AI Agent Settings
You can find how to do settings.
Last updated
Was this helpful?
You can find how to do settings.
Last updated
Was this helpful?
Control and update the name, description, display name and more for your AI Agent.
This is the most important area for your AI chatbot because it is where you will customize the AI agent for each client/ business.
This is the internal name of the chatbot in your account. This name is not seen by users as long as a display name is added as well.
This is the name shown to users on the chat widget.
This user-visible name represents your chatbot. We advise including 'Al' for clarity. If this is not set, then Chatbot's name will be shown.
Enter the chatbot's initial message(s) to users, separated by a new line for multiple messages.
This message greets users upon their initial interaction with the chatbot. To add multiple messages, separate them by new lines.
Time in seconds for auto-showing initial message pop-up. Negative value disables it.
Predefined clickable messages for user prompts. Separate multiple messages with new lines.
Example: "What type of documents can I upload to the Knowledge Base"
Chat Widget Pop Up
When enabled this will open the entire chat widget to the user on the webpage.
The Advanced settings are found in the "Update" section of each AI chatbot.
Here are all the items you can edit inside of the advanced settings. By default we add all of the most optimal settings to every chatbot when created. If you want to customize the bot even further, this is a good place to start.
Select the AI LLM model you wish to use for your AI chatbots.
What is a Model?
The word 'model' refers to the different AI software being used to power your AI Agents.
What's the Difference?
Each model comes with its own unique cost and abilities. The most powerful AI model is GPT-4o and the most affordable AI model is GPT-3.5.
This is where you can select between the current available models from OpenAI.
Given that these AI agents are designed to assist companies in customer support and sales, the advantages of each model can be evaluated based on these business-specific needs.
GPT-3.5
1 response = $0.01 /message
Cost-Effective for High Volume: Ideal for handling large numbers of customer queries efficiently, making it suitable for businesses with high traffic and cost-sensitive budgets.
Quick Responses: Provides fast answers for straightforward FAQs or repetitive customer queries.
Basic Task Handling: Can efficiently manage tasks like providing product details, checking order statuses, or answering simple policy-related questions.
Low Implementation Complexity: Easy to set up and sufficient for businesses that don’t require advanced or nuanced interactions.
GPT-4o
The most powerful model to use for advanced uses cases and function calling.
1 response = $0.04 /message
Best used for all use cases
Multimodal Capabilities: Can process and respond to both text and image inputs, which is invaluable for e-commerce or technical support.
Enhanced Creativity: Excellent for crafting engaging sales pitches, personalized messages, or marketing copy to attract customers.
Complex Problem-Solving: Handles nuanced queries and provides detailed explanations, improving customer satisfaction.
GPT-4 Turbo
1 response = $0.08 /message
Enterprise-Level Scalability: Designed for businesses managing large-scale operations with high customer interaction volumes while maintaining speed and accuracy.
Extended Context Understanding: Can process larger customer histories or complex sales inquiries without losing track of the conversation, leading to more personalized interactions.
High Precision: Reduces errors in responses, critical for handling sensitive topics like billing issues, legal policies, or complex sales agreements.
Sales Optimization: Great for generating tailored upselling or cross-selling opportunities based on detailed product data or customer preferences.
Cost Efficiency for Advanced Tasks: While pricier than GPT-3.5, it offers a balance of cost and performance for businesses that need advanced capabilities at scale.
When enabled this feature will prioritize the data found in the question and answer section of the knowledge base.
The chatbot's responses by placing a higher emphasis on delivering precise answers to user questions.
Prioritize question/answer results over others when querying the chatbot.
Temperature parameter for the GPT Chatbot API.
Temperature parameter for the GPT Chatbot Completion API. determines the randomness of the response.
A low value (close to 0) makes the output more deterministic and conservative, while a higher value (up to 2) makes the output more diverse and creative.
For most businesses setting the temperature to 0.05 to 0.20 will be a good place to start testing.
Feature Description: Temperature Setting for the Chat Box
The temperature setting in the chat box controls the randomness and creativity of the AI model’s responses. Recently, the numeric input for this setting was replaced with a slider, making it easier to adjust.
What Does the Temperature Setting Do?
Lower Temperature (Closer to 0):
Makes the model’s responses predictable and consistent.
Best suited for business use cases where reliability and accuracy are essential.
Higher Temperature (Closer to 1 or 2):
Makes the responses more varied, creative, and less predictable.
Suitable for scenarios where a bit of creativity is beneficial, but higher values (like 1 or 2) are rarely practical for most applications.
Range of the Setting
The temperature slider technically supports a range between 0 and 2.
While OpenAI documentation references both 0–1 and 0–2 ranges, setting the value to 2 is uncommon and generally not recommended for most business cases.
Recommended Usage
Start at 0 (Default Setting):
For most business-related use cases, start with a temperature of 0 for consistent and predictable results.
Gradual Adjustments:
Once the chatbot behaves as expected, experiment by incrementally increasing the temperature (e.g., 0.2) to introduce a small degree of creativity if needed.
Adjust settings one at a time (either the prompt or the temperature) to evaluate their impact effectively.
Avoid Extreme Values:
Values close to 1 or 2 typically lead to overly random and unpredictable outputs, which are rarely suitable for professional applications.
If you find a use case where setting the temperature higher (e.g., 1 or 2) delivers value, we encourage you to share your experience. Understanding unique scenarios helps improve the platform and refine recommendations.
Conclusion
For business use cases, a lower temperature (default: 0) is recommended for predictability.
For tasks requiring slight creativity, consider small increases (e.g., 0.2).
Avoid extreme settings unless you have a specific use case that demands it.
The temperature setting is a powerful way to balance consistency and creativity, ensuring the chatbot aligns with your specific needs.
This is an advanced feature for changing how the AI agent fundamentally works.
Modify the template used whenever a message is sent to ChatGPT.
The message template is displaying how the AI chatbot will process data from the knowledge base and add it to your base system prompt.
If you are looking to create an alternative AI agent that can do different things like classify text, write emails, or any other use case you could imagine.
Use this email to replace the agency's default support email for this chatbot, if one exists.
Enter your support email for this chatbot. This will be displayed to users in case of errors or other issues.
Enter the domains on which your Chatbot will work. Separate multiple domains with commas.
This feature is closely related to Visibility.
If a website visibility is set to public, then this section does not apply.
If embedding the chatbot on a website, enter the exact domain URL in this format: 'https://yourdomain.com'
Enable to show the data sources of the chatbot's response in the chat widget.
When an AI agent responds and is shown to a user, a little dropdown menu toggle will show up underneath the message response showing the links to where the AI agent found the answers.
This is only available for public websites scraped with the AI agent scraper.
This is helpful to enable when testing a new AI agent to see what data is being sent to OpenAI.
Enable Debug mode to display both the search results generated by the query and the data exchanged with the ChatGPT API for troubleshooting or analysis purposes.
