Uploading the client's business info

Every disappointing AI bot has the same root cause: nobody told it anything. The Knowledge Base — one of the bot's five modules — is where the client's business lives inside Chatfuel, and filling it well is the highest-leverage hour of the whole setup. A bot with a rich Knowledge Base quotes real prices, answers real objections, and sounds like the client's best salesperson. A bot with an empty one says "I'm not sure" until the lead leaves.

One split to keep straight: the Knowledge Base is what the AI knows; Automations are where and how it behaves. Business facts — company info, catalog, FAQ, working hours — go here. How the bot talks and acts goes into the automations. This page is about the knowing half.

For an agency this is a two-step job: collect the information from the client during onboarding, then load it into the client's company in Chatfuel.

What the Knowledge Base holds

• About the company — name, working hours, address, phone, email, website, payment options, and a short description. The identity facts the bot states with confidence.
• Catalog — the services or products, each with a name, description, and price. Keep it focused: under ~50 items, only what the client actually wants leads to ask about. The bot should never invent a price — if it's not in the Catalog, it's not quotable.
• FAQ — question–answer pairs for what leads actually ask: deposits, cancellations, delivery, parking, "do you work weekends". Short answers, 1–3 lines each, phrased the way customers phrase them.

Each piece has its place. Prices belong in the Catalog, not in the instructions; policies belong in the FAQ. Duplicating content across sections makes the bot inconsistent — one fact, one home.

What's not Knowledge Base material: the bot's role, tone, and house rules ("always offer a free consultation before quoting renovation work"). That's behavior, and it lives in Additional instructions on the automation — on Main bot for everywhere, or per source where a client wants a different voice for a different audience. Collect it in the same questionnaire; it just lands on the other page.

Step 1 — collect it from the client

Send the client a short questionnaire as part of onboarding, alongside the access checklist. What you need:

• Company basics: name, hours, address, phone, email, website, payment methods.
• Their top 5–15 services or products with current prices.
• The 10–25 questions their staff answers every single day — ask the person who runs their WhatsApp today; they know.
• House rules: what the bot should always offer, never promise, and when a human must take over.
• Tone: how they talk to customers — formal, friendly, emoji or no emoji.

The client answering in a voice memo is fine — you don't need a polished document, you need the facts.

Step 2 — load it into Chatfuel

Loading it is the same job through two equal doors. The Knowledge Base page in the left sidebar shows every section as editable fields — type the facts in, and changes take effect in new conversations immediately (here's a walkthrough of the Knowledge Base screens).

Or hand the raw material to the AI Co-Worker — by text, by forwarding a voice note, or by attaching files and images: a price list PDF, a brochure, a menu photo, a screenshot of the client's old FAQ page. It extracts the information, files everything into the right sections, and confirms what it saved:

• "Add my business hours, address, and top 10 FAQs to the knowledge base."
• "Add a new service to the catalog: gel manicure, $35, 45 minutes."
• "Update the business hours in the knowledge base."

The tone and house rules from your questionnaire go to the behavior side: paste them into Additional instructions on Main bot, or tell the Co-Worker how the bot should talk and it will write them in. And if you start with the "Set up WhatsApp" onboarding, its first step is exactly this: the Co-Worker shows what it already knows about the business and asks you to fill the gaps.

Test it like a lead would

Before the client ever sees the bot, interrogate it in the Test chat: ask for prices, opening hours, the tricky FAQ cases. Wherever it stumbles or improvises, the Knowledge Base has a gap — add the missing fact and ask again. Most "the bot sounds vague" complaints trace back to exactly this kind of gap; Troubleshooting the qualification bot covers the other failure modes. Once the answers hold up, teach the bot what to find out from every lead — that's lead qualification questions and logic.

Note: this isn't a one-time upload. Prices change, services come and go, new questions appear in real chats. Put a monthly Knowledge Base review into your client-maintenance routine: refresh the catalog, add FAQs for questions the bot answered badly, prune what's outdated. A stale bot quoting last season's prices costs the client trust — and costs you the retainer.