WhatsApp Commerce for Retail
AI chatbot for product discovery, ordering, and customer support entirely on WhatsApp.
Client
Premium D2C fashion retailer
Industry
Retail / D2C Fashion
Region
South Asia + Middle East
Duration
4 weeks build + 2 weeks soft launch
A premium D2C fashion brand was losing high-intent Instagram traffic at a generic web checkout. We built a WhatsApp-native commerce flow with an AI stylist concierge - handling browsing, sizing, checkout, re-engagement, and care all in chat. Result: WhatsApp became the brand's most profitable channel - 65% of orders flow through it, repeat purchase tripled, and AOV rose 28%.
Headline results
The client is a premium fashion brand with 80% of inbound demand starting on Instagram. The web checkout converted at industry-typical rates but felt nothing like the curated brand experience customers expected. A small team of stylists handled the highest-touch chat customers manually - and the volume capped at their hours. Email re-engagement was sliding under 12% open rate.
What we built and shipped
WhatsApp catalog & in-chat checkout
Customers browse curated collections via WhatsApp, ask styling questions, and complete checkout - all in chat with a payment link delivered at the end.
- Native WhatsApp catalogue with brand-curated collection drops
- Inline product card → sizing question → in-chat payment in under 90 seconds
- Cart persistence so customers can pause and resume across hours
AI stylist concierge
Trained on the brand's tone, sizing guides, and styling philosophy. Handles fit questions, recommends complete outfits, and hands off to a human stylist when needed.
- Brand-voice grounded conversations - never sounds like a chatbot
- Outfit-completion recommendations beyond single-item upsell
- Easy 'talk to a human stylist' keyword that hands off cleanly
Behaviour-triggered re-engagement
Behaviour-triggered WhatsApp campaigns based on past purchases, browsing, and abandoned carts - with personalised AI-written messaging instead of templated blasts.
- Re-engagement triggers tied to RFM cohorts, not calendar
- AI-written messages personalised to past purchase aesthetic
- Reply rate of 38% vs single-digit on email
Order tracking & returns in chat
Customers ask 'where's my order' and get a real answer pulled from the fulfillment system. Returns initiation handled in chat without any web visits.
- Live order status from the ERP, never a stale answer
- Returns initiated with a chat command - pickup booked automatically
- Proactive WISMO outreach for delays before customers ask
How it actually works
WhatsApp Business Platform
Official Business Platform with multi-number routing, per-brand sender identity, and full template approval pipeline.
Conversational stylist agent
Brand-tuned LLM with retrieval over the catalogue, sizing guides, returns policy, past-customer history - handing off to human stylists on signal keywords or low confidence.
Commerce backend
Direct integration with the existing Shopify store, payment gateway in-chat, real-time inventory and order sync.
Behavioural re-engagement engine
RFM cohorting + behavioural triggers feeding the AI message generator. Templated outreach is approved once; the actual sends are personalised per customer.
Phased delivery timeline
Catalogue, voice & templates
Built the WhatsApp catalogue, captured the brand voice from past customer service transcripts, secured template approvals from Meta for outbound flows.
Stylist agent + checkout build
Shipped the AI stylist with retrieval over catalogue and sizing data. Built in-chat checkout with the payment gateway integration.
Re-engagement & returns
Built RFM-cohort re-engagement and returns initiation in chat. Trained the human stylists on the handoff dashboard.
Soft launch + scale
Soft launched to past customers, gathered satisfaction data, then opened to the full Instagram traffic flow. Tuned weekly on conversion data.
Before vs after
Same business, same team - measurably different operating model after the engagement.
- Channel where most orders happenWeb checkout
- Average order valueBaseline
- Repeat purchase rateBaseline
- Re-engagement message reply rateUnder 5% (email)
- Chat satisfaction scoreN/A - was email or none
- Returns initiation flowWeb form + email
- Channel where most orders happenWhatsApp (65%)
- Average order value+28%
- Repeat purchase rate3× lift
- Re-engagement message reply rate38% (WhatsApp)
- Chat satisfaction score4.2 / 5 stars
- Returns initiation flowOne chat command
What changed and by how much
Operational and revenue metrics tracked from go-live, measured against the pre-engagement baseline.
