AI Approval Workflow
Gating specific actions of an AI agent behind a human decision: approve, edit, or reject.
Definition
An AI approval workflow is a control in which an AI agent proposes an action — a refund, a hand-over, a record change, an API call — and the action runs only after a designated person approves it, optionally with edited values. It turns the agent's mandate into policy rather than prompt wording.
In Depth
Good approval workflows are per action rather than all-or-nothing, so routine steps stay automatic while sensitive ones wait. They name who may decide, pass the decision to the next person when the first does not answer, and define a safe default when nobody does — typically rejection, with the customer told. They also record who decided, from where, and what values the action finally ran with.
In Lodgestory each chat action and each API tool on the AI Agent step is set to Off, Automatic or Needs approval. A teammate approves as proposed, approves with edits, or rejects with a reason the customer hears; a change nobody approves in time is rejected. Decisions arrive by ai_request.answered webhook and are available through the REST API and MCP tools.
Related Terms
Human-in-the-Loop AI Agent
An AI agent that pauses for a person at defined moments — to ask a question or get a change approved — and then continues.
Agent Handoff
Transferring a conversation from a bot or AI to a human agent.
Escalation Rules
Automated triggers that escalate conversations when SLAs are at risk.
Webhook
An HTTP callback that sends real-time event notifications to your server.
See it in action
Learn how Lodgestory implements this concept in the Human-in-the-loop AI agent feature.
Explore Human-in-the-loop AI agentReady to get started?
Experience omnichannel customer engagement with Lodgestory.