API stands for Application Programming Interface. That tells you essentially nothing. Let me try again.
The waiter analogy (I know, I know)
You're at a restaurant. You know what you want to eat. The kitchen knows how to make it. But you don't go into the kitchen and cook it yourself, and the chefs don't come out and take your order. There's a waiter. The waiter takes your request in a format the kitchen understands, and brings back what was made for you.
An API is the waiter. It's the messenger between two systems that need to talk to each other but aren't directly connected. One system makes a request. The API translates it. The other system does something and sends back a response.
A real example
Weather apps. Your phone's weather app doesn't have its own satellites. It sends a request (via an API) to a weather service — "give me the current conditions for Adelaide" — and the weather service sends back data your app displays. The app and the weather database never actually touch each other directly. The API is what makes them interoperate.
Every time you pay online, a payment API is handling the transaction between the merchant and your bank. Every time you log in with Google, an authentication API is doing the handshake. APIs are everywhere.
Why it matters for AI
This is where it gets relevant for us. Most AI tools can be accessed via API — meaning you can connect them to other software without building the AI yourself. Want your customer management system to automatically draft follow-up emails using Claude? That's an API integration. Want to trigger an AI summary whenever a new document appears in a folder? API.
You don't have to be a developer to understand this. You do need to understand it exists, because a lot of the most useful AI setups involve connecting tools together — and APIs are what make that possible. When someone says "hook this up to an API," they mean: let these two systems talk to each other.
Do you need to use APIs directly?
Probably not at first. Most tools you'll use have API integrations built in already. But understanding the concept means you can start to see which things are connectable — and that's where the real compounding starts to happen. Two systems talking to each other, automatically, without you doing the manual work in between. That's the goal.