The good, bad, and ugly of AI

Note: This site may show light ads and occasional affiliate links. How that works.

AI is already in the tools people use between school runs and invoices: drafts, summaries, photo tidy-ups, “what does this email mean?” That can be genuinely handy. It can also sand the caution off a decision that needed a slower second look.

The good

Used as a helper that suggests, AI is often fine. It can sketch a reply you then rewrite. It can explain an error message in normal English. It can suggest a spreadsheet fix you still check. The important part is that you stay the person who sends, pays, or publishes.

The bad

It is confident when it is wrong. It will invent a citation, a policy clause, or a “fact” that sounds right. If you paste customer details or passwords into a random chat box, you have also handed them to a system you do not run. Convenience has a memory.

The ugly

Scams get a polish. Voices can be faked. Job applications and profiles can be mass-produced. The ugliest part for small teams is not sci-fi — it is a believable message that arrives at 4:55pm asking for a “quick payment change.”

Stop! Pause! Before you approve a payment change that arrived late on a Friday: Is it urgent? Do you know who sent it? Is it really from them? Verify on a number or email you already trust — never only inside the thread that asked.

My working rule is simple: let AI advise. Do not let it act with money, mailbox identity, or live systems unless there is a narrow job and a human confirm.

A self-check

When did you last let a helper draft something you sent without reading it properly? If you cannot remember checking the facts, read the next one twice before it goes out — especially if money, access, or a customer’s name is involved.