ChatCart: turning WhatsApp from a customer service headache into a D2C brand's best sales channel
A D2C snacks brand had thousands of customers messaging on WhatsApp, answered by hand, and marketing blasts that got the number flagged as spam. We built ChatCart: a WhatsApp catalogue and checkout, segmented and consent-based campaigns, abandoned-cart recovery, an AI assistant for common questions and a shared inbox for the team.

Three thousand unread chats
ChatCart's founder showed us a phone with thousands of unread WhatsApp chats: order questions, delivery updates, 'is this spicy?', and people who wanted to buy but did not want to visit the website. Two team members answered by hand. Occasional broadcast blasts to everyone got the number reported and restricted.
“Our customers already live in WhatsApp. We were making them leave it to buy from us.”
How customers wanted to buy
We read hundreds of chats and interviewed repeat customers. Many preferred ordering in chat, especially for repeat purchases. Most questions were answerable from the catalogue and policies. Customers were happy to receive offers about products they liked — not blasts about everything.
- Many customers preferred to order in chat
- Most questions answered by catalogue and policy
- Blasts caused spam reports
- Carts abandoned on the website could be recovered in chat

Commerce and care in one conversation
Customers browse the catalogue, add to cart and pay by UPI without leaving the chat. An AI assistant answers common questions using only the brand's catalogue and policies, and hands over to a person when needed. Campaigns go only to customers who opted in, segmented by what they bought, with frequency limits.
The team works from a shared inbox that shows each customer's orders alongside the conversation.
The team
A small team with an AI engineer for the assistant.
Chat analysis and launch.
Chat commerce flows and inbox.
WhatsApp integration, campaigns and inbox.
Assistant grounded in catalogue and policies.
5 people in total, working as one team.
Decisions we made
Agreed with the founder and marketing lead.
Link out to the website or buy in chat?
- Send website links
- Catalogue and checkout in chat
Our call: Catalogue and checkout in chat. Customers who preferred chat bought more when they did not have to leave it.
Blasts or segmented campaigns?
- Broadcast to everyone
- Opted-in, segmented campaigns with frequency limits
Our call: Opted-in, segmented campaigns with frequency limits. Relevance protected the number's quality rating and customer goodwill.
How much should AI answer?
- AI answers everything
- AI answers from catalogue and policies; hands over otherwise
Our call: AI answers from catalogue and policies; hands over otherwise. Grounded answers stayed accurate; people handled the rest.
Every feature, module by module
Everything that shipped for customers, the support team and marketing.
- 01WhatsApp catalogue
Browse products in chat.
- 02In-chat cart and checkout
Order without leaving.
- 03UPI payment links
Pay in seconds.
- 04Order updates
Confirmation and delivery tracking.
- 05Reorder in one tap
Repeat past orders.
- 06AI assistant
Answers from catalogue and policies.
- 07Human handover
Escalation with context.
- 08Shared inbox
Team works from one inbox.
- 09Customer context
Orders beside the conversation.
- 10Opt-in management
Consent captured and respected.
- 11Segments
By purchases and behaviour.
- 12Campaign templates
Approved message templates.
- 13Abandoned cart recovery
Reminders for opted-in customers.
- 14Frequency limits
Protect customers from overload.
- 15Campaign analytics
Reads, clicks, orders and revenue.
- 16Assistant analytics
Questions answered and handovers.
- 17Store integration
Catalogue and orders synced.
- 18Role-based access
Support, marketing and admins.
- 19Audit trail
Campaigns and consent recorded.

Launch
ChatCart launched with the catalogue and assistant first, then campaigns once enough customers had opted in. Chat became a meaningful sales channel, especially for repeat orders.
- Weeks 1–2Discovery
Hundreds of chats analysed; customer interviews.
- Weeks 3–4Design
Chat commerce and inbox flows.
- Weeks 5–11Build
Catalogue, checkout, assistant, inbox and campaigns.
- Weeks 12–13Soft launch
Catalogue and assistant live.
- Week 14Campaigns
Segmented campaigns to opted-in customers.
What we learned
Meet customers where they are. Buying in chat removed friction for repeat customers.
Consent and relevance protect the channel. Segmented campaigns avoided spam reports.
- Next.js
- NestJS
- PostgreSQL
- WhatsApp Business Platform API
- LLM assistant with retrieval over catalogue and policies
- Payment links with UPI
- AWS Mumbai

