Configuring Generative Agent

Configure how GenerativeAgent interacts with end users and define its behaviors and actions. You have full control over its capabilities and communication style.

When GenerativeAgent engages in a conversation, it starts by conversing with the user to understand their needs or objectives. It then consults its available Tasks list and selects the appropriate task to assist the user. If no suitable task is found, it escalates to a human agent.

Follow these steps to configure GenerativeAgent:

1. Define the scope for GenerativeAgent
2. Configure core conversation settings
3. Create Tasks
4. Create Functions for those Tasks
5. Connect your Knowledge Base
6. Deploy your changes

After configuration, use the Previewer to test GenerativeAgent and make further refinements.

​

Accessing the AI Console

Configuring GenerativeAgent requires access to our AI Console, our dashboard for configuring and managing ASAPP. You should have received login credentials from your ASAPP team. If not, please contact them for access.

​

Step 1: Define the Scope

Define a clear scope to ensure GenerativeAgent provides safe and accurate assistance.

Consider and decide on:

• The voice or tone GenerativeAgent will use
• The types of issues or actions you want GenerativeAgent to handle (represented as Tasks)
• Which APIs your organization needs to expose for GenerativeAgent to address those Tasks (called Functions)

A Task is any issue or action you want GenerativeAgent to handle. Define a set of instructions in human language, and add one or more Functions, which are the tools GenerativeAgent can use for that task.

A Function is an API call given to GenerativeAgent to fetch data or perform an action. Once you’ve mapped out the APIs, Functions, and Tasks, use the GenerativeAgent UI to enter your configuration.

​

Step 2: Configure Core Conversation Settings

Configure the Core Conversation settings, including:

• Your company name for GenerativeAgent to use
• The welcome message for new customer connections
• How GenerativeAgent should refer to itself
• How your human agents are referred to
• A sentence explaining GenerativeAgent’s desired tone

Work with your ASAPP team to configure these settings.

​

Step 3: Create Tasks

Tasks are the foundation of how GenerativeAgent performs. This is often where you will spend most of your time when configuring GenerativeAgent. When analyzing a conversation, GenerativeAgent selects the appropriate task and follows its instructions.

To define a Task:

1. Navigate to the Tasks page
2. Click “Create task”
3. Provide the following information:
   • Task name
   • Task selector description
   • Task message (optional)
   • General Instructions
   • Functions the task should use

​

Improving Tasks

As you configure tasks, refer to Improving Tasks for strategies and tools to improve task performance.

​

Step 4: Create Functions

Functions enable your GenerativeAgent to perform actions similar to a live agent. For example, an airline might need Functions to check refund eligibility and process refunds.

Functions must point to specific API Connections and versions. API Connections contain technical details for connecting to specific API endpoints.

To create a function:

1. Navigate to the Functions page
2. Click “Create Function”
3. Provide the following information:
   • Function name
   • Description of how GenerativeAgent should use this function
   • The API Connection to use (default is the latest version)

​

Step 5: Connect your Knowledge Base

Connect your knowledge base to ASAPP and determine what information GenerativeAgent should use when assisting users.

​

Step 6: Deploy Changes

After configuring GenerativeAgent, deploy your changes. You have two environments and a draft mode:

• Draft: Changes are automatically available for testing with Previewer
• Sandbox: Test GenerativeAgent with your real APIs and perform end-to-end testing
• Production: Serve live traffic to your end users

​

Next Steps

With a functioning GenerativeAgent, you’re ready to support real users. Explore these sections to advance your integration: