Agent led support tiering rebuild (L6)
Stop staffing tier one as a human queue. Rebuild the model so agents own resolution for the known corpus, humans own judgment and relationships, and the tiering map is redrawn around what the agent actually resolves instead of ticket volume.
The steps
- 01
Measure what is truly known
Tool: Zendesk
Cluster twelve months of tickets into known corpus, judgment, and unknown. Only the known corpus is a candidate for agent ownership. This number sets the ceiling for the entire redesign. Owner: support ops. DoD: a percentage split with the clustering method written down.
- 02
Redraw roles, not just routing
Tool: Manual
Define two human roles: resolution specialist for judgment cases and corpus owner who maintains the content the agent answers from. Corpus ownership is a real job with real time, not a side task. Pitfall: keeping tier one titles and hoping volume falls. That produces idle humans and stale content. DoD: new role descriptions approved with headcount attached.
- 03
Set the escalation contract
Tool: Agentforce
Write the exact conditions where the agent must hand off: confidence below threshold, contractual or billing content, regulated topic, second failed attempt, explicit customer request for a human. Owner: support leader plus legal. DoD: handoff contract published and instrumented.
- 04
Report resolution quality, not deflection
Tool: Intercom
Deflection is not a quality metric. Report agent resolution rate, reopen rate, handoff rate, and CSAT split by agent resolved versus human resolved. Publish all four together every month. Owner: support ops. DoD: the monthly report includes all four metrics side by side.
Tools in this playbook
Next playbooks
Unfamiliar terms are defined in the AI and Revenue Dictionary. Related frameworks live in the framework library.