Service level agreement (SLA)· Support operations
Service Level Agreement (SLA) — The Clock on Every Support Conversation
Quick answer: A support SLA is a time target your team commits to, either to a customer in a contract or to itself as an internal goal. The usual targets are how fast the first human reply arrives, how fast each later reply arrives, and how long it takes to solve the issue. Help desk software starts a timer when a conversation matches the SLA, can pause it while you wait on the customer, and marks the conversation as breached if the time runs out.
Where the word comes from
Outside support, an SLA is a section of a contract: an IT provider or software vendor promises uptime or response times and pays a penalty if it misses them. Support teams borrowed the term for any written response promise. In small businesses most support SLAs are internal targets with no penalty, though B2B companies often put them in customer contracts, sometimes with faster targets for higher-priced plans. Intercom's article lists both uses: response times agreed with VIP customers and internal support goals.
What a support SLA measures
| Target | What the timer measures | Vendor names |
|---|---|---|
| First reply | Time from the customer's first message to the first reply from a person | Zendesk "First reply time"; Intercom "First Response Time"; HubSpot "time to first reply" |
| Next reply | Time from each later customer message to the next reply | Zendesk "Next reply time"; Intercom "Next Response Time" |
| Regular updates | Time between public agent replies on a long issue | Zendesk "Periodic update" and "Pausable update" |
| Resolution | Time from the first message until the issue is closed or solved | Zendesk "Total resolution time", "Requester wait time", "Agent work time"; Intercom "Time to Close" and, for tickets, "Time to Resolve"; HubSpot "time to close" |
For live chat, first reply and next reply matter most, because the customer is usually still on the page. Resolution targets matter more for email and tickets that run over days. Zendesk recommends choosing only one resolution metric per policy; reply and update metrics can be combined.
How help desks apply an SLA
Three settings decide which conversations get a timer and how it runs.
- Which conversations. Zendesk policies have conditions and set targets by the system Priority field (Low, Normal, High, Urgent); a ticket with no priority gets no SLA, and the order of policies decides which applies, so Zendesk advises putting the strictest at the top. Intercom applies an SLA through a workflow step, often when a new Messenger conversation or ticket starts, and only one SLA can be active on a conversation at a time. HubSpot applies SLAs to all tickets in the inbox or by ticket priority.
- Which hours count. Zendesk lets you pick calendar hours or business hours for each priority. HubSpot offers "Operation hours" or "24/7", and notes that holidays are not yet accounted for. Intercom sets targets in calendar hours divided across your office hours, so a message received at night can be due the next working day.
- What pauses the clock. Intercom can pause timers while a conversation is snoozed or waiting on the customer. Several of Zendesk's metrics pause when a ticket is Pending, the status that means you are waiting on the requester. HubSpot states that tickets in "Waiting on contact" are not excluded from SLA calculations.
What a breach looks like
A breach means the timer ran out before the action happened. Help desks show this in the inbox so agents can pick the most urgent conversation first. Intercom colors badges red when overdue and orange with under 5 minutes left; if an SLA is missed while a conversation is snoozed, Intercom reopens it. Its reports sort SLAs into hit, missed, and "fixed", meaning the target was missed but the reply or close happened later. HubSpot shows "due soon" and "overdue" states with limits you set. See our Intercom review for how the inbox handles this in practice.
SLA versus service level
The two terms are often mixed up. An SLA is a promise about each conversation: this chat must get a reply within 2 minutes. A service level is a percentage over many conversations: 80 percent of chats answered within 2 minutes. Contact centers usually manage to the percentage, because no team hits a strict time on every single chat at a busy peak. Staffing maths such as Erlang C, used in the chatbot handoff design guide, predicts the percentage, not individual breaches. A practical rule is to set the per-conversation SLA in your help desk and report the share of conversations that met it each week.
Where chatbots fit
A bot reply does not normally count as meeting a human SLA. Zendesk states that autoreplies fulfill reply and update metrics only with its Copilot add-on, and that reply-time SLAs for live chat must be turned on, otherwise only public replies count. The timer usually matters from the moment of human handoff. Bots still help: they answer what they can so fewer chats reach the chat queue, collect details so agents reply faster, and tell visitors the honest wait. Outside staffed hours, an after-hours chatbot can state when the team will reply, which is the business-hours SLA in plain words.
FAQ
What is an SLA in customer service? A promised time limit for replying to or resolving a customer request, tracked as a timer in the help desk.
What are the most common support SLA metrics? First reply time, next reply time and resolution time. Some help desks add periodic updates on long tickets.
Does an SLA have to be in a contract? No. Many small teams use SLAs only as internal targets. B2B companies often add them to customer contracts.
What happens when an SLA is breached? The help desk marks the conversation as overdue so it can be handled first, and counts it as missed in reports. Contract SLAs may also carry penalties set in the contract.
Do SLAs count weekends and nights? Only if you choose calendar hours. Zendesk, Intercom and HubSpot all let the timer follow business or office hours instead.
Does a chatbot reply meet a first reply SLA? Usually not. Zendesk counts autoreplies only with its Copilot add-on, so the SLA normally measures the wait for a person.
Related terms
- Chatbot first response time — the wait most chat SLAs are written about.
- Chat queue — where the wait builds when every agent is busy.
- Average handle time — the input that decides how many agents a target needs.
- First contact resolution — the outcome a fast reply should lead to.
- After-hours chatbot — how a business-hours SLA is explained to visitors at night.
- Human handoff — the moment a human SLA timer usually starts to matter.
- Conversation tagging — how breaches can be grouped by topic in reports.
Sources
- Zendesk, Defining SLA policies — support.zendesk.com/hc/en-us/articles/4408829459866, read 15 November 2026. Source of the seven SLA metrics, priority-based targets and the blank-priority rule, policy order, calendar versus business hours, pause behavior, the live chat and autoreply notes, the 15-second minimum and plan availability.
- Intercom, Set SLAs for conversations and tickets — intercom.com/help/en/articles/6546152, read 15 November 2026. Source of the target types, office hours handling, pause options, inbox badge colors, hit/missed/fixed reporting, workflow application and the one-active-SLA rule.
- HubSpot, Set SLAs in the inbox — knowledge.hubspot.com/inbox/set-slas-in-the-inbox, read 15 November 2026. Source of time to first reply and time to close, operation hours versus 24/7, the holiday note, the "Waiting on contact" note and due soon/overdue states.
- Chatbotscape evaluation methodology. /methodology (continuously updated).