Set Godin, in his The ever-worsening curse of the cog says something profound:
The end result is that it's essentially impossible to become successful or well off doing a job that is described and measured by someone else.
I still think managers can describe parts of a useful, non-cog job. But jobs where a person's qualities, preferences, and non-technical skills don't matter are going to be outsourced or automated.
Here's an example: If you're looking for a manual black box tester, why? If that person has to wait until the product exists to test, what value does that person add to development of the product? That's a cog position, unless you can articulate some specific value that this person will add, that can't be automated or outsourced.
Another example: if you're looking for a fill-in-the-blank language (Cobol, Java, C#, any language) programmer, what value is that person supposed to add? Why aren't you looking for people who live to break software in various ways, if you're looking for testers? Why aren't you looking for people who solve a variety of problems using a variety of functional skill expertise? (If you're hiring for other positions, think about the variety of functional skills required by the position.)
The more varied the required functional skills, the more valuable your position is, and the less cog-like it is. The less cog-like, the less of a commodity it is. So it's clear why candidates want not-cog positions. But why should employers care? Because you're in business to deliver results and build capacity. You can't deliver results without people who can adapt their skills to your context. You can't build capacity without people who can learn new things. Two very un-cog-like reasons. Without delivering results and building capacity, the organization dies, eventually. It's very expensive in the medium- and long-term to hire for cog positions. It's cheaper to higher fewer non-cogs, but people who can deliver for you.