Guidance on how to understand the AI Studio skill execution flow

AI Studio skill execution flow


The workflow table below shows how Zoom Virtual Agent uses a Marketplace Action to retrieve data from an external system during a conversation.

The Virtual Agent identifies the user's intent, gathers any required information, invokes the configured action, and uses the API response to generate a personalized reply.

This pattern enables secure, real-time access to external data while maintaining a natural conversational experience.

| Step | Description | Example |
|------|-------------|---------|
| Start the interaction | The user submits a request. | "Check my order status" |
| Trigger the skill | The Virtual Agent identifies the intent and triggers the matching skill. | — |
| Collect required input | The skill prompts the user for missing data. | "Provide your order ID." |
| Capture user input | The user provides the required value. | `12345` |
| Invoke the Marketplace action | The skill sends the input to the configured Marketplace action. | — |
| Process the request | The external API processes the request and returns order data. | — |
| Map response data | AI Studio maps API response fields to skill variables. | — |
| Generate the response | The Virtual Agent uses mapped variables to generate a natural language response. | "Your order is being processed and will arrive by March 25." |




Step	Description	Example
Start the interaction	The user submits a request.	"Check my order status"
Trigger the skill	The Virtual Agent identifies the intent and triggers the matching skill.	—
Collect required input	The skill prompts the user for missing data.	"Provide your order ID."
Capture user input	The user provides the required value.	`12345`
Invoke the Marketplace action	The skill sends the input to the configured Marketplace action.	—
Process the request	The external API processes the request and returns order data.	—
Map response data	AI Studio maps API response fields to skill variables.	—
Generate the response	The Virtual Agent uses mapped variables to generate a natural language response.	"Your order is being processed and will arrive by March 25."