Knowledge Base

A chatbot knowledge base you completely control

Chatbot knowledge base and support documentation: centralized knowledge management, FAQ page curation and business rules that update the AI in real time.

In short

A chatbot knowledge base in StoreChart supplies the verified information your AI assistant relies on to answer customer queries. Built by scanning website pages, articles, services and FAQs, it is enriched with your business profile and rules, updating live bot answers within minutes.

Answers from your own knowledge base

Built from your site, edited by you

  • The agent answers from what's in the knowledge base, and points to WhatsApp or phone when the information is missing
  • Service policies, terms, operating hours and documentation are stored once and used in every relevant answer
  • Editing, deleting or switching off an item reaches the answers within about a minute
  • Rules you write, up to 20, shape what the agent says and what it won't promise

Knowledge base capabilities

What goes into the knowledge base and how you control it

FAQ entries
Questions and answers are generated from the scan, and you can add, edit or delete any of them.
Product information
Products with name, price, category, specs and page link are extracted from the scan, and you can edit each one. A product you switch off isn't suggested.
Business policies
Shipping, return and warranty pages are stored as articles, and the agent uses them in every relevant conversation, so answers stay consistent for every visitor.
Updates within a minute
Adding, editing, deleting or switching off an item reaches the agent's answers within about a minute, with nothing to publish.
Business profile and rules
Who you are, phone and WhatsApp, general instructions and up to 20 rules you can switch on and off, to steer answers toward your tone and policies.
Re-scanning
When products, prices or policies change on your site, run the scan again from the dashboard and review the results.

Build a knowledge base in 3 steps

From a site scan to ongoing fixes

1
Scan your site

Enter your site's address, and the scan extracts products, articles and FAQs from up to about 250 pages.

2
Review and organize

Edit or switch off items, sort questions into categories and tags, and add a business profile and rules.

3
Check and improve

Go through conversations the agent handled, spot questions where the answer was missing or inaccurate, and add or fix content in the knowledge base.

What to put in your chatbot knowledge base

The five layers of the knowledge base

The knowledge base behind your website chatbot has five layers, and each plays a different role in an answer:

  • Business profile: who you are, what you sell, shipping, returns, contact hours, phone and WhatsApp. The scan drafts it, and you correct it.
  • FAQs: a question and an answer, with a category, tags and linked pages on your site.
  • Products: name, price as shown on the site, category, description, specs, keywords and a page link.
  • Articles: longer texts by topic, usually the policy pages the scan picked up.
  • Rules and instructions: up to 20 rules you can switch on and off, plus a general instructions field for tone.

How the chatbot reads the knowledge base

The agent doesn't read the whole knowledge base for every question. A matching FAQ is used almost word for word, it gets a short list of products, and a very long article is trimmed. Rules go into every answer.

A few rules of thumb follow from that:

  • FAQ, not article: write your returns policy or shipping cost as an FAQ.
  • One article per topic: a focused article works better than one long page that covers everything.
  • Rules: the place for things that must never be forgotten, such as what not to promise and what not to discuss.
  • General instructions: for tone, such as how to address visitors, answer length and emoji.
  • Facts: put shipping cost and opening hours in the profile or an FAQ, so they're updated in one place.

A working order for a new knowledge base

Starting from an empty knowledge base or a partial scan? This is a practical order:

  1. Business profile: phone and WhatsApp first, because that's where the chat points people when it has no answer.
  2. FAQs: shipping, returns and payment, which visitors ask about before buying.
  3. Products: specs, keywords and a short description of who each one is for.
  4. Rules: last, once you've read a few real conversations.

Example: returning a sale item

Illustrative only. A fashion store sees in the conversations list that visitors ask "Can I return an item I bought on sale?", and the chat answers vaguely from the policy page.

  1. The owner adds an FAQ with a two-line answer.
  2. She gives it the category "Returns" and the tag "Sale", and links it to the sale page.
  3. She adds a rule: "Don't promise a cash refund, only store credit".
  4. She checks the wording in the preview, which shows the change straight away.
  5. Within about a minute, the chat on the site answers with the new wording.

Re-scanning and the StoreChart catalogue

The scan is a snapshot of the day it ran. The chat doesn't check against your store what exists right now.

  • New product on the site: doesn't exist for the chat until the next scan.
  • Removed product: can still appear in answers. Switch it off in the knowledge base until you scan again.
  • Seasonal FAQ: switch it off rather than delete it, so you can bring it back next season.
  • StoreChart catalogue: the chat can take categories and contact details from there, and products only when the scan found none at all. Products from there have no page link.

Who it's for and what's not included

  • Good fit: a store with policies customers ask about all the time, such as made-to-measure sizes, installation or delivery to remote areas.
  • Policies not written on the site: the business profile and rules are the only place the chat can learn them.
  • Live data: the knowledge base holds no stock or order status.
  • WhatsApp AI agent: builds its own knowledge. A fix in one doesn't update the other, so update important policies in both places.
  • Product recommendations: draw on the same product list, so every improvement to it improves them too.

Frequently asked questions about the knowledge base

Where does the content in the knowledge base come from?

Mostly from the site scan, which extracts products, articles and FAQs. You add the rest yourself: a business profile, rules and instructions.

Does the agent learn from my website's content?

Yes, through the scan. Review what was extracted before relying on it, and edit, delete or switch off any item. The scan doesn't re-run by itself, so run it again after you change your site.

Does an edit to the knowledge base take effect immediately?

Almost. Changes reach the agent's answers within about a minute, with no new version to publish.

Are there ready-made templates to start from?

There's no library of Q&A templates. The starting point is the scan, which creates FAQs from your own pages, and you edit them as needed.

Can I import questions and answers from a file?

Not from a CSV or Excel file. Add questions manually, or let the scan create them from your site's pages, for example from an existing FAQ page.

Can I organize entries by category?

Yes. Each question can have a category and tags, such as shipping, returns or product care, which makes items easy for your team to find and update.

Can I see which questions went unanswered?

There's no separate list of unanswered questions. The conversations list shows every chat, so you can see what visitors ask and add what's missing to the knowledge base.

An accurate knowledge base for your chatbot

Scan your site and refine the knowledge base over time