How will we know whether it actually worked?

Agree this before anything is built. Afterwards, everyone remembers the old process as either much worse or much better than it was.

Take the baseline first

Before we build, write down the current number. How many hours a week. How many days to turn a quote around. How many errors last month. It takes an afternoon and it is the only thing that makes the later conversation honest.

Almost nobody does this, including me for the first year, and then everyone argues from memory.

Measures that hold up

• Time to complete the specific task, measured the same way before and after.
• Volume handled without adding people.
• Error or rework rate — often the real saving, and usually invisible in the before picture.
• Adoption. Are people actually using it in week eight? This is the one I care about most, because everything else is theoretical if the answer is no.

Measures that do not

Satisfaction surveys taken during the excitement of a launch. Hours "saved" calculated by multiplying an optimistic estimate by everyone in the building. And any figure your supplier produced without showing the working.

The blunt version

Someone framed the whole thing for me in one question, two years out: is 55k a good investment or a bad investment?

That is the only measure that matters, and you should be able to answer it with a number you took yourself, not one I gave you.