E-commerce
June 28, 2026
You have an AI chatbot on your Shopify store. Your customers also contact you on WhatsApp. The question is not "should we have a WhatsApp bot?" but "should we connect the same AI brain to both channels?"
Gartner estimates that more than 60% of WhatsApp Business Platform conversations will involve an AI agent by the end of 2026, up from 28% in 2025 (Kanal, WhatsApp statistics 2026). eGrow notes that a high-performing WhatsApp AI agent targets a 70%+ autonomous resolution rate and latency under 3 seconds (eGrow, WhatsApp AI agent playbook 2026).
This guide #132 covers AI chatbot + WhatsApp integration: architecture, dataset, handoff, limitations, and deployment. Separate from WhatsApp support ops (#131) and the comparison WhatsApp vs onsite chat (#18).
Summary
Should you connect an AI chatbot and WhatsApp for your store?
Connecting an AI chatbot and WhatsApp makes sense when your messaging volume justifies the API investment and you want a consistent cross-channel experience.
Yes, if...
Volume: 50+ WhatsApp conversations/month or predictable seasonal peaks
Messaging-first markets: LATAM export, Spain, Portugal, Africa
Mature website bot: validated corpus, 50%+ deflection, tested handoff
WhatsApp API: BSP + helpdesk already planned
Not yet, if...
Website bot is unstable (hallucinations, CSAT < 4), team has no WhatsApp process (#131), Business App alone with 5 messages/week. Consolidate your onsite bot first.
What "connecting" means
Not copy-pasting the website widget into WhatsApp. It is sharing your corpus, intents, Shopify data, handoff rules, and brand voice via an omnichannel orchestration layer.
What architecture for an omnichannel AI bot?
An effective Shopify WhatsApp chatbot architecture centralizes intelligence, not interfaces.
Recommended Layers
Channels: website widget, WhatsApp API, Instagram DM (optional)
AI Orchestrator: intent classification, RAG corpus, Shopify tools
Data: unified orders, catalog, policies, conversations
Helpdesk: Gorgias/Zendesk handoff + customer timeline
Anti-pattern: two separate bots
Website bot Qstomy + BSP keyword WhatsApp bot = contradictory answers, double maintenance, fragmented CSAT. One brain, multiple surfaces (ZynfoAI, 2026 omnichannel orchestration).
Bi-directional Shopify Sync
Order lookup, fulfillment status, tracking, real-time stock. Fulfillment webhooks trigger utility template + bot context if the customer responds.
How do I share the corpus between the website and WhatsApp?
The WhatsApp bot shared corpus guarantees the same policy responses, deadlines, and feedback.
Sources to synchronize
Terms and conditions hub: returns, delivery, warranty
Validated agent macros: field-tested answers
Fit notes and PDP specs: if pre-purchase WhatsApp bot
Support knowledge base: a single source of truth
WhatsApp format adaptation
Same content, different form: shorter messages (3-4 lines max), native Meta reply buttons, interactive lists for intent menu. No 800-word HTML blocks pasted into chat.
Clean terms and conditions hub data, bot brand voice (#125).
Which intents should you prioritize for automation on WhatsApp?
Prioritize high-volume and low-risk WhatsApp bot intents.
Phase 1 (week 1-2)
WISMO: status + tracking link
Returns: portal link + refund timeframe
Delivery time: ETA by zone
Support hours: + handoff if urgent
Phase 2 (month 2)
Pre-purchase sizing, product compatibility, promo conditions, abandoned cart (marketing opt-in). Measure each intent 14 days before adding the next one: unmatched > 10% on an intent = corpus or handoff to be corrected. See choose questions to automate.
Do not automate
Disputes, chargebacks, regulated products, exceptional refunds, VIP negotiation. Same red list as the website bot (chatbot limitations (#124)).
How to manage the 24-hour window with an AI bot?
The 24-hour WhatsApp window changes the bot behavior compared to the always-open website widget.
During the open window
Customer wrote: free-text AI responses, no template, target latency < 3 s. Interactive buttons to guide intent. Rich media: product photo, PDF size guide.
Closed window
The bot cannot reply in free-form if the customer has not reinitiated the conversation. Proactive follow-up = Meta-approved utility or marketing template. Do not send AI-generated responses outside the window: rejection or non-delivery.
Utility template + bot
Shipping flow: order_shipped template → customer replies "where is the package?" → window reopened → WISMO bot. Ops details: WhatsApp support (#131).
How of to organize the bot-to-agent handoff on WhatsApp?
The WhatsApp bot-to-human handoff must be smoother than with a widget: the customer does not switch to another tab.
Escalation triggers
Confidence < 85 % (95 % if regulated)
Negative sentiment: 2 consecutive messages
Keywords: agent, exceptional refund, lawyer
3 intent failures: loop without resolution
Explicit request: talk to someone
Handoff UX
Message: "I am putting you in touch with [First Name], who can already see your order #XXX." Full transcript + order context in Gorgias. Agent takes over in the same WhatsApp thread, not a parallel email.
What is the technical stack to connect an AI bot to WhatsApp?
