Mid-market Indonesian news outlet
True 24/7 monitoring, run by an agent
We made an AI agent the on-call engineer for a news outlet we manage: every 30 minutes it checks the site and the systems behind it, diagnoses anything broken, writes the fix, redeploys, and logs what it did. It ran that loop for weeks.
Check cadence
every 30 minutes
Continuous unattended run
weeks at a stretch
Repair mode
self-diagnose → fix → test → deploy
Human involvement
escalation boundary only
The operation
A mid-market Indonesian news outlet publishes around the clock, with a public website readers rely on. News doesn't keep business hours; the systems behind it run continuously, and any silent failure means a stale or broken site for every reader until someone notices.
That's the classic small-team problem: the system needs 24/7 attention, and hiring a night-shift engineer for one product makes no sense. They brought us in to keep it running.
The standard approach, and why we didn't use it
Standard practice is monitoring dashboards plus an on-call human: alerts fire, someone wakes up, reads logs, ships a fix. The monitoring is automated; the understanding and the repair are not. A person who can diagnose and fix at 3 AM is the expensive part, and it doesn't scale for a small operation.
So we made the diagnosing and fixing an agent's job too.
How the sentinel works
A coding agent (a Claude-based instance with access to the codebase, logs, and deploy pipeline) wakes every 30 minutes and runs the same loop a good engineer would: check that the site is up and publishing, read the logs for anything abnormal.
When something is broken, it doesn't page a human first. It investigates: reproduces the error, reads the relevant code, writes a fix, tests it, and deploys. Every action is logged. When a problem is ambiguous enough that a wrong guess could make things worse, it stops and escalates to a person. That boundary is the same human-in-the-loop principle we build into every client system.
The result ran for weeks at a stretch with no human in the monitoring loop: failures caught within a cycle, routine breakage repaired and redeployed by the agent itself, humans involved only at the escalation boundary.
Why this matters if you're a client
This is the engine inside our Care tier. When we say your workflow is monitored 24/7, we mean an agent checks your system every half hour, fixes what it safely can, and wakes a human for what it can't, with a readable log of everything it did.
This is not a demo. The same monitoring loop has kept a real production news operation running, unattended, for weeks.
From the sentinel's log
One incident, no humans woken
A representative cycle: article pages start erroring at 3:30 AM, the agent finds the cause, ships a tested fix, and the site is healthy again before the next check. A person reads the log in the morning.
Want a system like this in your business?
We build one workflow at a time, deliver it into your accounts, and monitor it after. Pricing is published up front.
Start with a free workflow audit
A 45-minute call. We map three processes agents can take over, estimate what each would save, and quote what it would cost. No obligation.