What AI is and where it breaks — Learn AI

DPR's friendly robot, standing in for your AI intern — Meet your AI intern

Start with the basics. ChatGPT, Claude, and Gemini aren't magic brains. They're text tools trained on huge piles of writing.

Think of AI as a fast, well-read intern. It writes, explains, and summarizes in seconds because it has read a huge slice of the internet — but it wasn't in yesterday's meeting, and it doesn't know your pricing sheet. When it doesn't know, it guesses in a calm, confident voice. That's the danger: it's useful because it sounds sure, and risky for the exact same reason.

What the intern read

A huge slice of the public internet
How to write, explain, and summarize
Patterns from millions of documents

What it never learned

Your meeting yesterday
Your pricing sheet and your customers
What's true right now

What it's good at

Turning rough notes into clean writing, like emails you don't want to write twice.
Explaining something complicated in plain words, so you're not stuck decoding jargon.
Coming up with options when you're staring at a blank page.
Tidying and reformatting text when you don't want to do cleanup by hand.
Getting you a decent first draft fast, so you can mark it up instead of starting cold.

Where it'll trip you up

Exact facts, names, numbers, and dates. It can invent them, and people call this "."
Recent events. Most tools only know up to a unless they're connected to the web.
Math and counting. It guesses unless it has a calculator, so don't trust a total without checking it.
Anything where a confident wrong answer is expensive. That includes legal, medical, money, or customer promises.

Try it

Trust it, or check it?

Five things AI just did for a small business. For each one, decide: use it as-is, or double-check first? Tap your answer to see the rule.

It turned your rough, messy notes into a short, polite customer email.

RuleTrust it as-is (after a quick read). It's just reshaping words you already gave it — that's its strength, and it's low-risk.

It says a clause in your contract "is standard and fine to sign."

RuleDouble-check it. That's legal judgment. AI can sound completely sure while being wrong — send it to a person, not the AI.

It listed "this year's three latest changes" to a tax rule.

RuleDouble-check it. Recent facts and dates are exactly where AI guesses. Confirm each one at the official source.

It added up a column of invoice amounts and gave you the total.

RuleDouble-check it. AI guesses at math unless it's using a calculator. Never trust a total you didn't verify.

It gave you three subject-line options for your newsletter.

RuleTrust it as-is. Ideas and options are a real strength. Pick the one you like — nothing here has to be "correct."

Tap an answer on each card to reveal the rule.

A quick before / after

Risky: "Is this contract clause legal?" You'll get a confident yes or no, and you can't tell if it's wrong.

Smart: "Explain what this clause means, and what I should ask a lawyer." Now it's explaining. You keep the judgment.

The rule is simple. Let AI do the reading and drafting. You keep the deciding.

Mistakes almost everyone makes

Treating it like Google. It isn't looking things up. It's predicting good-sounding text, so facts still need checking.
Assuming it knows your business. It doesn't, until you tell it. (That's Lesson 4.)
Believing it because it sounds sure. Confidence isn't proof. A calm answer can still be wrong.

Let AI do the reading and drafting — you keep the deciding.

Takeaway

AI is a fast intern that read a lot and remembers nothing about you. It won't admit doubt often enough. Use it for drafting, explaining, and ideas. Double-check anything that has to be exactly right.

Try this today: ask an AI tool to explain one thing from your own work. Use a policy, invoice, or customer email. Notice what it gets right and where it's guessing.

← All lessons Next: Your first 15 minutes with an AI assistant →