#

How to change the behaviour of AI Agent?

Created by Ayush Sharma, Modified on Tue, 30 Sep at 6:37 PM by Ayush Sharma

Your AI Agent is more than just a chatbot it’s the voice of your brand for learners . The way it responds, and handles conversations directly shapes the experience people have with you. That’s why configuring its behaviour is important.

By setting the right personality, tone you can ensure your agent feels natural, professional, and aligned with your brand while effectively handling sales and support conversations.


Step 1: Go to AI Agent in the left navbar of your Graphy dashboard and click on Overview 


Step 2: Set Personality and Tone from behaviour

Your AI Agent’s behaviour decides how it responds, and interacts with learners. Setting this carefully ensures that your agent feels natural, trustworthy, and aligned with your teaching or brand style.


Here’s what you can configure:

1. Language

Allowing learners to interact with your agent in the language they’re most comfortable with makes conversations easier, builds trust, and removes communication barriers. 


2. Initial Greeting Message

This is the very first message your agent sends in new conversations.

  • By default, it’s set to: “Hello! How can I assist you today?"
  • If you wish, you can edit this to match your own style and tone.

A friendly, clear greeting sets the tone, makes learners feel welcomed, and encourages them to start asking questions.

3. Message on No Answer

Define what your AI Agent says if it doesn’t know the answer.

  • By default, it’s set to “I'm sorry, but I don't have an answer for that."

  • If you wish, you can edit this to match your own style and tone. 

Instead of leaving learners confused, your agent will politely handle gaps in knowledge. This keeps interactions professional and avoids frustration.

 

4. Creativity Level

Control how strictly your agent sticks to the provided knowledge that you have added in brain. You can set it to:

  • Strict

    • Uses only the content you’ve added in Brain.

    • No outside knowledge is included.

    • Best for when you want fully controlled, accurate, and source-specific answers.

  • Adaptive

    • Uses a mix of your content + general knowledge.

    • Strikes a balance between relying on your uploaded material and filling gaps with broader context.

    • Best for when learners ask questions that partly connect to your content but need extra clarity or examples.

  • Creative

    • Responses are freer, more conversational, and explorative.

    • Goes beyond just your content can add analogies, scenarios, or storytelling.

    • Best for making the learning engaging, fun, and easy to grasp.

5. Dynamic Questions

Enable your agent to ask follow-up questions at the end of responses.

This makes conversations engaging, keeps learners involved, and encourages them to dig deeper into the topic.


6. Enable Internet Search

Allow your agent to look up answers online if it doesn’t find them in your content.

Keeps responses fresh, updated, and accurate for queries outside your uploaded knowledge base.


7. Show Citations

When enabled, your agent shows the sources it used to generate an answer.

Builds trust and credibility. Learners know exactly where the information comes from.


8. Recency Bias

Gives priority to the most recent documents you’ve added in the brain.

Ensures your agent always uses the latest and most relevant knowledge, not outdated materials.


9. Favor New Content

De-prioritizes content the learner has already seen.

Helps learners explore fresh material instead of repeating the same answers.

 
Step 3: Use the Agent Preview to test your setup and ensure responses sound right. Testing avoids miscommunication and ensures your AI Agent is fully aligned with your standards before learners interact with it.








Was this article helpful?

That’s Great!

Thank you for your feedback

Sorry! We couldn't be helpful

Thank you for your feedback

Let us know how can we improve this article!

Select at least one of the reasons
CAPTCHA verification is required.

Feedback sent

We appreciate your effort and will try to fix the article