AI Behavior lets you shape how your AI represents your brand — from its tone and persona to the topics it never covers. It works alongside your Knowledge Base to keep every AI-generated reply and comment both accurate and on-brand.
Setting up the AI Behavior
- Click Manychat AI in the sidebar.
- Switch to the Behavior tab.
- If you are using the Behavior for the first time, you’ll see the Set AI Behavior button. Once you click it, the AI will scan the Instagram page connected to your Manychat account. It will use the information from the page’s bio and recent comments to learn more about your business and how you communicate with your contacts.
- After analyzing your Instagram page, the AI will pre-populate three fields: AI Role, Brand Voice & Persona, and Guardrails & Constraints.
- AI Role: The persona or job title the AI represents (e.g., "Marketing Coach and Growth Strategist"). This influences how the AI introduces itself and frames its responses.
- Brand Voice & Persona: The tone, communication style, and formatting rules the AI follows in every response (e.g., "professional, concise, no emojis").
- Guardrails & Constraints: Explicit rules for what the AI should never say or do (e.g., "Never provide legal or financial advice. Do not guarantee revenue results. Never disparage competitors."). When a user asks a question that violates a guardrail, the AI either redirects or declines to respond. Note that guardrails greatly reduce off-topic responses, but the AI may still occasionally answer outside those boundaries.
- Check the information in each field to ensure it is relevant and aligns with your brand voice. Feel free to adjust the text to your liking. Remember, it’s just a suggested starting point, not a final version.
- Once you are happy with the results, click Save to enable the AI Behavior.
- After the feature is enabled, check how AI performs in the Playground by clicking the Test AI button.
💡You can edit each field at any time by pressing the Edit button. Save again to apply the changes.
FAQ
What's the difference between Behavior and Knowledge?
Knowledge defines what your AI knows and can answer. Behavior defines how it communicates. Both work together — you need both for a well-configured AI
Why does the AI still sound wrong after I set Behavior?
Try being more specific in the Brand Voice field. Instead of 'be professional', try 'speak in a professional, direct tone. Keep replies under 3 sentences. Don't use emojis.' Then test in the Playground.
Why does AI ignore a Guardrail and answer something it shouldn't?
Guardrails significantly reduce the unwanted responses, but AI is not fully deterministic. Try making the instructions in the Guardrails field more specific. For example, instead of “don't give advice”, write “never provide specific legal or financial advice regarding business structures or taxes”.
Will Behavior affect my existing AI conversations?
Behavior applies to all future AI conversations immediately after you save. It does not retroactively change past conversations.