Note

Start with one function

How to pick the first function to hand off, and why one working function beats five half-built ones.

When an owner decides to put AI to work, the first question is usually where to start. The answer I give is one function, with one owner, chosen on purpose.

A function is one team or process family, such as quoting, scheduling or billing. Workflows sit inside it. A function has an owner, a way of working today, and a standard for what good looks like. That makes it something you can build around and hand over.

How to pick the first one

Look for the function that meets three tests.

It routes through you. Ask where people wait on your answer. If a question about quotes, scheduling or approvals ends up on your desk every time, that function is a candidate.

It repeats. A function that runs weekly gives you a way to check the work. One that happens twice a year does not.

It has someone who can own it. The person who will run it after the handoff should be on your team, and they should be able to say yes or no on the calls that need a person. If you can't name that person, the function is not ready yet. Finding that out early is useful.

Why only one

Some functions take a week to set up and some take three months. Nobody knows which until the work is scoped. If you start five at once, you find out about the slow ones late and you have five things half done.

One function gives you something finished. The named owner runs it without me, and it holds to a written standard. That is the handoff test, and it is the only proof I count. Once one function passes it, the next one goes faster, because your team already knows how a handoff works.

What to expect from the work

Each function gets a written standard, a handoff to the named owner, and a short weekly memo on what was done and what is next. I don't count hours. The question is whether the function runs.

If you want to see the steps I follow for each function, they are on the method page.