Glossary

Product & Platform

AI Supervisor Zowie's monitoring and performance interface for reviewing AI Agent interactions. The Supervisor tables in the shared dataset expose the same data visible in this UI.

Inbox Zowie's human agent workspace for managing support conversations. The source of Inbox tables in the shared dataset.

GenAI Zowie's generative AI layer powering the AI Agent's reasoning and responses. The source of AI Agent session events — messages, tool calls, and reasoning steps — in the Events table.

CRM Zowie's customer data service, which stores end-user profiles. The source of the Users table.

Data Model Concepts

Unified data model The normalized data hierarchy (Users → Tickets → Interactions → Events) that consolidates data from across Zowie's product services into a single consistent schema.

Environment ID (environment_id) A unique identifier for a customer's Zowie account instance. Present as a primary identifier on all tables and used as a clustering key for query performance. All data in a linked dataset belongs to a single environment.

Chatbot ID (owner_user_id)
Unique identifier of the chatbot configuration instance that handles interactions. Each chatbot represents a specific AI agent setup.

Customer slug A short identifier for a customer account (e.g., acme for "Acme Co.") used as a prefix in all table names in the shared dataset.

Linked dataset The BigQuery dataset created on the customer's side when subscribing to Zowie's Analytics Hub listing. Provides a read-only view of Zowie's shared data.

Snapshot fields Fields on the Tickets table (prefixed snapshot_*) that capture a user's CRM profile at the moment the ticket was created, preserving a historical record even if the profile has since changed.

Entities in the Data Model

Users table Contains profile fields (name, email, phone), authentication level (authenticated, recognized, or anonymous), external user ID, and arbitrary custom properties set by the customer, of end user

Tickets table Contains single support requests. Request can originate from an AI Supervisor interaction or a standalone Inbox thread.

Interactions table Contains a segment of a ticket owned by a single actor — AI Agent, human agent, or queue. A new interaction is created each time ownership changes.

Events table Contains individual messages, actions, or system events within an interaction. The most granular level of data in the unified model.

End User The customer interacting with the AI Agent or human agent

Human Agent A support team member handling a conversation in Inbox. Referenced as owner = 'human_agent' in the Interactions table.

Interaction Concepts

AI Agent interaction An interaction segment handled by the AI Agent, sourced from AI Supervisor. Includes contact reasons, intents, sentiment, and knowledge sources used.

Inbox synthetic interaction A human agent or queue interaction segment derived from Inbox chat metadata, synthesized into the unified interaction model.

Standalone inbox ticket An Inbox thread that was never handled by the AI Agent — has no corresponding AI Supervisor interaction.

Contact reason A classification of why a customer reached out, recognized by the AI Agent. Stored as recognized_contact_reasons and final_contact_reason on Interactions.

Final contact reason The definitive contact reason assigned at the end of an AI Agent interaction, used for volume and automation rate analysis.

Event Types

AI Agent (GenAI React) Events

These events trace the AI Agent's reasoning loop — from receiving a user message through LLM calls, tool use, and response generation.

Event Type	Description
`UserMessageV2`	End-user sends a message to the AI Agent. Contains full input text, channel info (Email/Chat), react configuration, user details (language, region).
`AIAgentResponseV2`	AI Agent produces a response. Contains generated reply text, or `cannotHelp: true` when the agent can't answer (triggers handover).
`ReasoningResultV2`	Output of a single AI reasoning step — the agent decided what to do next (use a tool, respond, escalate).
`ToolRunV2`	AI Agent initiated a tool execution (e.g., KnowledgeSearch, intent lookup).
`ToolUsedV2`	Tool execution completed, result returned to the agent.
`ToolKnowledgeSelectedSourcesV2`	Knowledge sources selected and ranked by the knowledge search tool. Shows which FAQ/KB articles were retrieved.
`AvailableToolsV2`	Snapshot of tools available to the AI Agent at a given reasoning step. Changes as tools are added/removed during the turn.
`ParallelTasksInfoV2`	Metadata about parallel task execution within the AI Agent reasoning loop.
`IntentCandidatesV2`	Intent classification candidates identified by the AI. Includes confidence scores and goal extraction.
`IntentResponseV2`	An intent was matched and a response was triggered from the FAQ/intent system.
`ReactErrorV2`	Error in the React (AI reasoning) pipeline. Includes error type and details (e.g., timeout).
`ReactGuardV2`	A post-tool guard/safety check was triggered during AI reasoning.

Zowie Engine (Workflow) Events

Events from the deterministic workflow engine that handles structured conversation flows, API integrations, and decision logic.

Event Type	Description
`SessionStarted`	New Zowie Engine workflow session started for the interaction.
`GoToWorkflow`	Navigation to a top-level workflow.
`GoToSubWorkflow`	Navigation from a parent workflow into a sub-workflow.
`ReturnToParentWorkflow`	Returned from a sub-workflow back to the parent.
`GoToModule`	Navigation to a reusable module within a workflow.
`BlockEntered`	A specific block (step) in a workflow was entered. Very high volume — tracks every step execution.
`BlockError`	Error occurred executing a workflow block.
`Response`	Workflow sent a text response to the user. Contains the response text and strategy.
`Log`	Debug log message emitted by a workflow block.
`KeyRetrieved`	Retrieved a stored key-value pair from session/conversation storage. Highest-volume event type.
`KeyStored`	Stored a key-value pair to session/conversation storage.
`ApiCallMade`	External HTTP API call made from a workflow block. Contains full request/response details.
`UserActionReceived`	Raw user action received by the engine (text message, button click). Lower-level than `UserMessageV2`.
`RuleMatched`	A condition/rule matched in a decision table block.
`DefaultActionFallback`	No rules matched in a decision table — default/fallback action was triggered.
`ExtractorStarted`	A data extractor (custom JS function) started executing.
`ExtractorSuccess`	Data extractor completed successfully.
`ExtractorFailure`	Data extractor failed with an error.
`LoopDetected`	Infinite loop detected in workflow execution — safety mechanism that halts processing.
`Observation`	Observation/debug info emitted during AI processing within workflows (e.g., prompt instructions, context).

Inbox Configuration

Inbox Group A grouping of human agents configured within Inbox, used for routing and organization.

Inbox Queue A waiting state in Inbox where conversations are held before being assigned to an agent.

Inbox Topic A classification label configured in Inbox, used to categorize conversations.

Inbox Thread A conversation between an end user and a human agent within Inbox.

Inbox User A human agent account configured within the Inbox instance.