Template · Knowledge operations
WhatsApp Business Knowledge Base Template for AI Replies
An AI reply is only as reliable as the business facts behind it. This worksheet turns scattered product, policy and escalation knowledge into a compact source your team can review.
Quick answer
What should a WhatsApp business knowledge base include?
Include your official business identity, products and services, prices or pricing rules, minimum quantities, lead times, payment methods, shipping coverage, return policy, exclusions and escalation rules. Add an owner, last-verified date and source link for facts that can change.
Keep each entry concise and operational. The goal is not to copy your entire website; it is to give an AI reply tool enough verified context to answer common questions and enough boundaries to avoid unsupported promises.
“Standard lead time is 10–15 business days after artwork approval” is a fact. “If the requested date is earlier, escalate instead of promising delivery” is an instruction. A safe knowledge base needs both.
Original resource
Download the WhatsApp business knowledge-base worksheet
The Markdown worksheet includes product, payment, delivery, return and escalation sections plus the exact MSG.AI field limits. Complete it with the people who own those policies, then copy the approved content into the extension.
Download the knowledge-base template
This is a copy-and-paste planning worksheet, not an automatic importer. That distinction keeps a human review step between source documents and customer-facing AI context.
Recommended schema
Use four compact knowledge blocks
| Block | Include | Avoid |
|---|---|---|
| Business | Name, scope, regions, hours, languages, official links | Long origin stories that do not affect a reply |
| Products/services | Best fit, inclusions, exclusions, price rule, MOQ, lead time | Unverified claims and obsolete catalog text |
| Policies | Payment, delivery, return, refund and privacy rules | Legal paraphrases that conflict with the source policy |
| Escalation | Triggers, owner, handoff path and expected update time | Instructions to guess when information is missing |
Write facts as answer-ready sentences
Prefer “We ship to Singapore and Malaysia; other destinations require a freight check” over “Shipping: international.” Include units, currency, time zones and the event that starts a lead-time estimate.
Filled example
Example for a B2B office-chair supplier
Business introduction
Acme Seating supplies ergonomic office chairs to commercial buyers in Singapore and Malaysia. Support hours are Monday–Friday, 09:00–18:00 Singapore time. English and Chinese service are available.
Ergo Pro chair: commercial ergonomic chair with adjustable lumbar support and armrests. Minimum order: 20 units. Standard lead time: 10–15 business days after deposit and final color confirmation. Delivery outside Singapore and Malaysia requires a freight quote.
Escalation rule: do not promise stock, custom colors, delivery dates, discounts or warranty coverage unless the current order record confirms them. Escalate damage claims and orders above 200 units to a sales manager.
This example is fictional and illustrates structure only. Replace every fact with approved information from your business.
MSG.AI implementation
Fit the knowledge into the product’s field limits
- Business name: maximum 80 characters.
- Business introduction: maximum 1,500 characters.
- Up to 20 products or services.
- Each service name: maximum 60 characters.
- Each service description: maximum 500 characters.
- Additional information: maximum 2,000 characters.
- The extension composes a knowledge prompt within a 6,000-character budget.
Prioritize high-frequency questions and costly mistakes. If the source material exceeds the limits, shorten descriptions, group similar offerings and link agents to the authoritative system for volatile details.
Knowledge operations
Assign owners and review dates
- 01
Collect authoritative sources
Use current catalog, pricing, fulfillment and policy documents. Resolve conflicts before summarizing them.
- 02
Name an owner
Assign responsibility by section: sales owns product rules, operations owns delivery, and support owns escalation language.
- 03
Record verification
Add the source, last-checked date and next review. Trigger an immediate update when a policy or product changes.
- 04
Retest real questions
Use anonymized common and difficult queries to check whether replies stay inside the approved facts.
Quality check
Test normal questions, edge cases and missing facts
- Direct: “What is your minimum order?” The answer should name the correct product and quantity.
- Conditional: “Can 200 units arrive next Friday?” The reply should state conditions or escalate, not invent availability.
- Out of scope: “Can you guarantee this medical result?” The reply should refuse unsupported claims and route appropriately.
- Conflict: provide a customer statement that contradicts the knowledge base and check that the draft asks for verification.
- Missing: ask about a product not listed; the assistant should acknowledge uncertainty rather than fabricate details.
MSG.AI stores the knowledge configuration through extension storage and includes it only when the feature is enabled and contains usable content. Its AI Reply workflow creates an editable draft from recent chat context. See the step-by-step product guide and the AI assistant comparison.
Turn approved facts into faster drafts
Add your business knowledge to MSG.AI
Start small, test real questions and keep a person responsible for the final message.
Frequently asked questions
FAQ
What is a WhatsApp business knowledge base?
It is a curated collection of business facts and response rules used to answer customer questions. It can include products, prices, service areas, delivery, returns and escalation boundaries. It should be verified, concise and regularly maintained.
Is this knowledge-base template free?
Yes. The Markdown worksheet is free to download and adapt. It is a planning document rather than legal advice or an automatic import file. Review the completed content with the owners of each business policy.
What information should not go into an AI knowledge base?
Avoid passwords, payment-card data, unnecessary personal information, secrets, unsupported marketing claims and facts without a trusted owner. Keep customer-specific or rapidly changing data in an authoritative system rather than a static knowledge block.
How often should the knowledge base be updated?
Update it immediately when a product, price, region or policy changes, and review it on a fixed schedule. High-risk sections such as payments, refunds and delivery promises deserve more frequent checks and a named owner.
Does a knowledge base prevent every incorrect AI reply?
No. It improves grounding but cannot guarantee correctness. Source material can be incomplete, outdated or ambiguous, and a model can misinterpret context. Test difficult cases and require human review for customer-facing commitments.
How does MSG.AI use the knowledge base?
When the feature is enabled and populated, MSG.AI adds the saved business context to an AI Reply request alongside recent text from the active chat. It then places the generated reply in the composer for the operator to review.
MSG.AI field limits and behavior were verified against the codebase on September 1, 2026. The filled business example is fictional and provided only to demonstrate structure.