Call Sample Review Gate (QA)

Call Sample Review Gate (QA)

1. 01

   Pull ten real call recordings per client before any tuning session

   Mix inbound and outbound, and include at least two calls that ended in a transfer or hangup. Ten is the floor for hearing a pattern rather than a one-off.

2. 02

   Score escalation accuracy against the client's written handoff rules

   Count how often the agent routed to a human when the caller asked for one, and how often it kept a caller who should have been transferred. Track both error directions separately.

3. 03

   Measure response latency at the turn level, not the call level

   Averages hide the pauses that make callers talk over the agent. Note the slowest three turns in each recording and the conditions that preceded them.

4. 04

   Verify consent and disclosure language on every outbound sample

   Confirm the recording notice, AI disclosure, and opt-out path fire before the agent makes its first substantive statement. Outbound rules differ by jurisdiction and by client vertical.

5. 05

   Reconcile usage cost against the call log for the same period

   Divide total platform spend by completed calls and by booked appointments. A cost per booking figure is the number a client will ask for, so have it ready.

6. 06

   Log the labor that remains after deployment

   List every task a person still performs: transcript review, CRM cleanup, rescheduling, escalation callbacks. This list becomes the honest scope of the retainer.

7. 07

   Write the findings into a one-page scorecard the client can read without a walkthrough

   Include the sample size, the error counts, the latency outliers, and the cost per booking. Send it before the review call so the conversation starts from shared numbers.