availability (SLA)

in English: uptime / availability (SLA)

basicwidespread

Availability shows what percentage of the time an IT system or online service is functioning and actually accessible.

In plain terms

Think of a convenience store open day and night. If it is closed for one day out of 365 due to an unexpected burst pipe, it was available approximately 99.7 percent of the time. The same is measured for online services: in online banking, email, or an AI chatbot.

The number is given as a percentage, and the 'nines' matter. 99 percent sounds good, yet it allows more than three and a half days of downtime per year. 99.9 percent is roughly nine hours, and 99.99 percent is less than one hour.

An SLA (Service Level Agreement) is the contract in which the provider promises this number and describes what you receive if they fail to meet it. Uptime is the measured fact, while the SLA is the committed promise.

In AI, this has become important because more and more companies' workflows rely on an external model. If the model provider goes down, the customer service chatbot, the coding agent, and everything built on it go down with it.

How it works

The provider continuously monitors the system: automated test requests check, for instance, every minute whether it responds and whether it is fast enough.
The uptime is divided by the total time to get the percentage, which is usually calculated monthly.
The contract defines what counts as downtime: planned maintenance and brief, few-second glitches are often excluded.
High availability is achieved with reserves: multiple servers, multiple data centers, and automatic failover ensure that a single machine failure does not stop the whole system.
If the measured value falls below the committed level, the provider typically credits the next bill, but you usually have to request this yourself.

An everyday example

On Monday morning you open your usual AI assistant and only see: 'The service is temporarily unavailable'. The provider's status page shows a red signal indicating the outage, and this downtime at the end of the month lowers the reported availability.

Who is affected

entrepreneur, executiveIf your company's operations depend on external AI or cloud services, the agreed-upon availability determines how much downtime to expect.

developerYou must design the system so that a backup solution exists if an external service goes down. → Code

employeeIf your daily work tool goes down, your deadlines do not shift, so it is worth knowing how reliable it is. → Work & HR

bank customer, investorOnline banking and mobile payment availability depend directly on the stability of the underlying systems. → Money

A common misconception

Myth99.9 percent availability means the service practically never goes down.

In realityThis amounts to roughly nine hours of allowed downtime per year, which could fall on a single workday. Maintenance time is often not even included in the calculation.

Limits and risks

An SLA is not a guarantee against downtime: it only records what you receive if the provider fails to meet the agreed level.
Compensation is usually just a credit to your account, which does not cover lost revenue due to downtime.
Free and consumer plans often come with no commitments at all; only business subscriptions do.
"Available" does not mean usable: a service that is slow or provides erroneous answers may still count as functional in reports.

Similar, but not the same

reliability – Reliability measures how rarely something fails, while availability indicates the proportion of time it is usable, meaning repair speed is also factored in.
response time – Response time indicates how quickly the system replies, while availability indicates whether it replies at all.
service level agreement – The SLA is the contract itself, while availability is just one, albeit the most well-known, metric it records.

Related concepts

Broader

cloud service, IT infrastructure, quality of service

Narrower

high availability, scheduled maintenance, status page

On the edge

service level agreement, malfunction, redundancia, response time, data center

A brief history

The term originates from telecommunications and mainframe computing. Decades ago, telephone exchanges were expected to deliver 'five nines', i.e., 99.999 percent uptime, which amounts to roughly five minutes of downtime per year. SLAs became widespread in the 1980s and 1990s in IT outsourcing and network contracts. With the emergence of cloud services, they also became common for mass-used online services.

Where things stand

As of late 2026, this is a standard element of cloud and AI providers' business contracts, and the major providers show current service status on public status pages. AI model outages are getting more attention because more and more companies' daily work depends on them.

Status: 2026 November

In the news

Appears in 7 Agent Age stories; the curve shows weekly occurrence in the news magazine.

Meta Advertising Week: which AI ad tools are available now and which are in testingOct 19, 2026 · Marketing
IPercept raises $16.5M for AI predictive maintenance as Infinite Uptime and Tech Mahindra partner on physical AI in manufacturingSep 29, 2026 · Ipar
Google AI Mode in Hungarian: availability, AI Overview visibility and SEO measurementSep 21, 2026 · Marketing
Copilot harness reaches general availability in JetBrains; Copilot Studio billing starts September 1Aug 24, 2026 · Code
Azure DevOps remote MCP server reaches general availability without Claude, ChatGPT or Cursor supportAug 21, 2026 · Code
OpenAI adds two Latin American markets for ChatGPT ads, bringing the total to nineAug 17, 2026 · Marketing
Grafana's gcx CLI and MCP server reach general availabilityAug 17, 2026 · Code

Also known as

availability · uptime · availability · uptime · availability · SLA · service level agreement · service level agreement

Sources

This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.