Today we stay inside this week's AI playbook, but we narrow it to machine learning. People hear machine learning and picture a room full of math professors. You do not need that. Machine learning just means a program that gets better at a guess by looking at past examples. By the end of this you will know how to train a small one on your own records, the kind every business already has sitting in a spreadsheet.

First, find a guess you already make by gut. Which leads turn into customers. Which orders are likely to be returned. Which invoices get paid late. You already half-know these patterns from doing the work. Machine learning is just teaching a program to make that same guess, faster and on every record, instead of only the ones you happen to look at.

Second, gather the history with the answer already attached. Pull last year's leads and mark each one: became a customer, yes or no. That marked column is the teacher. A model learns by reading hundreds of past rows where the answer is known, so it can spot the same shape in a new row where the answer is not known yet. No history, no learning. The data you already have is the fuel.

Third, open a free Google Colab notebook and use a plain library called scikit-learn. Ten lines of Python loads your spreadsheet, splits it into a part to learn from and a part to test on, and trains the model. You do not write the math. You call a function named fit. The hard part was already built and given away years ago. Your job is to bring the right data to it.

Here is our self-proof. Inside our studio we run a small scorer that rates each video before it ships, and it learns from the ones we marked good and bad in the past. It is the same idea, pointed at our own work. We did not buy a fancy platform for it. We trained it on our own marked examples, and we let it flag the weak ones so a person looks closer.

So machine learning, stripped down, is four moves: name a guess you already make, label your old records, train a small model in a free notebook, and check it against rows it never saw. Start with one column you wish you could predict. You can see how we use our own learned scorer, live and in the open, at office.temerarii.xyz.

Chapter six. Staff training: how to build teams that actually execute on AI, starting with one task each person runs every week.