AI Agents
Overview
AI Agents interact directly with your customers. Conversations are assigned to AI Agents in the same way they are assigned to human agents. AI Agents can send messages, set fields, requeue/escalate, and close conversations, much like human agents.
AI Agents are one of three types of bots in Quiq:
- Custom bots hosted externally and written against Quiq's Bot API
- Bots built in Quiq's Classic NLU Bot Designer
- AI Agents built in AI Studio
AI Agents inherit the full feature set of classic assistants. Therefore, the Bot Designer Guide is still a useful resource for building AI assistants as we continue to update our documentation.
Flow Execution
AI Agents fully own the conversations assigned to them. When a conversation is first assigned to an AI Agent, the flow starts at a special Assigned Behavior . From there, flow execution continues until it reaches a location where it's waiting for a new message from the customer. This is sometimes called a flow resting point.
The following are possible resting points in a customer agent's flow:
- Send Message Behavior : if 'wait for response' toggle is enabled
- Call API Behavior: if 'wait for response' toggle is enabled, or the API response payload specifies to await a response
- Prompt AI Behavior: if it doesn't change locations in it's success/error handlers
- Set Timer Behavior
AI Agents are the only type of agent that rest at a location in the flow and then resume execution from that same location when the next event (e.g. customer message or timer expiration) occurs. An agent's reaction to an event starts at its last resting point (or Assigned node) and continues until it rests again. This is referred to as a 'user turn'. The events logged in the Debug Workbench correspond to user turns.
AI Agent Context
The context that AI Agents execute contains the standard data common across all agent contexts:
- Session fields
- Search results
- API results
- Buffer state
The defining aspect of the AI Agent context is the conversation, which includes full transcript history (including message delivery statuses, rich messaging elements, etc), field state (conversation and customer fields) and much more.
Full Reference: CustomerAssistantContext Object
Ai Agents have special behaviors such as a behavior dedicated to sending a message and waiting for a response.
AI agents likewise have special actions like starting the typing indicator
Ai agents have special global triggers called Special Events that are useful when automating conversations
Updated 8 days ago