Not every question asked by a Discord member will have an answer available to Novus AI.
A member may ask about information that is missing from your Knowledge Base, outside the purpose of your assistant, or requires assistance from a human staff member.
Fallback Responses define how Novus AI should handle these situations.
What Is a Fallback Response?
A fallback response is used when Novus AI cannot provide an appropriate answer using the information and instructions available to it.
Instead of guessing or creating unsupported information, the assistant can provide a controlled response explaining that it does not currently have enough information to answer the question.
A good fallback response can also direct the member toward another source of assistance.
Why Fallback Responses Are Important
AI models are capable of generating responses even when complete information is unavailable.
For community-specific information, this can sometimes result in an answer that sounds reasonable but is not actually correct.
This is especially important for questions involving:
Prices
Rules
Policies
Event information
Account-specific concerns
Services
Rewards
Schedules
Procedures
Technical instructions
Other community-specific information
A properly configured fallback helps reduce unsupported or inaccurate responses.
Create a Clear Fallback Message
Your fallback message should be simple and helpful.
For example:
I don't have enough verified information to answer that question right now. Please contact our support team for further assistance.
You can customize the message to match your community's support process.
For example, you may direct members to:
A support channel
A ticket channel
A Discord role
A staff member
An FAQ or documentation page
Another appropriate resource
Tell the AI When to Use the Fallback
Your AI Instructions should clearly explain when Novus AI should stop attempting to answer and use the fallback behavior.
For example:
If the requested community-specific information is not available in the provided Knowledge Base or relevant instructions, do not guess, estimate, or invent an answer. Use the configured fallback response instead.
This gives the assistant a clear rule for handling missing information.
Fallback vs. Escalation
Fallback and escalation are related but serve slightly different purposes.
A fallback tells the member that Novus AI cannot provide a reliable answer.
An escalation directs the member toward someone or somewhere that can continue handling the concern.
A fallback response can include an escalation.
For example:
I don't have enough information to confirm this. Please create a ticket in the support channel so a staff member can assist you.
This allows Novus AI to safely stop the AI response while still helping the member find the next step.
Route-Specific Fallbacks
Different AI routes may require different fallback behavior.
For example, a general support route may direct members to your normal support team, while an account-related route may direct them to a private ticket channel.
When a specific topic has its own support procedure, place that escalation behavior inside the appropriate route instead of putting every possible procedure in your general AI Instructions.
This keeps your configuration organized and easier to maintain.
Avoid Overusing the Fallback
A fallback should protect against unsupported answers, but it should not replace information that belongs in your Knowledge Base.
If members repeatedly receive a fallback for the same legitimate question, consider adding the missing information to your Knowledge Base.
Common unanswered questions can help you identify gaps in your existing content.
Keep Fallback Information Updated
If your fallback directs members to a specific channel, role, staff team, or support process, keep that information current.
For example, if you change your support channel, update the corresponding fallback or route instructions.
Outdated escalation information can be just as confusing as an incorrect AI response.
Test Your Fallback
After configuring your fallback behavior, deliberately ask Novus AI questions that are not covered by your Knowledge Base.
Check whether the assistant:
Avoids inventing information
Clearly explains that the information is unavailable
Uses the expected fallback message
Directs the member to the correct support option
Preserves configured Discord channel or role mentions
Uses the appropriate language for the member
Also test questions that are covered by your Knowledge Base to make sure Novus AI does not use the fallback unnecessarily.
Example Configuration
A basic fallback instruction could be:
Use available Knowledge Base and route information when answering community-specific questions.
If the requested information is unavailable, do not guess, estimate, or invent an answer.
Inform the member that the information cannot currently be confirmed and direct them to the appropriate support option when one is provided.
You can expand this behavior through individual AI routes when different concerns require different escalation procedures.
Improving Your Fallback Over Time
Pay attention to the questions that frequently trigger fallback responses.
If a valid question appears repeatedly, it may indicate that your Knowledge Base needs additional information.
A useful improvement cycle is:
Identify frequently unanswered questions.
Verify the correct information.
Add or update the relevant Knowledge Base content.
Test the question again.
Confirm that Novus AI now provides the expected answer.
This allows your assistant to become more useful as your Knowledge Base grows.
Next Step
Your assistant now has instructions for handling situations where reliable information is unavailable.
The next step is understanding the factors that can affect AI usage and response limits.
Continue to AI Response Limits & Usage.