Two examples from our own delivery work, and what the next step towards agents looks like.
Talk is easy. Progress is specific.
Cloud teams are under pressure to keep up with security updates, keep services running, and reduce risk, all while controlling costs. Many still rely on manual work for database patching, disaster recovery, and routine infrastructure tasks.
In our last post, The Five Levels of Cloud Operations, we described five levels of cloud operations, from Fragmented to Autonomous. It’s easy to agree with a model like that. It’s harder to see what moving from one level to the next actually looks like in a real team.
So here are two examples from our own delivery work. Neither involves AI agents. Both show the step that matters most for most teams today: moving routine work from people to automation.
Example 1: Patching dozens of databases from one job
The starting point. A global clinical research organization runs an Oracle E-Business Suite on Oracle Database. Database security patching was done by server, by hand, by skilled DBAs. Every patch cycle took hours of their time and carried the risk of a missed step.
What changed. We replaced the manual process with a single automated job. It patches many database servers in parallel, with the scripts version-controlled, and the logs kept in one place.
The result. DBA time saved on every patch cycle, a much shorter patching window, and the same steps run the same way every time. We applied the same approach to other routine work, such as moving approved data fixes into production through an automated, approval-based job instead of a DBA running scripts by hand.
Why it matters now. Security updates are arriving more often than ever. Oracle now releases a security patch update every month. When patching runs from one job, a more frequent patch cycle is far easier to keep up with.
Example 2: Disaster recovery that one person can run
The starting point. A global environmental services company runs Oracle E-Business Suite around the clock across many countries. A disaster recovery failover needed several DBAs working for most of a day.
What changed. As part of moving the estate to Oracle Cloud Infrastructure, we automated disaster recovery using Oracle Full Stack Disaster Recovery, run through automated jobs. One job sets up recovery plans. Another runs the failover, including the database switchover and the steps specific to E-Business Suite.
The result. One person can now run a failover, in a fraction of the time and effort it used to take. And because recovery is a repeatable job rather than a long manual exercise, it can be tested properly.
Where these sit on the five levels
Both examples moved from Level 2 to Level 3. The work that used to rely on people following runbooks now runs as automated jobs. People still start or approve them, and that’s exactly as it should be at this stage.
Neither example uses AI agents, and we’re not going to pretend otherwise. But this is the groundwork that makes agents possible. An agent can only help when the process it works with is already consistent and automated.
What the next step looks like
Level 4 is where an agent starts to help with decisions. In the patching example, that could mean an agent checking the results of a patch run, spotting failures and recommending what to do next, for a DBA to approve. In the recovery example, it could mean an agent checking that everything is ready before a drill and flagging anything that would stop it.
We’re already running this kind of agent on our own systems in watch-only mode. It investigates an alert, finds the root cause with evidence and recommends a fix. A person decides whether to apply it.
That’s the journey: get the routine work automated first, then bring in agents that help your team decide, and only then hand over more, as trust grows.
See it live on 19 November
Join our webinar, The Journey to Autonomous Cloud Operations, on Thursday 19 November 2026. We’ll walk through the five levels and show real work running today, across AWS, Azure and Oracle Cloud.