Working with AI assistants

A new colleague has joined every office and every phone: the assistant that answers in full sentences, in seconds, in any language you type — writes the letter, explains the tax, drafts the proposal, corrects your code, and never sighs. Used well, it is the most patient apprentice in history, and it is already part of honest work in every field on this shelf. Used carelessly, it is the forward-that-lies with better grammar. This lesson is the difference, and it is now a basic skill, like the keyboard was.

Where it shines: drafts — the first version of a letter, a proposal, a poster's wording, written in seconds and then made yours; explanations — a concept from this shelf said five simpler ways until one lands; translation and tone — the firm email softened, the Pidgin polished for a formal client; and brainstorming — ten names, twenty post ideas, three prices, asked without embarrassment. The working method is the one every editor knows: it drafts, you decide. Your knowledge of the actual work — the client, the market, the truth — is what the assistant does not have and cannot fake. The name on the work is still yours, and the judgement must be too.

Where its confidence must be checked

The assistant does not know when it does not know. It will state a wrong date, invent a policy, cite a law that does not exist — fluently, in beautiful sentences, without blinking. It is the voice-clone lesson in text: fluent is no longer evidence. So the rule is one sentence: everything checkable gets checked before it leaves your hands — the figure, the date, the policy, the legal claim, checked at the source the way the sixty-second check taught. And the privacy line is drawn hard, as always: nothing that belongs to a client goes into the box — not their data, not their invoices, not their logins — and never your own passwords or the codes that die. What you type there has left your house. Treat the chat window like a public street: fine for ideas, dangerous for keys.

Then the honesty question every trade is now settling: must you tell the client? The working answer here: the client pays for an outcome — the site that works, the letter that lands — and tools have always been allowed. What is not allowed is passing off its mistakes as your work, or claiming hours you did not work. Draft with the apprentice, verify with your own eyes, stand behind the result with your own name. Do that, and the strongest tool ever handed to a self-taught worker is simply yours — free, patient, and waiting in the same browser you already know how to use.

• Give it one real task today: the first draft of a letter you have been postponing. Then rewrite it until it sounds like you.
• Adopt the checking habit: every figure, date or claim that will leave your hands is verified at a real source.
• Nothing confidential goes in: no client data, no logins, no OTPs. The chat window is a street, not a filing cabinet.
• Use it as a teacher, not an oracle: ask it to explain anything from this shelf in simpler words, then test the explanation on somebody.

The apprentice, not the master

Every tool on this shelf arrived with the same warning label: it does what you tell it, not what you mean — the spreadsheet, the find-and-replace, the calculator. The assistant is that warning at its loudest, because it fills silence with confidence. The people it will serve best are exactly the people this series has been building since lesson one: those who read before they send, check before they trust, and sign nothing they have not understood. The apprentice is remarkable. Keep the master's chair.