Skip to main content

Command Palette

Search for a command to run...

How to Use Restricted Topics in ZazzyAgent AI Agents

Updated
View as Markdown

An AI Agent can understand a very wide range of subjects.

Your business usually doesn't want it discussing all of them.

For example, a customer support Agent may need to answer:

  • Product questions

  • Pricing

  • Delivery

  • Support

  • Booking

It doesn't necessarily need to discuss unrelated subjects.

Restricted Topics help define subjects that your AI Agent should avoid.

What are Restricted Topics?

Restricted Topics are subjects that you don't want the AI Agent to engage with as a normal business conversation.

For example:

Competitor-specific comparisons

or:

Questions unrelated to the company's services.

Why use Restricted Topics?

They help keep the Agent within its intended role.

For example:

A real estate Agent should discuss properties, viewings and company information.

You may not want it giving unrelated advice about other subjects.

Open AI Configuration

Open the AI Configuration area in ZazzyAgent.

Locate:

Restricted Topics

Add the topics that should be outside the Agent's intended scope.

Write meaningful restrictions

Good:

Competitor pricing and direct competitor comparisons.

Legal advice outside the company's services.

Unrelated financial advice.

Poor:

Everything else.

The more clearly you define the topic, the easier it is for the Agent to stay focused.

Add a predefined response

ZazzyAgent allows you to define the response used when the customer asks about a restricted subject.

For example:

I can help with questions about our services, pricing and bookings, but I can't assist with that topic.

Keep the message simple.

Restricted Topics vs Knowledge

These perform different jobs.

Knowledge

defines information the Agent can use.

Restricted Topics

define subjects the Agent should avoid.

For example:

Knowledge

Our refund policy is 30 days.

Restricted Topic

Legal interpretation of a third-party dispute.

Restricted Topics vs System Prompt

Your System Prompt defines the Agent's overall behavior.

Restricted Topics provide a specific mechanism for keeping certain subjects outside the Agent's normal scope.

They can work together.

For example:

System Prompt

You are a customer support assistant.

Restricted Topics

Political discussions, competitor disputes and unrelated advice.

Example: Real Estate Agent

The Agent can answer:

What properties do you have?

How much is this apartment?

Can I schedule a viewing?

Restricted subjects might include:

Legal advice beyond approved company information.

Guarantees about future property values.

Example: Clinic Assistant

The Agent can answer:

What services do you offer?

What are your hours?

Can I book an appointment?

The Agent shouldn't make unsupported medical diagnoses.

The exact restrictions should reflect the clinic's approved use case.

Example: Financial business

The Agent may provide approved business information.

But it should not automatically move into personalized financial advice simply because a customer asks for it.

Use appropriate human escalation for situations that require qualified professional review.

Use a helpful fallback

A restriction doesn't have to end the conversation.

For example:

I can help with information about our products, services and pricing. What would you like to know?

This redirects the customer to an area the Agent can support.

Restricted Topic vs Keyword

A keyword is a specific word or phrase.

A topic is a broader concept.

For example:

Keyword

Bitcoin

Topic

Cryptocurrency investment advice

Topics are useful when you want to cover a broader category of conversation.

Don't create an enormous restriction list

You don't need to list every possible sentence.

Instead, define meaningful categories.

For example:

Competitor comparisons and competitor pricing.

This is easier to maintain than dozens of nearly identical phrases.

Test variations

Suppose you restrict:

Competitor pricing.

Don't test only:

What is your competitor's price?

Also test:

Is your competitor cheaper?

Compare your pricing with theirs.

How much would I save by using them?

The restriction should address the actual topic, not only one exact sentence.

Test normal business questions

Make sure valid questions still work.

Test:

What are your prices?

What services do you offer?

How do I book?

A restriction that blocks legitimate customer questions is too broad.

Human handoff for restricted subjects

Some subjects may be better handled by a human rather than a generic restricted response.

Your overall Agent instructions can define when an issue should be escalated.

See How to Hand Off Instagram AI Conversations to a Human in ZazzyAgent.

Don't reveal internal rules

Avoid telling customers:

That topic appears in our Restricted Topics configuration.

Instead, simply redirect the conversation naturally.

Example

Customer:

Which of your competitors is best?

A suitable response could be:

I can explain our own products, services and pricing. I can also help you understand which of our options fits your needs.

The Agent stays within its intended business role.

Common problems

AI still discusses restricted subjects

Review the topic definition and System Prompt.

Normal questions are blocked

Your restriction may be too broad.

Response feels unhelpful

Improve the predefined restricted response.

Customer should have gone to a human

Add a clear escalation rule for that type of request.

Business Scope
      ↓
What the AI should handle
      +
Restricted Topics
What the AI should avoid
      +
Human Handoff
What needs a person

These three layers give your AI Agent a much clearer operating boundary.

AI & Conversational Automation

Part 21 of 22

Learn how to create and train ZazzyAgent AI Agents, connect business knowledge, write effective system prompts, use AI actions, detect customer intent, automate tasks, and hand conversations to humans.

Up next

Can I Build an AI Agent Without Knowing How to Code?

A few years ago, building software that could understand a customer, use business data and take actions usually meant involving a developer. AI has changed that. Today, a non-technical founder, market