Skip to main content
NexPatch
Operations · Service Level Agreement

SLA for AI systems.

The SLA for AI systems at NexPatch sets out how fast we respond to an incident, staggered by severity, and during which hours that commitment applies. It also sets out the escalation path if an incident isn't resolved in time, and the written follow-up after every major incident.

Health Check
Rolling Update
Failover Cluster
Verified Uptime

Severities and what they mean

Not every incident is equally urgent, so we distinguish three severities. This classification happens at the start of every incident together with the reporting contact on your side, so it is clear from the outset which deadline applies and who needs to be informed internally.

SeverityMeaningResponse time
P1, criticalSystem fully down or a core function unusableSet in the contract
P2, highA significant function is limited, ability to work is affectedSet in the contract
P3, normalMinor issue with no immediate effect on operationsSet in the contract

A general figure here would not be worth much: for an internal tool with low impact the deadline has to look different than for a system with a direct effect on your customer-facing business. So we set the specific response time per service tier during the assessment and put it in writing in the operations contract. What this page does define are the severity levels and the escalation path that apply.

Service hours

We distinguish two service-hour models. Under business hours, we are reachable during normal working hours on business days, which covers most internal tools whose use is also concentrated in working hours. Under the extended model, coverage runs longer each day, suited to systems with a direct effect on customer-facing business - an agent handling incoming cases outside normal office hours, for example. Which model fits your system, we decide together with you during the assessment; see also Operations.

The escalation path

Every contract has a named contact on our side responsible for your system, plus a deputy in case they are unreachable. This prevents an incident from getting stuck with a single unavailable person. If an incident is not resolved within the agreed deadline, it escalates automatically to our management, who then get directly involved in resolving it, without you having to trigger that escalation yourself. After every P1 or P2 incident, we produce a written follow-up covering cause, action taken and, where useful, suggestions for avoiding a similar incident in future, so an incident also produces an improvement, not just a fix.

How the SLA connects to our operating target

Our general operating target is 99.9 percent availability across the systems we support. That is a target for our own operations, not an individual guarantee for any single customer or system - the exact contractual commitments for your case are in the relevant operations contract. Detail on operations overall, including what is included and what it costs, is at Operations; detail on the underlying infrastructure at private AI infrastructure.

Frequently asked questions