William Stanley Jevons, writing about British coal in 1865, pointed out something nobody expected. More efficient steam engines did not reduce coal consumption. They increased it.

Efficiency made steam power cheaper per unit of work, which made it economical for uses that had not been worth it before, which expanded demand faster than the efficiency saved.

Making something cheaper to do means more of it gets done.

In a business

Automation. Automating a report does not give back the hours it used to take. It makes reports cheap, so more get produced, and the time returns to roughly where it was. The gain is real — you have more reports — and it is not the gain that was budgeted.

Faster delivery. Cutting the time a piece of work takes usually results in more work of that kind being sold or accepted, rather than in slack.

Cheaper meetings. Video calls removed travel and made meetings nearly free, and the number of meetings rose to consume the saving. Parkinson's law and this one are pointing at the same thing from different angles.

Cheap storage and cheap code. Both got radically cheaper, and both are consumed in vastly greater quantity, with the total spend roughly flat and the complexity higher.

The useful form

When you make something more efficient, decide in advance what the saved capacity is for, and commit it before it evaporates.

If automating a process is meant to free a day a week, the day needs a destination on the calendar or it will be absorbed by more of the thing that just got cheap. That is the same condition comparative advantage has: the gain from freeing time is entirely in the redeployment, and without one there is no gain to collect.

It is also a reason to be skeptical of business cases that count savings. The saving is usually real per unit and usually does not appear in the total, because volume moves.