Objective
This article explains how to use Knowledge documents effectively in Personalization to get accurate, relevant AI-generated email content. It covers how document referencing works, how to write prompts that reliably pull from attached documents, and how to handle multiple documents and buyer persona targeting.
Applies To
Before You Begin
Before attaching documents to a Personalization, keep the following in mind:
- The AI does not automatically use content from attached documents. You must explicitly reference the topic you want in your prompt.
- Only attach documents that contain content relevant to the email you are generating. Unrelated documents can cause incorrect or off-topic content to be retrieved.
- Knowledge documents and Buyer Data are separate features. Buyer Data surfaces account-level signals (such as recent acquisitions or office openings). Knowledge documents let you supply your own content — product brochures, persona guides, battle cards, and so on.
- Admin needs to put documents into the Personalisation Agent channel first, to let them be accessible in Personalisation. Learn more
Procedure
1. Reference topics, not file names
You do not need to name the document file name in your prompt. Instead, call out the specific topic or type of content you want the AI to use from the document.
"Write an intro for the prospect using the attached doc."
"Write an intro for the prospect. Use the pain points and value proposition from the document context."
2. Use the right keyword to trigger document retrieval
The system uses keywords in your prompt to identify when it should fetch content from attached documents. The recommended keyword is document
document— for example, "from the document" or "as described in the document"document context— matches the internal section name the system uses and is the most explicit signal
3. Keep only relevant documents attached
Attach only documents whose content you actually want used in the email. Irrelevant documents introduce noise and can cause the AI to retrieve off-topic content. Fewer, more focused documents consistently produce better output.
4. Structure multi-document setups clearly
If you are using multiple documents — for example, separate files for different products or different buyer personas — follow these guidelines:
- Use one document per use case where possible. Separate product content from persona messaging, for example.
- If combining content in a single document, add clear section labels (for example, "For VP of Engineering:" or "For CEO:"). Without explicit labels, the AI cannot reliably infer which section applies to which context.
- Mirror your document labels in the prompt. If your document uses the heading "VP messaging," reference the same term in your instruction — for example, "use the VP messaging from the document."
5. Use variables for persona-based targeting
To generate different content based on the prospect's role, combine a resolved variable such as {{title}} with a matching label in your document.
"Using the {{title}} messaging from the document, write a two-sentence opening that speaks to their priorities."
When the sequence runs, {{title}} resolves to the prospect's actual title, giving the AI a clear signal to match the right section in the document.
Additional Information
After defining the prompt use Generate Preview feature to generated email and see how documents were used and which one.
Was this article helpful?
That’s Great!
Thank you for your feedback
Sorry! We couldn't be helpful
Thank you for your feedback
Feedback sent
We appreciate your effort and will try to fix the article