AI Florist Singapore: a three-agent commerce platform selling flowers on web, WhatsApp, and Telegram
Buying flowers in Singapore is emotional, occasion-driven work. You are sending a get-well bouquet to a friend in hospital, arranging an anniversary surprise, or trying to find the right sympathy arrangement for a family you have never met. Every other florist online makes you scroll through fifty bouquets and hope the photo matches the moment. We wanted to test a different idea: what if you could just describe the occasion in your own words and have a bouquet designed for you on the spot? The catch was timing. We wanted to put the idea in front of real customers in 48 hours, not 48 days, and we wanted to keep building on it after launch.
3
AI agents live
49+
SEO pages live
Web · WhatsApp · Telegram
Sales channels
What changed since we started
- ↑ newNot ranked#1
Brand position on Google
For "aiflorist" and "ai florist"
- ↑ 0 → 49+049+
SEO pages live
12 occasion landings, 18 comparison pages, 19 blog posts
- ↑ 3x13
Channels people can shop on
Web, WhatsApp, and Telegram
- Fleur · Flare · Flair
AI agents in production
Design, grow, and serve, in that order
- 3.82x
Return on Meta ad spend
30-day snapshot at SGD 147 budget
- 1,049 → 3 purchases
7-day chat funnel
Page views, 154 chats, 12 add-to-orders, 3 sales
- 3
Real 5-star Google reviews
Cited on the storefront in structured data
- 172 modules
Platform scale
Across three Cloudflare-edge applications
The Challenge
Buying flowers in Singapore is emotional, occasion-driven work. You are sending a get-well bouquet to a friend in hospital, arranging an anniversary surprise, or trying to find the right sympathy arrangement for a family you have never met. Every other florist online makes you scroll through fifty bouquets and hope the photo matches the moment. We wanted to test a different idea: what if you could just describe the occasion in your own words and have a bouquet designed for you on the spot?
The catch was timing. We wanted to put the idea in front of real customers in 48 hours, not 48 days, and we wanted to keep building on it after launch.
Why we built a chat-first florist for Singapore
Most florist sites treat you like you are shopping for sneakers. Endless rows of product photos, generic "roses" and "lilies" categories, no real guidance for the moment you are actually buying for.
Singapore flower buyers do not shop that way. The job to be done is closer to "help me say the right thing" than "show me your range". A chat interface is the cleanest match for that job: tell the assistant what is happening, see a bouquet designed in seconds, check out without leaving the conversation.
To know whether the idea worked, we needed real customers placing real orders. A 48-hour launch forced ruthless choices. We shipped the bare minimum on day one, then kept adding everything else over the following months.
Three AI agents, one storefront
AI Florist runs on three AI agents, each with its own job:
- Fleur is the shopper-facing one. You describe the occasion in your own words on the website, on WhatsApp, or on Telegram, and Fleur designs a bouquet for you using AI image generation, shows you the price, and routes the order straight to checkout. Same brain across all three channels.
- Flare keeps the search and social presence growing. It writes blog posts, schedules social content, and ships SEO landing pages around occasions people actually search for in Singapore. Anniversary flowers, sympathy bouquets, get-well arrangements, and so on.
- Flair handles what happens after the sale. It opens cases for delivery exceptions and gives you a live link to track your order without having to call us.
Behind all three is a measurement layer that ties every order back to the channel and the search that drove it, so we can see what is working week to week.
A working commerce business, not just a demo
AI Florist is not a marketing demo. Real customers place real orders through Fleur conversations every week, leave 5-star Google reviews, and come back. The storefront ranks first on Google for the brand name. The 49 plus occasion pages keep adding fresh search visibility without us having to lift a finger. Meta ads return roughly four times what we spend on them.
This is the clearest demonstration of what we can build end-to-end for a client. Brand, storefront, AI agents you can actually use, a content engine, post-purchase service, and search visibility, all run by a tiny team. Not a brochure site. A working business.
Screenshots
The Results
3
AI agents live
49+
SEO pages live
Web · WhatsApp · Telegram
Sales channels