Genesys Cloud Copilot AI actions overview

AI-generated summary

Genesys Cloud Copilot utilizes an AI action-based pricing model where discrete operations performed by AI agents are charged based on consumption rather than subscription fees. The system categorizes actions into knowledge actions, read actions, write actions, and analytical actions, each representing a single atomic operation using an API or tool. Chargeable actions include data modifications such as creating users or assigning roles, while non-chargeable actions encompass knowledge explanations and utility calculations like date processing. Action consumption scales linearly with request complexity, particularly when operations involve multiple entities. For instance, assigning a role to four users requires eight actions compared to two actions for a single user, as the AI agent must perform separate read and write operations for each entity. However, batch operations can optimize action consumption by processing multiple entities within a single action when available. The platform dynamically determines the exact number of AI actions required per request based on user prompts and context. Specific use cases demonstrate varying action requirements: knowledge queries incur no charges, role assignments to individual users require three actions (user lookup, role lookup, and role assignment), bulk role assignments scale per-user lookups while consolidating role and assignment operations, and queue assignments for multiple users optimize by performing single queue lookups with consolidated assignments. Recognition features require two actions (user lookup and recognition transmission), while analytics operations utilize utility actions for date range resolution at no charge, with analysis actions charged separately. All chargeable AI actions are tracked and visible in analytics, providing transparent monitoring of Copilot usage and associated costs. This action-based pricing structure enables organizations to manage expenses while performing administrative, analytical, and employee recognition tasks within the platform.

Genesys Cloud subscribers can add Genesys Cloud Copilot to their subscription. Charges for Genesys Cloud Copilot are based on the number of AI actions Genesys Cloud Copilot uses to complete the requested tasks.

An AI action is an atomic operation that an AI agent performs using an API or a tool. Each step required to complete a task counts as a separate AI action.

Genesys Cloud Copilot action types

Action type	Visible in analytics	Charged	Example
AI action	Yes	Yes	Create a new user
Knowledge	Yes	No	Explain how to set up a progressive dialing campaign
Utility	No	No	Calculate relative dates as part of processing a prompt

Action scaling considerations

Genesys Cloud counts AI actions per atomic operation. The exact number of AI actions required to complete a task depends on the user request (the prompt) and the user context. Genesys Cloud Copilot determines the number of AI actions dynamically for each per request. When a request involves multiple entities (for example, multiple users), the number of actions typically scale linearly because each entity requires its own set of read and write operations.

For example, to assign a role to one user based on their name requires two actions (one to find the user by name, and one to assign the role to that user). To assign the same role to four users requires eight actions (two actions for each of the four users).

When available, batch operations can reduce the number of actions by allowing multiple entities to be handled in a single AI action. However, prerequisite steps such as validation and lookup may still occur per entity.

Example use cases

Use case	Prompt	Actions	Total AI actions
Knowledge question	How do I create a work type?	Query Genesys Cloud Resource Center.	1 (no charge)
Assign a role to a user	Add Agent Copilot role to Jane Doe.	Find the user by name. Find the role by name. Assign the role to the user.	3
Assign a role to multiple users (two users)	Assign Agent Copilot role to John Doe and Adam Smith.	Find the role by name (once). Find the user by name (per user). Assign the role to the user (per user).	5
Add multiple users to a queue (four users)	Add John Doe, Jane Doe, Joe Bloggs, and Adam Smith to the Support queue.	Find the user by name (per user). Find the queue by name (once). Assign all users to the queue (once).	6
Analytics insight	Which virtual agents performed worst last week?	Resolve “last week” into a date range (utility action). Analyze the results.	1 (utility, no charge), 1
Send recognition	Send a recognition to John Doe with the type “Thank you” and the title “Thank you for closing that important sale!”.	Find the user by name. Send the recognition.	2

[NEXT] Was this article helpful?

Get user feedback about articles.