When all live agents are at capacity, incoming chats have no fallback. Visitors who can't be routed to a human either wait or drop off, with no automated coverage. Conversely, when an agent unexpectedly becomes available mid-conversation, there's no mechanism to hand the chat back to a human.

The AI should automatically step in when all agents are occupied, and step back when a human becomes available, with routing switching dynamically based on real-time agent status.

Teams with variable staffing need both halves of this. A visitor who started with AI shouldn't stay with AI once a rep comes online, and a visitor who can't reach a rep shouldn't be left with nothing.