
“AI saves each person five hours a week.”
I hear that number a lot. My question is always the same. Where did the hours go?
Usually nobody knows. They got soaked up by the week. More email. Longer meetings. A slower Friday.
That’s the AI ROI myth. Saved hours aren’t a return until someone uses them. And even then, they’re only half the story.
AI pays back in two ways.
Efficiency: the same work, with less effort. Effectiveness: better work than before. Fewer mistakes. Safer sites. Better calls.
Most business cases count the first. The second is often worth more.
Take a legal team. Efficiency is a cleared backlog and contracts back in three days, not ten. More work stays in-house, so the outside counsel bill falls. Effectiveness is risky clauses caught before signing.
Or a crusher on a mine site. Here the return is mostly effectiveness. Failures caught days early. Fewer unplanned stops. Safer repairs. But watch the false alarms. A few of those and the crew stops listening.
So before go-live, I decide two things. Where the saved hours go. And what “better” means.
I record the baseline first. I put a dollar figure on both, using numbers finance trusts. And finance signs off the benefit, not the AI team.
If the number can’t survive the CFO, it isn’t a return.