Workflow

Support Conversation Deflection Scorecard

For Support ops, CS ops, RevOps, or customer experience leader when an AI support agent resolves, escalates, or drafts responses for customer conversations. The workflow produces valid deflection, false deflection, and escalation quality. Success means deflection rate increases without increasing reopen rate, escalation delay, refund risk, or negative sentiment.

LaneSupport and lifecycle
OwnerSupport ops, CS ops, RevOps, or customer experience leader
Briefs4
Trigger

An AI support agent resolves, escalates, or drafts responses for customer conversations.

Owner
Support ops, CS ops, RevOps, or customer experience leader.
Review
Review customer-facing, CRM, routing, budget, and approval-changing outputs before they ship.
Metric
Deflection rate increases without increasing reopen rate, escalation delay, refund risk, or negative sentiment.

Run it

Operating shape.

Inputs
  • Conversation transcript.
  • Resolution status.
  • Escalation reason.
  • Help center article used.
  • Customer sentiment.
  • Refund or churn risk.
  • Human review outcome.
Outputs
  • Valid deflection.
  • False deflection.
  • Escalation quality.
  • Missing content.
  • Risk flag.
  • Agent improvement backlog.

Steps

Control the first pass.

  1. Sample AI-handled conversations.
  2. Mark whether the resolution actually solved the issue.
  3. Identify cases that should have escalated sooner.
  4. Record missing help content.
  5. Flag churn, refund, or high-value customer risks.
  6. Update agent rules and knowledge base gaps weekly.

Prompt

Starting point.

Score these AI-handled support conversations for valid deflection, false deflection, escalation quality, missing content, sentiment risk, and improvement backlog. Separate safe-to-automate issues from human-required issues.