I think the hardest to do will be the secretary and the paralegal, so those are the ones I'm going to do.
The secretary works for an insurance company. He may have very advanced clerical skills, but little or no knowledge of insurance. So how can we rotate him to other jobs in a way that will actually be useful?
I think the way to do it is to retrain all clerical staff and claims adjusters to perform both roles---so that each claims adjuster files their own paperwork from start to finish. This constitutes both job rotation and job enlargement. This would have a number of benefits for customers---if one person is in charge of your file from start to finish, it's less likely to get mixed in, misunderstood, or simply lost. This would also allow the claims adjusters to build closer relationships with their clients, since each one would have fewer clients and do more work for each one.
That would provide one of the most powerful intrinsic motivations of all: social contact. Interacting socially with other human beings is such a powerful motivation that it can override literally everything else, even survival as a soldier falls upon a grenade to save his comrades. It is the one thing we don't take away from prisoners unless we want to give them the maximum possible punishment. Allowing claims adjusters to build closer relationships with clients instead of seeing them as numbers on a page could radically improve their well-being, and thus we might not even have to raise salaries in order to attract the better talent we would need. If we did need to raise salaries, we could also afford to raise premiums, because customer satisfaction is also likely to significantly improve.
In terms of the job characteristics model, we have definitely expanded skill variety and improved task identity. The plan could be structured to increase autonomy and task significance as well. Feedback was probably already fairly good, but now it is directly tied to the outcome of each claim, rather than specifically to the paperwork portion of many claims.
The paralegal does research. She probably knows the law fairly well, but she is not licensed as a lawyer, so she can't go to trial by herself. We can't simply rotate lawyers and paralegals, and lawyer wages are so astronomical that we really can't afford to use the lawyers' time at anything except that which absolutely requires their attention. So we seem to be in a bind.
What I think we can do is at least rotate the paralegals with each other, so that each one performs multiple roles in the firm within the broader umbrella of paralegals. This is job rotation. They could be assigned to whole cases: Instead of one paralegal for research, another for discovery, another for writing motions, we would have each paralegal do research, discovery, and motions on a single case. This would give the cases more unity, and reduce the likelihood of gaps or redundancies that could have been introduced by cycling between different hands.
This could provide some intrinsic motivation as well, as it would tie each paralegal to her own case, which she can be more emotionally invested in. While the overall workload would not decrease, the number of names, faces, and events she must be familiar with would, because she isn't trying to keep track of three cases at once. Still, we'll probably have to raise salaries, and it's not clear that this would actually be cost-effective for the firm.
In terms of the job characteristics model, the main benefit is in terms of task identity and task significance---this is a single case, a unified whole, and its outcome depends upon your work. Autonomy is unlikely to improve much, as paralegals are by necessity strongly bound to the authority of the licensed lawyers they work under. But feedback could improve, as each paralegal would be accustomed to following the whole case to its conclusion, rather than moving on as soon as their particular task is done.