Which problems remain even when powerful automation tools like Ansible are used?

The tools make DevOps work easier, more predictable and scalable, but they don't solve everything: the scripts and configuration files themselves still need managing, and it still has to be decided how automation enters and affects each stage of the application lifecycle.

Tools such as Ansible wrap shell functions, OS utilities, API calls and other control-plane elements behind one interface, which brings ease of use, predictability, discipline and the ability to work at scale.

Two problems are left over:

1. The automation code itself. Playbooks, scripts and config files are code. They need versioning, review, testing and a single place where the current truth lives, or you end up with the same drift you had with manual changes, just in YAML.
2. The pipeline. Someone has to decide how automation is triggered and how it moves a change from development through testing to production.

Both point to the next idea: store infrastructure as code in version control, and let changes to that repository drive deployment. That is GitOps.

Go deeper:

• Ansible (software) — Wikipedia — agentless, playbook-based configuration management.