AI can generate code. Production is another problem.
Generating an application is becoming increasingly easy.
Running it reliably is not.
A production system still needs to answer questions such as:
- What happens when a service goes down?
- Who is allowed to access a production resource?
- How do we detect abnormal behavior?
- How do we roll back a bad deployment?
- How do we prevent an agent from executing a dangerous action?
- How do we control cloud costs?
- How do we know whether an AI-generated infrastructure change is actually safe?
This is where I think DevOps is evolving rather than disappearing.
The interesting shift: from automation to autonomous automation
Traditional DevOps automation is usually deterministic.
A CI/CD pipeline might look like:
Commit → Tests → Security Scan → Build → Deploy
We define the steps and the system executes them.
AI agents introduce something fundamentally different.
Instead of only executing predefined steps, an agent may be able to:
Observe → Reason → Decide → Act → Evaluate
Imagine an agent detecting increased latency, analyzing logs and traces, identifying a problematic deployment, proposing a rollback and potentially executing the remediation.
That is incredibly powerful.
It is also potentially dangerous.
The more autonomy we give these systems, the more important identity, permissions, observability, security policies and human oversight become.
This might actually make Platform Engineering more important
For me, this is the most interesting consequence.
If developers increasingly use AI agents to build and deploy software, someone still needs to design the platform in which those agents operate.
That platform needs:
- IAM and least-privilege access
- secure execution environments
In other words:
AI may automate operational work, but it also increases the need for well-designed operational boundaries.
The question may therefore change from:
“Who operates the infrastructure?”
to:
“Who designs the platform that humans and AI agents are allowed to operate?”
That sounds a lot like Platform Engineering.
DevOps engineers may become platform architects for humans + agents
I am currently studying AWS, Terraform, Kubernetes, CI/CD and cloud infrastructure, and this shift has changed how I think about learning DevOps.
Learning individual commands is important.
But understanding the principles behind them seems even more important.
Why least privilege?
Why Infrastructure as Code?
Why observability?
Why immutable infrastructure?
Why security gates?
Why controlled rollbacks?
An AI agent may eventually execute many of these operations faster than a human.
But someone still needs to understand why the guardrails exist and how they should be designed.
That may become one of the most valuable skills for the next generation of Cloud and Platform Engineers.
My prediction
I don't think AI will eliminate DevOps.
I think it will eliminate a growing amount of manual DevOps work.
And that distinction matters.
Engineers may spend less time executing repetitive operational tasks and more time designing:
platforms, policies, architectures, reliability mechanisms and guardrails for autonomous systems.
DevOps might move from:
“automate everything we can”
to:
“build safe platforms where humans and AI can automate responsibly.”
And personally, I find that future much more interesting.
What do you think?
I would love to hear from people already working with AWS, DevOps, SRE or Platform Engineering:
1. Do you think AI agents will reduce the demand for DevOps engineers?
2. Which DevOps skill becomes MORE valuable as agents become more autonomous?
3. Would you allow an AI agent to automatically roll back a production deployment without human approval? Why or why not?
I'm especially curious about the third question.
👇 DevOps engineers, SREs, Platform Engineers and AWS builders — what would you trust an AI agent to do in production today?
#AWS #DevOps #PlatformEngineering #SRE #CloudComputing #AI #AIAgents #AWSCommunity #CloudEngineering
Loading article