Understanding AI Behavior
Whisper's AI is powerful, but it's important to understand how it learns and why it behaves the way it does.
Updated: 29 May 2026
Whisper's AI is powerful, but it's important to understand how it learns and why it behaves the way it does.
How Whisper Learns#
Whisper builds its knowledge base through several methods:
Onboarding Research
When you first create your organization, Whisper performs a deep crawl of your provided e-shop URLs to understand your catalog, policies, and brand tone.
Feedback Notes
You can manually add specific instructions in the Feedback & Training section of your WhisperBot settings.
File Uploads
You can upload reference documents (PDFs, text files, etc.) for the AI to learn from.
Conversation History
The AI learns from resolved threads, adapting to how your operators handle different situations.
Why Responses May Be Inaccurate#
If the AI provides an inaccurate response or draft, it usually means it lacks the correct context or specific instructions.
Modifying Suggestions with Copilot#
If a drafted email response isn't quite right, you don't need to rewrite it manually. Use the Copilot prompt at the bottom of the open thread to instruct the AI on how to improve it (e.g., "Make this shorter and more polite").
Re-analyzing a Conversation#
If you've recently updated your integrations or customer data, the current AI suggestions might be based on outdated information. Click the Re-analyze button in the context panel to force the AI to fetch the latest data and generate new suggestions.
