SLA stands for Service Level Agreement. It is a contract that defines what a vendor or internal team promises on three dimensions: availability of the service, response time to customer messages, and resolution time for issues. It also defines what happens if those promises are missed, which usually means service credits or, in rare cases, contractual exit rights.
In support specifically, SLA usually refers to response-time and resolution-time targets per priority level, often per customer tier.
What goes into a typical support SLA
A typical SLA has four components. First, priorities: how tickets are classified, usually critical, high, medium, low. Second, response time per priority: critical tickets get a first reply in fifteen minutes, low tickets within one business day, that kind of thing. Third, resolution time per priority: how long until the issue is fully closed. Fourth, exclusions: what does not count, like force majeure, customer-caused outages, or scheduled maintenance.
Customers paying more usually get tighter SLAs. Enterprise contracts often add a guaranteed uptime percentage (the SaaS provider's availability SLA), penalties for breach, and a named contact.
SLA versus SLO
Inside a vendor, the same numbers are usually called SLOs (Service Level Objectives). An SLO is the internal target. An SLA is the customer-facing promise. SLAs are always more conservative than SLOs, because no vendor wants to commit to its own ceiling.
If a vendor says they will respond in fifteen minutes (SLA), they are probably internally targeting five minutes (SLO). The gap is the buffer for bad days.
How AI changes SLA economics
The classic SLA assumes humans are doing the work. Hitting a fifteen-minute response time around the clock requires people working nights, weekends, holidays. The cost is significant, and small teams cannot do it without expensive coverage.
An AI agent that handles tier-one questions in seconds shifts the math. The human team focuses on tickets the AI cannot handle, which is where their judgement actually matters. The SLA on those tickets can be tighter because the queue is shorter.
In Keloa
In Keloa, the AI handles routine questions in seconds, which makes first-response SLA easy on the channels we support: widget, email, and Slack. Resolution SLA depends on the team behind the AI, but the AI shortens the queue so humans can focus where it counts. For our own service guarantees, see pricing and the DPA on security.