Vesmona
Join the private beta

Guide

What is an AI SRE?

“AI SRE” is shorthand for software that takes on the site-reliability work an on-call engineer does: reading signals, investigating incidents, finding root cause and, increasingly, acting on it. The category has split into two shapes, and the difference that decides everything is who holds the keys.

The read-only assistant

One shape investigates, correlates and writes a solid summary, then hands back. It is genuinely useful, and it is also where the work stops: the fix is still yours to type, at whatever hour the pager went off. Nothing it produces can touch a system, which is safe, but it leaves the last and riskiest mile to you.

The autonomy dial

The other shape executes with its own credentials, governed by a configurable threshold. Turn the dial up and it acts on its own; turn it down and it asks first. The catch is that safety has become a setting, and a setting can be misconfigured. The blast radius is whatever those standing credentials can reach.

A third position: execution with structural containment

There is a third option most of the field does not message. Execute, but only through a gateway the customer operates, against an allowlist the customer publishes, with zero standing credentials in the control plane and a policy gate plus human approval on the only path to execution. A signed job from the vendor's own control plane is still refused if it is not on the customer's list. Containment is the architecture, not a setting.

What to ask any AI SRE

  • What credentials does it hold to our systems, and where do they live?
  • Is approval architecture, or a setting someone can turn off?
  • What bounds a mistake, or a prompt injection, once the model is wrong?
  • Can it run entirely on our own hardware, on-premises?
  • Does it auto-remediate security incidents? (The safe answer is no.)

See also: Security · Prompt injection

Run it against your real incidents, as a design partner

Vesmona is in private beta. We're looking for a small group of teams who'll run the platform on real incidents and tell us where it falls short. You get early access, a direct line to the engineering team, and a real say in the roadmap. We get the operational reality (and, where it's earned, the reference) that makes a platform product-ready.

Become a design partner