Your AI assistant on WhatsApp can answer customer questions on its own, day or night. But it will only ever answer with what it actually knows about your business. If you have not told it your delivery time or your return policy, it cannot invent a good answer — and you would not want it to. The Knowledge page is where you teach it.
Think of it as the employee handbook for your assistant: the things a new staff member would need to know before answering customers.
How the AI Uses What You Write
When a customer asks a question, Whapick searches your knowledge and your product catalogue, finds the entries that best match, and the AI writes a short, friendly reply grounded only in what it found.
Two things follow from that, and they are worth remembering:
- It answers from your words, not from the internet. It will not make up a policy or a price. If the answer is not in your knowledge or catalogue, it will say so honestly and offer to connect the customer with your team — which is far better than a confident wrong answer.
- Anything you have not written, it does not know. The quality of its answers is directly the quality of what you put here. This is the single highest-value thing you can spend time on.
Product prices and availability are handled by your product catalogue, so you do not need to list them here. Knowledge is for everything else: how you work, what you charge for, where you deliver, what your rules are.
The Types of Knowledge
Every entry has a type. Picking the right one helps the AI understand the entry’s role and lets you filter your list later:
- FAQ — questions customers keep asking.
- Policies — returns, refunds, warranties, privacy.
- Delivery Info — areas covered, timeframes, fees.
- Payment Info — how and when customers can pay.
- Pricing — what things cost and any rules around it.
- Company Info — who you are, where you are, what you do.
- Contact Info — how to reach a human.
- Sales Info — how ordering works, minimum quantities, lead times.
- Instructions — how you want the assistant to behave (tone, what to always mention).
- General — anything that does not fit the others.
Nothing breaks if you pick the wrong type — it is a filing label, not a rule. But using them well makes a big knowledge base navigable.
Adding One Entry
Click Add Knowledge and fill in three things:
- Type — from the list above.
- Title — a short label, like “Return Policy” or “Delivery Areas”. This is for your own navigation; it also helps the AI find the entry.
- Content — the actual answer, written the way you would say it to a customer.
There is also an optional Keywords field — comma-separated words a customer might use, like return, refund, exchange. Adding a few helps match common phrasings the title might miss, but it is a helper, not a requirement.
Adding Many at Once
If you are setting up for the first time, doing this one entry at a time is slow. Click Bulk Upload and you get the same three convenient routes as contacts:
- Upload a file — CSV, JSON or Excel, with columns
type,title,content,keywords. Rows missing a type, title or content are skipped, and the page shows you the results. - Paste data — drop in CSV or JSON directly.
- AI Parse — paste unstructured text (your website’s about page, an old FAQ, a notes file) and the AI turns it into structured knowledge entries. You review and edit them before saving, and it tells you how many AI credits it used.
Bulk import is the fastest way to go from “the AI knows nothing” to “the AI can handle my common questions” in one sitting.
Switching Entries On and Off
Every entry has an enabled toggle. This is more useful than it first looks: when a policy is temporarily different — a holiday closure, a paused delivery area — you can switch the entry off instead of deleting it, and switch it back on later. Deleting is for entries you no longer want at all; disabling is for things that are just not true this week.
Finding and Editing
Use the type filter at the top to narrow the list, and the search to find an entry by name. Click an entry to edit its title, content, keywords or type. Everything is editable, forever — knowledge is never locked.
What Good Knowledge Looks Like
The AI answers in short, natural WhatsApp replies. Write your entries the same way:
- Be specific. “We deliver within Lagos in 24–48 hours; ₦1,500 for orders under ₦20,000 and free above that” beats “We deliver fast”.
- Write one entry per topic. A separate entry for delivery, one for returns, one for payment. It is easier to find, easier to edit, and it matches customer questions more precisely.
- Answer the question. Lead with the answer, then the detail. “Yes — you can exchange within 7 days if the item is unused.”
- Keep it current. An outdated entry is worse than none, because the AI will state it confidently. Review your knowledge whenever something changes.
And do not over-write. A handful of clear, accurate entries beats a long document. The assistant only needs the facts.
Common Questions
Does my knowledge have a size limit?
Entries are stored in your catalogue space, which is counted towards your plan. In practice a normal knowledge base is tiny — text is small. If you are near a plan limit you will be told before anything is saved.
What is the difference between knowledge and my products?
Knowledge is for how your business works (policies, delivery, FAQs). Products hold names, prices and availability. The AI searches both, so put each fact in the right place.
What if the AI still answers something wrong?
It almost always means the fact was missing, vague, or out of date. Find the relevant entry — or add one — and make it explicit. That is the whole fix.
Can I write instructions about tone?
Yes — that is the Instructions type. Use it for behaviour you want the assistant to follow, like always mentioning free delivery over a certain amount.
Conclusion
The Knowledge Base is how your AI stops being a generic chatbot and starts sounding like your business. Add your FAQs, policies, delivery and payment details — by hand for a few, by bulk import for many — and keep them current. The assistant will then answer your customers correctly and consistently, and hand over to a human only when it genuinely does not know.