The team was not slow. It had no way to see which customers were waiting longest - so it answered whatever was at the top of the inbox.
The starting position
Seven agents shared a Google Workspace group. Each worked newest-first, because that is how a mailbox sorts. Easy questions got quick answers; hard ones sank, and some resurfaced only when the customer chased.
The median first response, measured afterwards from the mailbox, was 19 hours of wall-clock time.
What they changed
Three things, in order.
- One SLA policy: 4 business hours to first response for paying customers, 1 business day for trials, on a 9-to-7 calendar with holidays.
- Routing by mailbox and category, with round-robin inside the two teams.
- Each agent's day started from "Awaiting my reply" and "SLA at risk", sorted oldest first, instead of the inbox.
What did not help
An early attempt to create a "Waiting on engineering" status that paused the clock made the report look better and changed nothing for customers. They removed it after two weeks and let those tickets count.
The result
By week eight the median first response was 3 business hours and 91% of first-response targets were met. The nine-day-old emails were gone in the first week and did not return. No agents were added.