Measure the wait, not the wow

There is a metric most AI initiatives are quietly optimised for, and it is the gasp in the room. The demo lands, heads nod, the budget is approved. Months later nothing measurable has changed, and no one is quite sure why.

The reason is that the gasp and the gain live in different places. The gasp comes from novelty. The gain comes from removing the wait — the hours work spends in a queue, the loops of rework, the lead time a customer actually feels.

Lead time is the honest number

Lean teaches you to watch a unit of work travel from request to delivered, and to time it. Most of that clock is not spent working — it is spent waiting. Waiting for approval, for context, for someone to pick the task back up. That waiting is invisible on a demo stage and obvious on a value-stream map.

When you measure the wait, the right intervention often turns out to be unglamorous. Sometimes it is a model. Just as often it is removing a handoff, killing an approval step, or putting the right context in front of a person before they ask for it.

Instrument before you celebrate

We agree the baseline before anything is built: how long does this take today, how often is it redone, where does it stall? Without that number, every result is a story, and stories are easy to tell and impossible to defend.

It is a less exciting way to start. It is also the only way to know, a quarter later, whether you bought a capability or an applause line.