AI Agent Idle Action
The idle action is an AI Agent (v2) advanced setting: when a customer goes quiet for a configured duration, the AI Agent automatically takes one action — nudge the customer with a follow-up message, assign the room to a human, or resolve the conversation with a closing message. Today it is single-action: an agent can do one of those after one timer, and the only way to chain "nudge then resolve" is by coupling to the channel-level global idle rule in bot settings.
This initiative makes the idle action a self-contained, user-configurable sequence owned entirely by the AI Agent — up to 3 ordered steps, each with its own idle duration, action, and message — so a PM or customer Bot Builder can configure "idle 10 min → follow-up → idle 10 min → resolve" directly on the agent without touching channel settings.
Scope Changes
Engineering surfaces this initiative touches (controlled vocab). Kept in sync with the scope_changes frontmatter above.
- Backend —
chatbot: extendai_agents.parameters['idle_rule'](JSONB) to carry an orderedstepsarray; reworkprocess_idle_rule_action.rb+send_message_with_resolve.rbto advance/reset a per-room step pointer instead of chaining to the global channel idle rule; new feature flagai_agent_sequential_idle. - Frontend —
chatbot-fe: turn the single-action picker inAiAgentIdleActionForm.vueinto a step-list builder (add/remove up to 3 steps); extendidleActionUtils.tstransforms and theIdleRulestore interface. - Design — Figma for the step-list builder UI — TBD, designer not yet assigned.
QA Lane
Lane B — keeps a human QA gate. No Lane-B trigger is present (an idle-timeout action on existing agent flows), but no E2E test specs exist for this initiative yet, so the Lane-A entry bar (100% E2E, spec-mapped coverage) is unmet — Lane B by default. Classified 2026-06-29.
PRDs
- Sequential idle follow-up (ADJUSTMENT) — make the AI Agent idle action a user-configurable, self-contained sequence of up to 3 steps. Confluence: AI Agent — Idle Action — Adj: Sequential Idle Follow-up.