What AI Cannot Do

Lesson 6 · taught by Pip

We've spent a few lessons admiring what AI does well. Now let's be honest about where it stops. Knowing the edges of the map keeps you from sailing off them.

The plain idea: a chat assistant is very good at working with words, and that is all it is doing. It has no eyes, no ears, no stake in the outcome, and no memory of your business unless someone gives it one. It can sound certain about things it has no way of actually knowing.

Think of it like a brilliant new hire on their first morning — sharp, well-read, eager — but who has never walked your floor, met your customers, or felt the weight of getting it wrong. Knowledge is not the same as experience, and experience is not the same as responsibility.

Three things it simply can't do

It has no senses. AI doesn't see your shop, hear the tone in a caller's voice, or smell that the kitchen is behind. It only knows what's typed to it. If a customer is quietly furious but politely worded, the AI reads the words, not the mood. A person reads the room. The AI reads the transcript.

It has no judgment of its own. It can lay out options, but it doesn't care which one is right. It won't lie awake worrying about a decision, and it has no instinct that says "this feels off, let me double-check." It follows patterns in language — confidently, even when the confident answer is wrong. The instinct that something doesn't add up is yours, not its.

It has no built-in knowledge of you. Out of the box, an AI knows a great deal about the world in general and nothing specific about your prices, your policies, or your Tuesday. It can't know your refund rule unless you've taught it. (This is exactly why a Portmint assistant is trained on your business — that's what the one-time Training & Testing Fee is for: turning a clever generalist into your specialist.)

Why this matters for accountability

Here's the part that's easy to skip. When something goes wrong, a person is answerable for it. The AI is not. It cannot be sorry, cannot make it right, cannot learn its lesson the way a teammate does. So the rule is simple: an assistant can draft the email, suggest the answer, and handle the routine — but a human stays accountable for anything that carries real weight. Money, promises, safety, a tricky customer — those stay on a person's desk.

Used this way, AI isn't a replacement for your judgment. It's a tireless first mate that frees you to spend your judgment where it counts. The beacon doesn't steer the ship. It just helps you see.

Your turn

Make a quick two-column list: tasks where a wrong answer is annoying but cheap (drafting a reply, looking something up) versus tasks where a wrong answer is costly (refunds, legal, safety). The first column is fair game for AI; the second always keeps a human in the loop.

Next up, we'll look at one of the sneakiest failure modes: when AI states something false with a perfectly straight face. 🔦