YundaDeskCustomers on WhatsApp do not like waiting, and they like scripted-sounding replies even less.
For cross-border sellers running support on WhatsApp, the two most common mistakes are simple. First, not distinguishing between the 24-hour customer service window, where you can reply freely, and the period after it closes, where you can only send pre-approved message templates. Second, writing scripts so rigid that one follow-up question exposes them. This post collects scripts you can use directly, explains how message templates work, and shows how to feed the same scripts into an AI customer service system so agents and AI answer with one consistent voice.
Free-form replies vs. message templates
WhatsApp Business has a hard rule: once a customer messages you, you get a 24-hour customer service window to reply freely. Once that window closes, you can only reach out using pre-approved Message Templates. This is not a rule YundaDesk sets — it is built into the WhatsApp platform itself.
That means you need two separate sets of scripts:
- Free-form replies inside the window — for handling inquiries, follow-ups, and explanations, where tone can flex.
- Message templates — for proactive outreach outside the window, like shipping updates, payment reminders, or order confirmations. Content must be submitted for approval in advance and cannot be edited on the fly.
WhatsApp Customer Service Scripts and Message Templates: put channel response times on one scale
Opening lines and identity checks
On WhatsApp, customers assume you already know who they are. Skip the generic “how can I help you” and confirm the order or account first.
Hi, this is [store name] support. Could you share your order number or the email you ordered with? I will pull up your order right away.
Thanks for reaching out. Let me check your order first — I will ask if I need more details, and anything involving a refund or compensation gets confirmed by a team member.
Hi there. If this is about shipping, send me your order number and I will check the status now.
Shipping and order scripts
Shipping questions are the highest-volume scenario on WhatsApp, and a good fit for AI to handle first — as long as the knowledge base has clear processing times and carrier policies, instead of leaving the AI to guess.
| Scenario | Script |
|---|---|
| Shipping timeline | Orders ship within the processing time listed on the product page, and I will send tracking once it ships. Share your order number and I will confirm the current status. |
| No tracking update in a while | I understand the delay is worrying. International transit sometimes has gaps in tracking updates that do not mean the package is lost. Let me check the latest record, and I will escalate if it is past the normal window. |
| Requesting a shipping status update | I have logged your request. I will confirm whether the order has entered the fulfillment queue — I will not skip any required steps, and I will update you as soon as there is news. |
| Address change request | If the order has not been handed to the carrier yet, we can usually help. Please send the new address and order number, and I will confirm with our team whether there is still time. |
| Marked delivered but not received | This needs verification with carrier records, so I cannot draw a conclusion here. Please share your delivery address and a screenshot of the delivery confirmation, and I will escalate this to a team member. |
Refund and compensation scripts
These are scenarios where AI should only reassure the customer and gather information — never approve anything on its own. This is a hard line, no matter how urgent the customer sounds.
I understand you want this refund resolved quickly. Refund requests go through manual review. Let me log your order number and issue now, and a team member will follow up — you will not need to repeat yourself.
For compensation, we first need to confirm what happened with the damage or loss — could you send photos or a video? I am escalating this to the team that handles these cases; it is not something I can decide directly.
Payment reminders and cart nudges (message templates)
Payment reminders count as proactive outreach outside the service window, so they typically need to go out as message templates — and they still need to follow outreach guardrails that avoid annoying customers: no high-frequency pinging, no duplicate nudges when the customer is already talking to you on another channel, and nothing during quiet hours.
Cart reminder
Hi {{1}}, the items you picked out at [store name] are still in your cart and stock is limited. Reply anytime if you need help checking out.
Order confirmation
Hi {{1}}, your order {{2}} is confirmed and should ship within {{3}}. We will send tracking as soon as it goes out.
Shipping update
Hi {{1}}, your order {{2}} has shipped and is expected to arrive within {{3}}. Tap the link for live tracking.
The variables ({{1}}, {{2}}, etc.) hold order numbers, names, and timeframes. Since template wording cannot change after approval, keep the tone neutral and factual rather than promotional.
Multilingual replies and one customer profile
Cross-border sellers get WhatsApp messages from customers across different countries, languages, and time zones. Ideally, the AI replies in whatever language the customer writes in, and if that same customer also messaged through the website widget, everything shows up in one customer profile instead of forcing them to re-introduce themselves on every channel. Once that context is in place, payment reminders and shipping updates can respect the customer’s actual time zone instead of landing at 3 a.m.
Turning these scripts into something AI can use
Scripts sitting in a spreadsheet do not help much. The better approach is organizing them by scenario in a knowledge base, with clear conditions for when they apply and when they do not, then letting AI handle the low-risk cases first — routing refunds, compensation, complaints, and threats of negative reviews to a human after reassuring the customer and collecting details. For a practical walkthrough of setting this up, see building a knowledge base that actually feeds your AI.
Agents will run into questions the knowledge base has not covered yet. When that happens, correcting the AI does not change its behavior instantly — it generates a suggested learning item that only takes effect once a manager reviews and approves it, and every change stays traceable, testable, and reversible. For how that learning loop works, see how AI customer service actually gets smarter over time.
The hard part of WhatsApp support was never the wording — it is remembering that the service window and template approval rules exist at all. Separate free-form replies from message templates, draw a hard line around refunds and compensation, and let agents and AI pull from the same knowledge base, and payment reminders and shipping updates can move fast without ever feeling intrusive.