over in the first five days, and 190 orchard projects spotted in the first weeks