The quality of the chatbot’s recommendations depends directly on how well you train it. The chatbot relies on the shop’s product feed, but you can significantly improve its effectiveness by giving it additional knowledge. In this article, we’ll show you what to add to the chatbot’s knowledge base and how to do it. Chat works on its own as a channel, and the AI Sales Agent is the AI layer you attach to Chat or WhatsApp.
Where does the chatbot get its knowledge from?
The chatbot uses two sources:
Product feed – product data (names, descriptions, prices, attributes).
Knowledge base – information about your store, organized into topics. The first time you open the AI Sales Agent with an empty knowledge base, edrone reads your store website and adds what it finds.
The richer the knowledge base, the better the answers and recommendations.
Where is the knowledge base?
Go to AI Sales Agent → Chat, click the ⋮ menu at the top of the page and choose Knowledge Base. There you’ll find:
Topics customers ask about – 8 topics: Contact details, Delivery and shipping, Returns and exchanges, Payments, Terms and store rules, Privacy policy, Warranty and complaints, and Seasonal and current promotions. A counter shows how many topics are done (for example, "Topics completed: 3 of 8") and suggests the next step.
Knowledge sources – what edrone read on your store website. To read it again, click Fetch from store website. New information is added right away. Changes to existing entries wait as suggestions for you to review, so a new fetch doesn’t create duplicates.
Other information – entries that don’t fit any topic, like buying guides or product care tips.
There’s no entry limit: add as many entries as you need.
How to add knowledge?
Click Add knowledge and pick one of 4 ways:
Paste a page link – for example, terms and conditions, FAQ or your shipping page.
Upload a file – PDF or TXT, like your terms or a shipping price list.
Describe it in your own words – a few sentences are enough, and AI turns them into a structured entry.
Write the entry yourself – a title and content, straight into the knowledge base.
With the first three, AI prepares entry suggestions, and you review them before they go into the knowledge base.
What should you add to the knowledge base?
1. The 8 store topics
Start with the topics marked To add. They cover what customers ask about most:
Returns and complaints policy
Delivery times and costs
Available payment methods
Warranty terms
Contact details, terms, privacy and promotions
This lets the chatbot answer questions like "Can I return this product?" without sending customers to customer service.
2. Service descriptions
If you offer additional services (installation, customization, subscription), add their descriptions. The chatbot will be able to actively recommend them.
3. Buying guides
This is the most important part of training. Add materials such as:
"How to choose the right size of X?"
"Comparison of models A, B and C"
"Who do we recommend product X to?"
The chatbot turns this knowledge into personalized recommendations during the conversation.
4. Pages and files you already have
You don’t have to write everything from scratch. Paste links to pages like:
Your FAQ section
How-to articles on your blog
An "About Us" page with your brand description
Or upload PDF or TXT files, like product data sheets, user manuals and internal sales scripts.
How to ensure data quality?
Take care of your product feed:
Make sure product descriptions are detailed and complete.
Add product attributes (size, color, material, use). The chatbot uses them to filter recommendations.
Update the feed regularly so the chatbot doesn’t recommend unavailable products.
Keep the knowledge base up to date:
Changed your returns policy? Update the topic entry.
Added a new service? Add its description.
Seasonal changes to your range? Add the right guides.
After using Fetch from store website, review the suggested changes.
Test the result: on the AI Sales Agent → Chat page, the What your agent knows about your store panel shows how many products the agent can recommend and how many of the 8 topics are in the knowledge base. Use the chat preview to ask questions the way a customer would.
What to avoid?
❌ Don’t upload very long, unstructured documents. The chatbot handles divided, structured content better.
❌ Don’t leave the knowledge base empty. The chatbot will then be limited to information from the feed.
❌ Don’t forget to update. Out-of-date information can confuse the customer.
