Skip to main content

Conversational commerce: how to sell on WhatsApp with AI in 2026

Recover carts, recommend products, charge and run post-sale on WhatsApp with an AI agent. Tech stack, integrations with Shopify/Tiendanube/WooCommerce and real conversion metrics.

Online shopping setup with a laptop and shopping bags on a stylish couch.

E-commerce in LATAM has a structural problem: the average abandoned-cart rate hovers around 70% — and classic email recovery flows get back less than 5% of that. Meanwhile, 80% of shoppers use WhatsApp every day and open 98% of messages within 3 minutes.

Conversational commerce — selling, recommending, recovering and charging inside a WhatsApp conversation with AI — is the answer. This guide covers how to assemble the technical stack, which flows generate the most impact and what metrics to watch.

What is conversational commerce?

It's not just "having WhatsApp to answer questions". It's moving the entire funnel into a conversation:

  • Discovery: the customer asks and AI recommends products based on history and catalog.
  • Cart and checkout: the customer adds products in natural language and pays without leaving WhatsApp.
  • Post-sale: tracking, exchanges, returns, reviews — all in the same thread.
  • Re-engagement: segmented campaigns that reactivate cold buyers.

Four critical flows

1. Cart recovery

The customer adds 3 products to the cart on the website but doesn't complete the purchase. With a webhook from the storefront (Shopify, Tiendanube, WooCommerce, VTEX), AI fires a message 1 hour later:

"Hi Lucy, we noticed you left some items in your cart. Want help finishing the order? If you confirm now we can ship today."

With a direct payment button on WhatsApp. Typical conversion: 18-32% of carts recovered, vs 5-8% from traditional email.

2. Product recommender

The customer writes "looking for a road running shoe size 9 for hard pavement". AI:

  1. Queries the catalog in real time.
  2. Filters by category, size and use attribute.
  3. Returns 3 options with photo, price and stock.
  4. If the customer picks one, adds it to cart and offers checkout.

The difference vs a website search: AI can ask back, refine and handle objections.

3. Checkout inside WhatsApp

Three ways to charge:

  • WhatsApp Pay where available (Brazil, India). Payment inside the app.
  • Dynamic Stripe / MercadoPago / dLocal link generated by AI with the exact cart amount.
  • WhatsApp Business catalog with products loaded directly.

The dynamic-link flow is the most universal and works across LATAM.

4. Post-sale and support

Once the customer has bought, AI handles:

  • Shipping tracking (querying Andreani, Correo Argentino, OCA, FedEx).
  • Exchanges and returns per policy.
  • Automated reviews and NPS at day 7.
  • Cross-sell based on purchased product.

Tech stack

The pieces that need to connect:

Layer Tools
E-commerce platform Shopify, Tiendanube, WooCommerce, VTEX, Magento
Payments Stripe, MercadoPago, dLocal, Pago Fácil
Logistics Andreani, OCA, Correo Argentino, FedEx, Envíopack
WhatsApp Official Business API via Meta or BSP (360dialog, Twilio)
Conversational AI Platform like Keebai connecting the stack above
Analytics Meta Pixel + CAPI, GA4, customer data platform

Keebai connects the six with native connectors and bidirectional webhooks.

Metrics that matter

Not all metrics are equal. The five that reflect real ROI:

  • Cart recovery rate via WhatsApp: 18-32% is a good range.
  • Cost per acquisition from Lead Ads → conversation → sale: typically 30-50% lower than ads without conversation.
  • Conversion rate inside conversation: 8-15% (vs 1-3% typical web).
  • Average ticket on WhatsApp vs web: 10-25% higher due to guided upsell.
  • Post-sale NPS: automatically measurable at D+7.

Common mistakes and how to avoid them

Mistake 1: using WhatsApp as a spam channel. Mass messages without segmentation burn the database and get the number flagged. Solution: only send to genuine opt-ins and segment by behavior.

Mistake 2: not integrating real-time inventory. Recommending an out-of-stock product is worse than not recommending. Solution: direct webhook with the e-commerce platform.

Mistake 3: a bot that doesn't know when to escalate. If AI doesn't detect when to hand off to a human (serious complaint, off-catalog case, fraud), you lose the customer. Solution: configure explicit escalation rules with tags and priorities.

Mistake 4: ignoring the discovery phase. Focusing only on checkout leaves out customers who are comparison-shopping. Solution: conversational content for the awareness phase.

How to start

Suggested 14-day rollout:

  • Days 1-3: register WhatsApp Business API and integrate the e-commerce platform.
  • Days 4-7: configure payment connector and Meta-verified abandoned-cart templates.
  • Days 8-10: load catalog, exchange/return policies and brand FAQs.
  • Days 11-12: pilot with Meta Lead Ads traffic directed to WhatsApp.
  • Days 13-14: full rollout with cart, conversion and NPS dashboard.

Next steps

If your e-commerce is leaving 60-70% of carts unrecovered, see how Keebai automates recovery, recommendation and checkout on WhatsApp or book a demo and we'll set up a pilot with your catalog and current platform.

Back to blog

Give your brand a voice. 24 hours a day.

Book a 20-minute demo with our team. We'll show you how Keebai can transform your operation this week.