In the olden days, you wrote up a sequence of operations to be executed (typically on punched cards, a job), and handed them to the operator to be loaded. The operating system took such a batch of commands to execute off it's input queue, ran them until the end (or something failed), and took the next one. Output was collected (probably printed out) to return to you.
If the above is what is meant, the crucial difference is that in batch processing you get (perhaps the next day) the feedback that the job failed because of a syntax error or some equally irritating glitch. With a command line interface you run one command, can check it's output or correct the command before entering the next one.