Most automation projects start too big[4.1]

The pattern is always the same. A team decides to "automate the business", maps everything for two months, and ships nothing. By the time the first workflow is live, the people who asked for it have stopped believing in it.

The projects that work start embarrassingly small. One process. The one someone complains about every week. Lead triage, invoice matching, order exceptions. Something with a clear input, a clear output, and a person currently doing it by hand.

Why small wins

A single process gives you three things a grand plan never does.

• A measurable before and after. You know what 40 hours of triage costs.
• A short feedback loop. If the automation gets something wrong, you hear about it on Tuesday, not next quarter.
• Trust. The second project is easy to sell when the first one visibly works.

Where AI actually fits

Most of the value is not in the model. It is in the plumbing: getting the data in, handling the messy cases, and deciding what happens when the system is not sure. The rule we build to is simple. If the system is confident, it acts. If it is not, a human sees it. Nothing falls on the floor.

Start with one process. Ship it in weeks. Let the results argue for the next one.