cx-agent-experience-auditlisted

# Agent experience audit When quality or handle time slips, the default fix is **more training**. Often the binding constraint is **the job itself**: six systems per ticket, a macro that stopped working, an approval that requires a manager who is offline, a KB article that contradicts policy. Agents compensate with workarounds; WFM sees occupancy; customers see inconsistency. **Fix the top friction before buying more training.** An audit quantifies where time and errors go that agents cannot control. ## Step 1: pick representative work Do not audit from memory or a tour from engineering. - **Sample contact drivers** that cover most volume — include the worst handle-time and highest repeat-contact strata. - **Include channel variants** — voice often has different tooling from async. - **Include vendor agents** if they use narrower access; friction there is frequently yours, not theirs. - **Sessions:** shadow at least 3–5 agents across shifts, 90–120 minutes each, or equivalent ticket replay with screen recording if shadowing is impossible. State what you did not observe (overnight, specific language, new hire vs tenured). ## Step 2: shadow protocol For each session, the observer records **without coaching**: | Field | Capture | | --- | --- | | Contact id / driver | For traceability | | Systems opened (in order) | Name each; count hops | | Time in each system | Rough buckets: >30s matters | | Copy-paste or re-key events | Same data entered twice | | Dead ends | P