There is a number that surfaces in almost every AI business case I have seen in the past two years: 30 percent. Thirty percent productivity improvement. Thirty percent reduction in process time. Thirty percent fewer hours spent on task X. The figure varies slightly, but the shape is consistent — a confident, round number presented as the reason to implement the tool, the return on the investment, the justification for the budget line.

I am sceptical of this number not because AI tools are ineffective, but because the people presenting it have almost universally confused speed with productivity. They have measured how much faster a process runs and called that an improvement. They have not asked whether the process was worth running at all, whether the output of the process was the right output, or whether going 30 percent faster in the wrong direction constitutes any kind of gain.

The productivity myth is not that AI makes things faster. It does. The myth is that faster is the same as better, and that better process execution is the same as better business outcomes. These assumptions deserve considerably more scrutiny than most organisations are currently giving them.