A user would hand in his 'job' to the computer's operator, which was called 'submitting' the job. The operator would pass the job to the computer, and, when it was done, return the output, etc produced by running the job back to the user.
Early computers (which were very expensive) tended to charge for all resources used by the job - CPU time, main memory usage, printer output, etc. An accounting system measured all the resources used in running a job, added up all the costs, and charged the result to the account given when the job was submitted.
A job was often run as a process, but not all batch OS's supported the process abstraction. Some early interactive OS's retained the term, but used it to refer to a process, or a user's top-level process, etc.