Your first thirty days with AI

Most advice about getting started with AI is a list of tools. That is the wrong place to start.

Here is what I would tell you to do over thirty days if you were sitting across from me and you had decided this is the year. It costs nothing until week four. By day thirty you will have a real number that tells you whether to keep going.

Week one: find out where your time goes

Do not buy anything. Do not sign up for anything. Just watch.

For five working days, keep a note on your phone. Every time you do something and think I do this all the time or this should not be my job, write one line. Do not think about it. Just write it down.

By Friday you will have somewhere between fifteen and forty lines. About six of them will be the same thing said a few different ways. Those six are your list.

Two rules for reading that list:

• Skip the hard things. Hard usually means judgment, and judgment is not what you automate first.
• Circle the things that repeat and get skipped when you are busy. Dull, frequent, and unreliable. That is where the money is.

Week two: write one of them down

Take the one item on that list that bugs you the most. Write down how it actually happens today, step by step. What starts it. What happens next. Who does each part. Where it gets stuck.

Twenty minutes on paper. I wrote a whole post about this exercise, because it is the step everyone skips and the one that decides everything.

You are looking for one thing: the step that is supposed to happen every time, but only happens when someone remembers.

By the end of week two you should be able to finish this sentence: When ______ happens, ______ should happen next, every time. Right now it happens about ______ percent of the time.

If you can say that out loud, you are ahead of most businesses that have already spent money.

Week three: do it by hand, on purpose

Now you can start. You still are not buying anything.

For two weeks, do that missing step by hand, every single time. Set an alarm. Put it on a whiteboard. Whatever it takes to make that step happen every time using nothing but stubbornness.

This feels like a detour. It is the most useful two weeks of the month. Three reasons.

You find out if it matters. Some steps you were sure were costing you money turn out not to change anything. Better to learn that now, for free, than after you paid for something.

You learn what the step really involves. Doing it on purpose twenty times shows you the exceptions. The times you would not send the reminder. The customer you would call instead. Those exceptions are exactly what has to be spelled out later, and you cannot know them from memory.

You get a starting number. Count what happens. Replies, callbacks, payments, whatever the step is supposed to produce. Write it down.

That number is the most valuable thing you will make this month. Everything after this gets measured against it. And it belongs to you, not to somebody's sales pitch.

Week four: decide, then buy small

Now you know three things you did not know on day one. What the step is worth. What it involves. What it produces when it happens every time.

Now you can shop, and you are a different buyer than you were a month ago. You are not asking what can AI do for me. You are asking what will make this one thing happen without me, and what does it cost.

Buy the smallest thing that does that. Not a platform. Not a bundle. The one thing.

Then keep counting. Same number, same way, for the next thirty days. If it does not beat what you did by hand, you learned something real, and you are out one month instead of a year.

What you will have on day thirty

• A list of the six things that eat your week
• One of them written down clearly enough to hand to someone else
• A real number for what that step produces when it happens every time
• Either a small system that works, or proof that this one was not worth automating

Any one of those is worth the month. Having all four puts you ahead of most businesses that have been "looking into AI" for a year.

The part people skip

Weeks one through three have no software in them. That is why most people jump straight to week four.

I get it. Buying something feels like progress. Watching yourself work does not. But every time I have seen this go wrong, it went wrong in the same place. Somebody bought a tool for a problem they never wrote down. Then during setup they found out nobody in the building agreed on what the problem was.

The three boring weeks are the whole thing. The purchase is the easy part. It goes well or badly depending on what you knew before you made it.