Writing knowledge manually is the best option when you want full control over what your AI agent says. Use it for policies, FAQs, special instructions, service explanations, or any content that is easier to write directly than upload from a file or webpage.
Start from All Knowledge
Go to AI Agent → Knowledge Base → All Knowledge. Find the knowledge item you want to add and select Add.
In the Add [Knowledge Name] sheet, choose Write Manually. This opens the text editor for that knowledge item.
Use the editor
The editor lets you write structured content with rich formatting. Keep the content clear and customer-facing. The goal is not to write a script for the agent; it is to give the agent accurate information it can use in conversation.
Good manually written knowledge usually includes:
- what customers commonly ask
- what the answer should be
- exceptions or conditions
- exact wording for important policies
- anything staff would normally explain over the phone
💡 Tip: Write in plain English. If a staff member would not understand the wording quickly, the content is probably too complicated for a customer conversation too.
Use a template if available
Some knowledge types include a starter template. In the editor, choose Load template to bring it in. If you already typed content, Speako asks you to confirm because loading the template replaces the current draft.
You can also choose Start blank if you prefer to write from scratch.
Save your content
When adding new text knowledge, choose Save. The item then appears under Your Knowledge as configured.
If you open an existing configured item, you can preview it, choose Edit, then use Save Only or Save & Publish depending on whether you want to apply the change to the live agent right away.
Keep it maintainable
Short sections with clear headings work better than one long block of text. If a policy changes often, keep that section easy to find so your team can update it quickly later.
Before saving, read the content as if you were a customer asking over the phone. If the answer depends on a condition, include that condition directly. If staff would normally clarify something, write that clarification into the knowledge so the agent does not have to guess.