Composition of impact
Approximate breakdown of how this engagement contributed to the business outcome - the headline metric is a roll-up of these levers.
- WhatsApp checkout conversion32%
- AI stylist AOV lift24%
- Re-engagement reactivation revenue20%
- Repeat-purchase compound effect14%
- Returns & care cost-out10%
What we built it with
Messaging
- WhatsApp Business Platform (Meta)
- Multi-number setup with per-brand sender ID
- Approved template library
AI & commerce
- OpenAI GPT-4 with brand-tuned prompts
- Vector retrieval over catalogue + sizing
- Shopify Plus storefront integration
Payments & ops
- Razorpay + Stripe in-chat payment links
- ERP integration for live inventory
- Klaviyo for email holdouts
Analytics
- RFM cohort engine
- Conversation-level CSAT tracking
- Per-stylist dashboard for handoffs
What we de-risked along the way
Off-brand stylist replies
Mitigation: Brand-voice eval suite run on every prompt change, plus human-stylist review of the first 50 conversations per new collection drop. Two-strike escalation for any out-of-voice response.
Meta policy violations leading to number bans
Mitigation: Strict template approval pipeline, opted-in lists only for outbound campaigns, per-message quality scoring with auto-cooldown on high-complaint patterns.
Customer feels they are 'just' chatting with a bot
Mitigation: Honest framing at start of conversation, easy keyword handoff to a human stylist, satisfaction tracking per conversation. CSAT held at 4.2/5 stars.
What we'd carry into the next build
Brand voice trumps everything
An LLM with the right brand voice and the right knowledge feels like a stylist. The same LLM with generic voice feels like Customer Service Bot. We spent more time on voice tuning than on architecture and it was worth it.
WhatsApp is a sales channel, not a support channel
Treating it as 'support that also sells' kept us small. Treating it as 'our primary sales channel that also supports' unlocked the 65% share.
Re-engagement on WhatsApp is 4-6× email
Reply rates of 38% on WhatsApp versus under 5% on email mean re-engagement is suddenly a real revenue lever again. Migrate the high-intent segment off email.
Human stylists are still essential
The AI handles 70% of conversations to satisfaction. The 30% where a human takes over are disproportionately high-AOV customers who become loyalists. Do not over-automate the top end.
ROI & payback
Investment
Low-six-figures one-time build + mid-four-figures monthly run cost (messaging + model + ops)
Payback period
Inside 3 months on the AOV and repeat-purchase lift
Year-1 ROI
Estimated 10-15× ROI; this engagement now drives the majority of the brand's net new revenue
“WhatsApp went from a customer-service channel to our most profitable sales channel. The AI stylist isn't a chatbot - it feels like a real shopping assistant, and our customers love it. Our return-customer rate tripled in 90 days. I genuinely cannot imagine running the brand without it now.”
Questions about this engagement
What made the AI stylist different from a chatbot?+
Brand-voice grounding, retrieval over the actual product catalogue and sizing guides, awareness of past purchases for personalisation, and an easy human-stylist handoff for the 30% of conversations that genuinely need one. The voice tuning made it feel like a stylist instead of customer service - that distinction is everything for a premium brand.
How is WhatsApp commerce kept compliant?+
Built entirely on the official WhatsApp Business Platform with template approvals, opted-in lists, per-message quality scoring, and automatic cooldown on patterns that draw complaints. Unofficial automation tools are policy-violating and lead to number bans - we use none of that.
How did average order value rise 28%?+
Two drivers: the AI stylist actively recommends outfit completions instead of letting customers buy single items, and the in-chat checkout is fast enough that customers don't drop a third item the way they would on a web cart. Mid-funnel friction is gone.
Did the human stylist team lose work?+
No - they took the 30% of conversations that are most valuable, focusing on first-time premium-segment customers and complex styling consultations. Their throughput on those high-touch interactions roughly doubled because the AI handled the simpler conversations.
What about returns and customer care after the sale?+
Returns initiation, order tracking, and post-purchase questions are handled in the same WhatsApp thread. Customers do not have to go to a web form or write an email. The single-thread experience is a meaningful loyalty driver - there is no app to install, no portal to log into.
Let's buildReady to put AI to work in your business?
Book a free 30-minute strategy call. We'll map your highest-impact automation opportunities and give you a clear roadmap - no obligation.