Dani opens the mail to a letter from a debt collector: it says he's being sued over an old debt, with a court date. His stomach drops, and he does what more and more people do first — he opens an AI assistant. Here are the two versions of what happens next.
✕ The naive way
- He pastes in the whole letter — his full name, address, the account number, the amount — and types: "Do I actually have to pay this? Can they even sue me?"
- The AI answers warmly and confidently: old debts like this are usually "time-barred" and can't be enforced — and it even cites a specific law and a court case to prove it.
- Dani feels a wave of relief. It said exactly what he was hoping to hear.
- Trusting it, he throws the letter in a drawer and skips the court date.
- But "time-barred" was a general possibility, not a fact about his case — and his area's time limit hadn't actually passed. The case it cited didn't exist. He missed the hearing, lost by default, and his wages are now being garnished.
What went wrong
- Pasted private identifiers into a chat box — see what to never tell an AI.
- Took a confident answer at face value on a high-stakes, irreversible matter — exactly what the trust check flags red.
- The citation was hallucinated; he never verified the specific.
- It told him what he hoped — flattery, not confirmation.
- Whether a debt is time-barred depends on local rules and dates he didn't check.
- He let the AI be the decider on a legal matter instead of a human.
✓ The wise way
- Same letter, same fear. But first he removes the specifics — pastes the letter with his name, address, and account number replaced by "[REDACTED]".
- He asks it to explain: "In plain words, what is this letter, what does a court summons mean, and what are my general options? What questions should I be asking?"
- The AI explains the process and raises possibilities — the debt might be time-barred; he can demand validation; he must respond and show up. He treats these as leads to check, not answers.
- He asks it to argue against him: "What am I missing? What's the worst mistake here?" It flags that limits vary by location and that missing the date is the worst outcome.
- He verifies: looks up his local time limit, puts the court date in his calendar, and calls free legal aid. He can raise validation and show up — and he keeps his wages.
What he did right
- Redacted before pasting — kept the help, dropped the exposure.
- Used AI for its strengths — explaining, prepping questions — not as the decider.
- Treated every specific as a lead to verify at the source.
- Made it push back instead of please him.
- Sent the decision to a human (legal aid) and the deadline to his calendar — see health, money & legal.
The difference
Nothing changed but the how. The same assistant that nearly cost Dani his wages also, used differently, helped him keep them. The naive version asked AI to decide a high-stakes question and believed its confident voice; the wise version used AI to understand and prepare, then verified the facts and let a human make the call. That's the whole skill of this kit in one story: understand with the AI, decide with a human — and when it matters, run the answer through the trust check before you rely on it.