Infrastructure automation is not proved by a single successful run. Take a small Ansible playbook that provisions a service, applies configuration, and restarts it only when needed. Run it twice on clean test hosts, change one variable, and run it again. The candidate should show which resources changed and leave stable systems alone.
Then induce a failed task and practise recovery. Compare secret handling, inventory management, review of proposed changes, and how the tool works with your current CI. If existing roles must be rewritten, estimate that cost rather than calling the alternative free. The choice is sound when repeated runs and failures are predictable.
Akousa