The value is downstream.

Why do the usual numbers disappoint?

Time saved is the most common measure of AI value and the least convincing. Saved for whom, on work that anyone downstream actually needed? On its own it is soft, and hard to prove after the fact. Accuracy is a better measure, but not on its own either: accurate at what, feeding which decision? Both describe the task. Neither tells you whether anything got better.

So where should you measure?

One step downstream of where the AI worked, at the hand-off. Real work moves: it flows from one person or team to the next, and value accumulates along the way.

Take a common shape. An AI tool produces the first draft of a case summary. “Minutes saved drafting” is a soft number. But that summary feeds a reviewer, who now starts from something structured instead of a blank page; the reviewer feeds a decision-maker, who reaches a cleaner recommendation sooner. The decision was always the valuable thing, and the business already knew what it was worth. The value of the AI is not the minutes at the start; it is the better, faster decision at the end.

That is the whole shift. You are not inventing a business case, the process already appraised the outcome. You are simply showing that AI reached it quicker, or made it better. An outcome that carries forward carries its own proof.

Where does accuracy come in?

Exactly here. If the draft is trustworthy enough that the reviewer can act on it rather than re-check it, the value compounds down the chain. If it is not, the reviewer redoes the work and the value evaporates at the first hand-off. Accuracy matters not as a score in isolation, but as the thing that lets the next person build on the output instead of second-guessing it. Speed and accuracy turn out to be the same measure seen from two sides: does the next step get something it can use?

Why can you believe the promise?

A measure like this asks something of the firm making it. It only holds from a firm aligned to the outcome rather than the engagement, and that is how we are built. We are senior-led: the people who win the work do it. We are built to make ourselves less needed, not more; the capability stays with you, and the work ends when the outcome is reached. And we are already thinking about the next wave, because the engagement was never the point, the outcome was.

We have measured value this way through every technology wave since 1999, because it is the only measure that survives the wave. The technology changes; the question does not. Did the next person’s work get better? That is where we look, and it is where we have always looked.