Skip to main content

Set characteristics and behavior of AI Agent

In the Agent section, you can configure the bot's persona, how the bot will present itself to users. This configuration focuses on defining the identity, tone, and behavior of the bot, rather than the specific conversations it will have. The bot persona settings ensure that your bot provides a consistent and human-like experience in every interaction.

Define Persona​

Define the persona for your AI agent here. This section focuses on designing how the agent should respond to users.

note
  • All fields marked with * are mandatory.
  • Click Save after making the changes.

image

Agent Name​

Choose a name for the agent that resonates with your customers.

Example: Mia

Agent Persona​

Agent persona defines how the AI agent interacts with users. A well-defined persona ensures consistent, engaging, and aligned interactions. You can select from the following personas:

  • Polite and persuasive: Ideal for sales agents.
  • Witty: Best for marketing agents.
  • Empathetic and helpful: Suited for support agents.

How Should the Agent Welcome Users?​

Welcome message: This is the initial message users will see when the bot opens. You can write a custom welcome message or select from an existing flow.

Example: Hey, I'm Mia from Yellow Banking. How can I assist with your banking needs today?

Describe the Agent's Role and Scope​

In fewer than 300 characters, explain the agent’s role. This helps the agent understand what topics to cover and what to avoid.

Example: You are a banking assistant who helps users with Yellow Banking-related issues and queries. You can persuade people to take personal loans, check the status of existing loans, and answer general queries about loan processes.

How Does the AI Agent Handle Unanswered Queries?​

Fallback message: Specify the response when the agent cannot help with a query. You can write a custom fallback message or select from an existing flow.

Example: I couldn't assist with your current query as I am still learning. Please reach out to my human colleagues at [email protected].

Rules to Follow During Conversation​

Click +Add rule to include multiple rules for conversations. Each rule can be described in up to 200 characters.

Example:

  • Always respond within 50 words or less.
  • You should sound reliable and make users trust that Yellow Loans are an investment in their bright future.
  • Always ask users if they want to apply for a new loan as a follow-up.
  • If the user seeks help on multiple topics, consider it small talk and ask which topic they prefer to start with.

Click Delete icon to delete any rule.

How to Handle Out-of-Scope Queries​

In fewer than 500 characters, explain how the bot should respond to questions outside its persona.

Example: If users ask about topics beyond loan needs or banking, or discuss competitors, politely re-emphasize that you are here to assist with loan needs only.

AI Agent Settings​

Under settings, you can configure the model and conversation memory.

image

Account​

Select the account name for which the LLM is enabled.

Click +Add account to configure a new LLM account and follow these steps.

Model​

Choose from the enabled models to select which LLM the AI agent will use.

This can be powered by third-party LLMs such as GPT-3.5, GPT-4, Claude Opus, Google Gemini, Anthropic, and other models to generate customized responses but by default, in-house model is used to generate responses.

Conversation Memory​

Set how many previous exchanges are remembered for context-aware replies.

  • 4 messages: Recommended for bots using a knowledge base, as customer questions are usually straightforward and answered quickly within 3-4 messages.
  • 8 messages: Recommended for bots without a knowledge base, where longer conversations may be needed to resolve queries.