### Answer Your Knowledge Base can include Q&A-style content like "What if the member says they are not interested" - that's a valid use case [(1) ]() . The Knowledge Base is designed to provide context and information that your agent can retrieve during calls . However, there's an important distinction: **Use KB for**: Supporting information, FAQs, product details, policies - content that provides factual context **Use the prompt for**: Agent instructions, behavioral guidelines, conversation flow logic Based on community experience, when you link a Knowledge Base to an agent, it retrieves information for every reply, which can sometimes override your structured conversation flow [(2) ]() . If you're using multi-prompt or conversation flow agents and want deterministic behavior for specific scenarios (like handling objections) , you may want to put those instructions in your prompt instead [(2) ]() [(3) ]() . To prevent the KB from being invoked when you want the agent to follow specific node transitions, you can restrict Knowledge Base usage in your general prompt [(2) ]() . For example, you could add instructions telling the agent to only use the KB for certain types of questions, not for handling common objections or conversation flow decisions. **Recommendation**: Put objection handling instructions ("what if they say they're not interested") in your agent's prompt for more deterministic control. Reserve the KB for factual information that's too lengthy for the prompt [(1) ]() .

​ To learn more, see the sources I considered: [[1] Knowledge Base](https://docs.retellai.com/build/knowledge-base) [[2] Most of the times KB is invoked and the agent does not follow the node transitions](https://discord.com/channels/1186424376300679209/1407282116802252892) [[3] Help - In Multi Prompt Mode, KB is invoked when not required](https://discord.com/channels/1186424376300679209/1385077075865632809) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.

### Answer Yes, putting objection handling directly in your prompt will increase token usage compared to using the Knowledge Base. However, there's a trade-off to consider: **Token costs with KB**: Using Knowledge Base costs $0.005 per minute of calls that have knowledge base enabled [(1) ]() . When KB is invoked, it retrieves chunks (default: 3 chunks with 0.60 similarity threshold) and appends them to the LLM prompt under "Related Knowledge Base Contexts" . **Prompt engineering approach**: You can use sectional prompts to keep your instructions organized and focused [(2) ]() . For objection handling specifically, you don't need to write lengthy responses - just concise guidelines . **Recommended hybrid approach**: - Put brief objection handling guidelines in your prompt (e.g., "If caller says not interested: acknowledge and ask if there's a better time") - Keep detailed product information, policies, and FAQs in the KB [(1) ]() For example, instead of full Q&A pairs in your prompt, use concise style guardrails [(2) ]() :

## Objection Handling
If the caller says they're not interested: "I understand. Is there anything specific..."
If the caller is frustrated: "I hear your frustration, let me help resolve this..."

This keeps your prompt lean while maintaining deterministic control over conversation flow. The token increase from brief objection guidelines is typically less impactful than the unpredictability of KB overriding your node transitions [(3) ]() [(4) ]() .

​ To learn more, see the sources I considered: [[1] Knowledge Base](https://docs.retellai.com/build/knowledge-base) [[2] Prompt Engineering Guide](https://docs.retellai.com/build/prompt-engineering-guide) [[3] Help - In Multi Prompt Mode, KB is invoked when not required](https://discord.com/channels/1186424376300679209/1385077075865632809) [[4] Most of the times KB is invoked and the agent does not follow the node transitions](https://discord.com/channels/1186424376300679209/1407282116802252892) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.