Skip to content
AskVault Docs
Try Free →

SLA policies on conversations

Written by Joseph Russo, Principal Architect, Networks and Tech, AskVault · Reviewed by Allison Chan
Last updated: · 4 min read

What an SLA policy defines

An SLA policy has three settings:

  • Channel. One channel, such as widget or WhatsApp, or any channel.
  • First-response target. How many minutes your team has to send its first reply.
  • Active. Whether the policy is in use.

A policy can also store a resolution target, but nothing measures it today. There are no priority conditions, customer-tier conditions or escalation steps.

SLA policies are part of the Help Desk on Growth and above. Growth+

Creating a policy

The dashboard has no screen for creating, editing or testing SLA policies today, and the Setup Copilot can't create one either. Policies can't be created with an API key.

The rest of this page describes how a policy behaves once a workspace has one.

Which policy applies

When a conversation needs a deadline, AskVault picks one active policy:

  1. A policy for the conversation's exact channel wins over an "any channel" policy.
  2. Among equal matches, the oldest active policy wins.

Email Assistant and hosted-inbox conversations don't match a policy limited to the email channel today. Use an "any channel" policy to cover them.

When the clock starts

The deadline is set when a conversation is escalated in one of these ways:

  • A teammate clicks Escalate on a conversation in Chat History.
  • Your own system calls POST /v1/agent/escalate with an API key.
  • An AI action a teammate approved fails, and the conversation is handed to a person.
  • An AI action waiting for a teammate's approval expires before anyone reviews it (rolling out), and the conversation is handed to a person.

These common handoffs don't start an SLA clock today: the AI's escalate_to_human skill, the visitor's "Talk to a human" button, automation rules, and a teammate taking over or pausing the AI.

Each conversation gets one deadline. Escalating again does not reset it.

When the SLA is met

The SLA is settled by the first reply a teammate sends from the Help Desk inbox. AI answers don't count.

The pill then shows SLA met. It shows met even when that first reply arrived after the deadline, so check the report for late replies.

The SLA pill

Conversations with a deadline show a pill in the Help Desk list and conversation view:

  • On track. Time is left.
  • Near breach. 5 minutes or less left.
  • Breached. The deadline passed with no teammate reply.
  • SLA met. A teammate has replied.

The inbox has an SLA filter for these states, and saved views can filter on them.

Rolling out; not yet switched on. A first reply sent after the deadline will count as a missed target, not "SLA met". The pill then reads a grey Late instead of a pulsing red Breached, because nobody needs to act on it any more, and Breaching now lists only conversations still waiting for their first reply past the deadline. Late replies still count against compliance in the report.

Office hours

Set office hours in the Help Desk with Office hours in the side menu or the command palette. Pick a time zone and the working hours for each day.

When office hours are set, the first-response deadline counts only working minutes. A 1-hour deadline set at 4:30pm on a 9-to-5 schedule is due at 9:30am the next working day. With no office hours, deadlines run 24 hours a day.

There is no holiday calendar.

The SLA report

The Help Desk reports include SLA compliance over a window you choose, 30 days by default:

  • Counts of met, breached, near-breach and on-track conversations.
  • Compliance rate.
  • Open breaches. Up to 50 conversations that passed their deadline without a reply, with their channel and assignee.

A conversation that is resolved without a teammate reply stays counted as breached. There are no per-agent SLA columns and no CSV export for this report.

Breaches

When a deadline passes, the pill turns red and the conversation appears in the report's open breaches list. Nothing else happens:

  • No in-app, email or Slack notification.
  • No conversation.sla_breached or conversation.sla_risk webhook.
  • No automatic reassignment or priority change.

Snoozing a conversation does not pause its deadline.

Common pitfalls

No SLA pill on conversations. Either the workspace has no active policy, or the conversation was escalated in a way that doesn't start the clock. Rolling out; not yet switched on: setting a conversation's status to Escalated from the status menu or in bulk will start the clock and run auto-assignment, the same as the Escalate button.

Email conversations have no deadline. A policy limited to the email channel doesn't match them. Use "any channel".

Breached conversations you already closed. Resolving without replying leaves them breached in the report.

FAQ

Does the SLA clock pause when a customer doesn't reply?

No. There is only a first-response deadline, and it waits on your team, not the customer.

Can SLA policies trigger automation?

No. Automation rules have no SLA trigger.

Does the bot's answer count as the first response?

No. Only a teammate's reply from the Help Desk inbox settles the SLA.

How precise is breach detection?

The state is worked out each time the inbox loads, by comparing the deadline with the current time. The near-breach window is 5 minutes.

Was this page helpful?