Stack Shopify DTC WhatsApp AI bot integration 2026.
Components
Shopify: source for orders, products, customers
BSP WhatsApp: 360dialog, Twilio, MessageBird
AI Layer: Qstomy or agent platform + RAG
Helpdesk: Gorgias WhatsApp channel
Setup in 7 steps
Verify Meta Business Manager + display name
Connect BSP to WhatsApp API number
Webhook BSP → AI orchestrator
Sync website bot knowledge base to WhatsApp
Connect Shopify order lookup tools
Configure Gorgias handoff
Test 20 gold scenarios before going live
The standard WhatsApp Business App alone does not support this stack (w.app, e-commerce WhatsApp chatbot 2026). Realistic SMB budget: BSP €50-150/month + AI layer + existing helpdesk. Allow 2-4 weeks for setup, including Meta display name validation and utility templates.
What are the UX differences between a website chat and WhatsApp?
Adapt the WhatsApp bot UX to the channel, don't duplicate the website widget.
Website widget
PDP page context, visible cart, browsing session. Longer answers accepted. Proactive trigger based on variation hesitation.
Order ID or phone lookup context required. Short, async messages. Customer sends photos (defect, size label). No pop-up: the customer has chosen the channel.
Mandatory consistency
Same return policy, same processing times, same tone. The customer must not get two contradictory answers depending on the channel. Monthly mystery shopping: same WISMO question on website vs. WhatsApp.
What are the specific limits and risks of the WhatsApp bot?
WhatsApp bot limits: technical, compliance, Meta reputation.
Hallucinations and quality rating
Policy error feedback on WhatsApp = viral screenshot + reporting. Strict corpus, confidence threshold, no free generation on refund amounts. See hallucination prevention (#123).
Cost and latency
Every API message counts. Slow bot (> 5 s) = conversation abandonment. Cache frequent WISMO responses. Out-of-window templates are billed.
rules gdpr and opt-in
Proactive bot marketing (abandoned cart): explicit opt-in. Client-initiated support: contractual basis. Documented conversation retention.
How to measure WhatsApp bot integration?
WhatsApp chatbot integration KPIs: quality, cost, omni-channel consistency.
Essential metrics
Autonomous resolution rate: target 60-70% phase 1
Latency P95: < 3 s bot response
Handoff rate: monthly trend by intent
WhatsApp channel CSAT: compare vs website
Cross-channel consistency: mystery shop audit
Cost per resolved conversation: API + AI + agent share
Improvement loop
Weekly export of unmatched WhatsApp conversations → brief corpus update → redeploy within 7 days. Compare website vs WhatsApp deflection: gap > 15 points = intent poorly adapted to the channel. eGrow recommends a weekly review of unresolved conversations to feed the RAG (eGrow, playbook 2026).
How does Qstomy connect AI chatbot and WhatsApp?
Qstomy unifies website AI chatbot and WhatsApp: one corpus, one intent logic, one handoff.
Integration Features
Single corpus: sync hub conditions, macros, PDP chunks
Channel adapter: WhatsApp short format + Meta buttons
Shopify tools: order lookup, tracking, return link
Unified handoff: cross-channel Gorgias transcript
Confidence routing: automatic escalation with configurable threshold
Omnichannel analytics: deflection by channel and intent
Quantified DTC Cosmetics Scenario
FR cosmetics brand + ES export: website bot 54% deflection. WhatsApp API extension month 2: 280 conversations/month, 61% autonomous bot resolution (shared corpus), median latency 2.1s. Response discrepancy return policy website vs WhatsApp: 0% after mystery shopping 20 pairs. Marginal WhatsApp cost: +€190/month API vs -€420 agent time saved. WhatsApp bot CSAT 4.4/5.
Explore AI support, AI sales agent, Shopify, request a demo.
What operational playbooks are used to connect the two?
Playbook 1: website bot audit readiness
Before WhatsApp: website bot CSAT > 4.2, deflection > 45%, unmatched < 15% volume, handoff tested. If not: 30-day website consolidation first. Timeframe: 2-hour audit.
Playbook 2: map 10 shared intents
List the top 10 website AND WhatsApp questions. Validate that the corpus covers 100%. Write a short WhatsApp version (max 320 char.) per intent. Timeframe: 1 day.
Playbook 3: connect BSP + test webhook
Test number, 20 gold scenarios (WISMO, return, handoff, out-of-window). Validate latency < 3 s and button formatting. Timeframe: 2-3 tech days.
Playbook 4: cross-channel mystery shopping
Same test customer: return question on widget + WhatsApp. Identical policy? Consistent tone? Address discrepancies within 48 h. Monthly recurring.
Playbook 5: progressive go-live
Week 1: WISMO + WhatsApp return only. Week 2: + delivery times. Week 3: + pre-purchase if volume warrants. Review unmatched every Friday.
Useful links
Connecting an AI chatbot and WhatsApp is not a channel project: it is an omnichannel project. One corpus, one voice, one handoff. The WhatsApp channel amplifies what already works on the website.
Enzo
June 28, 2026
