Risk states reference
Every conversation with an active SLA policy moves through one or more risk states. This reference covers what each state means, how a conversation gets there, and what happens at the boundaries.
The four states
- SLA — Grey — Policy is active. Plenty of time remaining. — Carry on; no urgency
- At risk — Yellow — First threshold crossed. The window for a clean response is narrowing. — Look at it now if it’s yours; reassign if you can’t
- Urgent — Red — Last threshold before breach. Minutes-to-hours from violation, depending on policy. — Drop everything that isn’t more urgent
- Breach — Red, overdue — The SLA target was missed. Timer flipped to count up. — Respond as fast as possible to minimize impact
The visual indicators are consistent everywhere SLAs are surfaced: conversation list rows, the conversation header, the inbox sidebar counters, Analytics dashboards.
Escalation is one-way
A conversation’s risk state only climbs:
SLA → At risk → Urgent → Breach
It never walks back down within a single conversation. A timer pause (going outside office hours) does not return the badge to a calmer state. An agent reply ends a First Reply or All Messages timer, and the badge becomes Met or Breached.
This is intentional. The risk state captures the worst point the conversation reached during the SLA’s life — useful for retrospectives (“how many of our resolved conversations breached at some point?”). If a conversation hit At risk and you replied in time, a First Reply or All Messages badge shows Met.
Apart from an agent reply, the only reset is to close the conversation. On reopen (next section), some metrics restart their timer.
On reopen
When a conversation moves from done back to active (because the customer replied), each tracked metric’s timer restarts — with one important exception:
First Reply Time does NOT restart if an agent has already replied. That promise was kept. The First Reply badge stays as
Meteven after reopen.
The other metrics (All Messages, Resolution Time) restart fresh. Resolution Time in particular gets a new countdown — you have the full target duration again to resolve the reopened conversation.
On a team or channel change
When an agent or a team lead moves the conversation to another team, an automation assigns it, the AI hands it over, or the conversation switches channel, the SLA is chosen again for the new team and channel, and its deadline is counted again in the new team’s opening hours. The clock keeps the time the customer wrote in; it does not start again. The deadline can move later (the new team has shorter hours) or earlier (the new team has longer hours, or is open 24/7).
Two SLAs stay as they are:
- An SLA that an automation set.
- A First Reply SLA that is already answered. Its
MetorBreachedresult is final.
On close
When a conversation moves to done, the live countdown freezes. The badge flips from a counting timer to a static label:
- Met — the metric was satisfied within target
- Breached — the metric exceeded target before close
The final status is calculated once and recorded in the conversation’s audit trail. Even if the conversation later reopens and the badge changes, the historical record of the previous resolution survives in the Analytics data — you can see “this conversation breached on its second iteration” in reports.
A breached SLA can’t be un-breached
Once the timer hits zero, the SLA is breached. There’s no “set the badge back to Urgent” override. The breach goes into the historical record permanently.
What you can do: respond as fast as possible to minimize the negative-time count and the customer-impact window. The breach is logged regardless, but a 30-minute breach is meaningfully better than a 5-hour breach.
Audit trail
Every SLA assignment a conversation has ever had — including ones it had earlier in its life if the policy assignment changed — is preserved:
- The policy ID
- The metric type
- The priority band when the assignment started
- The start time, the due time, the actual response time
Useful for analytics, debugging “why did this breach?”, and any audit-style review.