Incident record
Example- What the AI read
- What it proposed
- Who approved it
- What changed, and in which pipeline run
- How to undo it
- What the runbook learned
AI-Assisted Operations: AI that changes nothing until an engineer approves
Your cloud’s own agent reads each alert against your logs, recent changes and runbooks, and drafts a fix. A named engineer approves before anything changes.
One alert, read in context
- An alert arrives, and the agent starts at any hour (on Google Cloud, an engineer starts it from the alert).
- It reads the metrics, the logs, recent changes, what depends on what, and the runbook.
- It proposes a cause, a fix and a way to undo it.
- A named engineer approves, and your pipeline applies the change.
- The record is kept in your account, and the runbook gains a line.
Ready before the first alert
- Signals in one placeThe metrics, logs and traces the agent can query.
- A map with ownersTags that say what each resource is and who owns it.
- Runbooks it can readThe same steps for the agent and the engineer.
- A read-only roleScoped to the resources your team supports.
- A way to reach youProposals land in your chat and ticket tools.
What every incident leaves you
Kept in your account, so it is yours if we leave.
The agent on each cloud
- The agent
- AWS DevOps Agent
- How we set it
- Read-only, with directed actions left off
- Where the record lives
- AWS CloudTrail, and the agent’s journal archived in your account
AWS and Azure bill their agent outside our fee, and Azure adds a fixed charge while the agent exists. Google Cloud’s investigations need Premium Support or access from your account team.
Where you will meet it
- Cloud Build & Launch (opens in a new tab)The AI assistant drafts and checks, and an engineer reviews every output.
- Data and AI (opens in a new tab)Small agents for cloud operations, built as scoped projects.
- Managed Services (opens in a new tab)An alert starts the AI first pass at every service level.
- Audit & Compliance (opens in a new tab)Findings are ranked, given an owner and carried into a remediation plan.
Questions
Does the AI change anything on its own?
No. It proposes and a named engineer approves. The one exception is a runbook you approved in writing for a known, safe fault, from the Professional level.
Who pays for the agent’s time?
Your cloud, outside our fee, and each prices it its own way: AWS by the second, Azure with a fixed charge plus use, and Google Cloud through Premium Support while its investigations are a preview. We scope it with you and keep the cost in view.
Do you sell an AI product?
No. We set up and run the clouds’ own agents, and build small agents as scoped projects.