When businesses calculate the cost of a manual process, they usually start with the obvious number: the time it takes, multiplied by the hourly rate of the person doing it. That number is real. It's also incomplete by at least half.

The full cost of a manual process has four components that rarely appear on the same spreadsheet — and the gap between what organisations think they're spending and what they're actually spending is usually where the strongest case for agentic AI is hiding.

The four components of true process cost

"The labour line is what you can see. The opportunity cost is what you're actually paying."

How to build a number the board will act on

The goal isn't precision — it's a number that's credible and directionally correct. The approach I use with clients:

In most cases, the number that emerges is 2–4× the figure the business had in mind when they said "it's not that expensive to do manually." That's not a dramatic finding — it's just the full picture.

What the comparison needs to show

The business case for agentic AI isn't just "what does the manual version cost?" It's "what does the agentic version cost, and what does the difference buy?" That means modelling the agentic option with the same rigour — including the governance overhead, the implementation cost, and the ongoing operating cost of the CAgO function that makes the whole thing defensible.

The businesses that get this right don't present a dramatic ROI slide. They present a credible process map with real numbers on both sides — and let the gap speak for itself.