The question arrives in every engagement, usually in the second week, usually quietly: what happens to the team? It deserves a straight answer, because the honest version is more interesting than either the utopian or the grim one.
When an agent takes a queue, the work does not vanish. It changes shape. Triage disappears; judgment concentrates. The roles that emerge are real jobs, and the operations that thrive are the ones that staff them deliberately instead of letting them happen by accident.
The three roles that appear
The reviewer owns the agent's uncertain cases. Everything the system escalates arrives with a full brief: what the agent saw, what it would have decided, why it did not. Reviewing is harder than the old triage and noticeably more senior; it is the best training ground the function has ever had. The escalation owner holds the boundary: which decisions the agent may never take alone, and whether that boundary should move as evidence accumulates. And someone, with the role written into their week, reads the evaluation report. Every week. The score, the regressions, the new failure modes. That half hour is the operation's immune system.
The org chart's new unit is not a department. It is an outcome, owned by one human, run by a squad.
Above the loop, not inside it
The pattern across engagements is consistent: people move above the loop, not out of the building. Fewer hands touch routine cases; more attention goes to limits, quality and the next queue worth rebuilding. The teams that resist this hardest in week two are, reliably, the ones defending it hardest by week twenty, because the new work is simply better work.