Design · The commitment

The human SLA

Software has long sold service levels: we will be up this much, respond this fast. Human oversight needs the same move. A human-in-the-loop SLA turns "a person can step in" from a hope into a commitment with a number behind it.

A human-in-the-loop SLA is a committed Time-to-Human and availability guarantee for the decisions that need a person. It is how meaningful human oversight becomes something an organization can stand behind and a customer can rely on, rather than a line in a policy. Putting a number on it is the difference between claiming a human is available and promising one, on the record, with consequences for missing it. That commitment is a core product of human judgment infrastructure.

What a human SLA commits to

At minimum, a guaranteed Time-to-Human for flagged decisions and an availability window during which that guarantee holds. It can extend to who the human will be qualified to decide, and what happens if the guarantee is missed. The form mirrors the service levels engineers already trust, aimed at the new thing: not system uptime, but human reachability.

Why I am not quoting a number yet

Because a committed figure has to be earned by real operation, not invented for a page. The whole argument here is honest measurement, so I will not publish a Time-to-Human target I have not stood behind in practice. The construct is ready. The number waits for the data, the same discipline I hold for every metric in this work.