AI terms explained for Non technical and Managers

Ask an AI agent to “just send that email” and sometimes it stalls out — not because it’s dumb, but because it doesn’t have hands. It can’t open Gmail, click compose, or hit send. And yet the same AI can post to Slack, update a Shopify order, or drop a row into a spreadsheet without you touching anything. The difference between those two situations is one word: API.

The window, not the kitchen

The clearest explanation of this I’ve come across is a drive-thru analogy from Cathryn Lavery, who writes The Non-Technical Technical Dictionary (nontechnical.dev) — a plain-English glossary of tech and AI terms, built for people who didn’t come up through a CS degree.

Her framing, paraphrased: your AI pulls up to a window. Bolted to the wall is a menu — send this email, update this product, fetch these orders, post this message. The AI reads the menu, places an order, and the kitchen does the work. It never walks inside. It never logs into your account. It just orders off what’s posted at the window, and the result comes back out.

That’s an API. It’s the window between your AI and somebody else’s software.

The menu is also the fence

The part people miss is that the menu isn’t just a convenience — it’s the safety mechanism. Whatever isn’t printed on the menu simply can’t be ordered. If “delete every customer” isn’t listed, your AI has no way to ask for it, no matter how it’s prompted. The company that built the API decided exactly what’s on offer, and everything else stays locked in the back, out of reach entirely.

That’s why businesses are comfortable opening a window at all: they’re not handing over the keys to the kitchen, just a menu they wrote themselves.

Every window needs a key

To place an order, your AI has to prove it’s allowed at the window — that’s the API key, a long random string that says “this request is mine, charge it to my account, let it through.”

One rule matters more than the rest: never hardcode that key into your code. Code has a habit of ending up on GitHub, and a lot of GitHub is public. Leave a key sitting in plain text and you’ve effectively taped a blank check to the window — there are bots that do nothing but scan public repos for exactly this, and a leaked key can be in use by someone else within minutes. Keys belong in a separate, hidden config file, not in the source itself.

Chatbot vs. employee

Once the window clicks, so does a lot of what agents actually are. Pulling last week’s orders and summarizing them, adding a subscriber the moment they sign up, posting to Slack when something breaks — none of that is the AI being clever with language. It’s the AI placing orders at windows that were already open.

As Lavery puts it: “Without APIs, you’ve got a chatbot. With them, you’ve got an employee.”

Credit where it’s due

The drive-thru framing above comes from Cathryn Lavery’s API entry on The Non-Technical Technical Dictionary — a 101-term dictionary that explains AI and dev jargon in plain English, one analogy at a time. If you’re building with AI without a technical background, it’s worth bookmarking (nontechnical.dev); the whole project is built on the idea that better vocabulary leads to better questions, which leads to better things built. Cathryn also writes at littlemight.com. Full credit to her for putting it this clearly — go read the original.