Product manager’s AI 101

First, trust your tech team, but here is how remove the worries of feeling like an idiot not knowing anything at all. Know that this is all simplifications but will give you a mental model of what is going on which will help you make decisions faster.

AI is prediction; AI is Automation.

But, they are separate.

AI prediction — tools that try to predict the future but instead look at probability compared with training data, which often overlaps but is not the same.

AI automation — tools that help reliably complete actions that are well structured based of varying inputs.

Interestingly, there is an overlap, as predicting what should happen based off of training data is not a huge step away from predicting the next word you expect in a sequence — this is the premise of how LLMs work (using the training dataset known as the entire internet).

Types of AI

Here I try to describe what they are simply, when to use them and the most common methods. I cover Supervised, Unsupervised, Reinforcement, and Deep Learning, plus Generative AI.

Supervised Machine Learning — getting a machine to run through your labelled training data many times until it builds its own way of solving the problem. You can think of it trying to build a flow diagram or math functions that best fit solving the problem. In the below, imagine trying to build a classifier that will say if some input is…