I think the decision to automate is usually based on the capital cost of the automation, with the productivity/consumption ratio of the automation usually being higher than the displaced workers.

Take a robot truck. The difference in energy consumption per mile will be nil, but the auto truck can easily and safely drive 168 hours a week, compared to the death march legal limit for humans, 70 hours.

The per mile cost of the automation system will also likely be close to nil (a few tens of thousands of dollars amortized over a million miles).

So a system that isn't really more expensive (compare the up front cost of the automation to several years of labor) is more than twice as effective.