What is the basic idea behind how an automation tool like Ansible works?

You declare the desired end state; the tool checks the current state and only changes what differs, then stops.

* The tool gathers facts first and only converges when the current state differs from the desired one. *

1. Define the end state. For example: "a web server is installed and running."
2. Gather facts. The tool looks at the system as it is now (the Ist-Zustand, actual state), which starts out unknown.
3. Compare. Is the current state already the desired state?
   • Yes: do nothing.
   • No: converge, meaning make exactly the changes needed (for example install the web server).
4. End in the desired state, and finish.

The key point is step 3. The tool does not blindly run a list of commands. It decides whether a change is needed, which is what makes it safe to run again and again.

Go deeper:

• Ansible playbooks (official docs) — how tasks express desired state, what "changed" vs "ok" means, and why runs are safe to repeat.
• What is Infrastructure as Code? (Red Hat) — contrasts the declarative "describe the end state" approach with imperative step lists.
• Ansible (software) (Wikipedia) — overview of the architecture and design goals.