Right now, Respond.io’s AI Agent doesn’t have “conversation event” context—meaning it isn’t aware of important state changes like a conversation being closed and then reopened by a new inbound message. Because of this, the AI Agent often behaves as if nothing significant changed, even when the platform clearly shows the conversation has re-entered an open state.

A common scenario looks like this:

There’s a second related behavior: when the AI Agent has already called a tool earlier in the conversation, it tends to avoid calling that same tool again later—even when it’s perfectly valid and necessary because the conversation state has changed (for example, after a reopen event). This “one-and-done” tool behavior reduces reliability in event-driven flows.

AI Agent should be able to understand and use conversation events—especially closed → reopened transitions—so it can make state-aware decisions. When a conversation is reopened and the new inbound message is clearly an acknowledgement with no new intent, AI should confidently re-close the conversation to keep the inbox clean and prevent unnecessary follow-ups.

In addition, conversation events should act as a strong signal that prior actions may need to be repeated. AI Agent should be allowed (and encouraged) to call tools again when an event indicates a new state, rather than treating a previously-called tool as “off-limits” for the rest of the thread. This helps the Agent behave more like a dependable operator: aware of what just happened, and comfortable repeating the correct action when conditions warrant it.