ChatGPT Commands That Actually Save You Time

Most people use these tools badly, and it is not their fault. They type a question the way they would type it into a search engine, receive a bland paragraph, and conclude that the technology is overrated. The difference between a useless answer and a genuinely useful one is almost always the instruction, not the model.

The underlying principle is simple. These systems respond to constraint. Give them a role, a context, a format and a standard to hit, and the quality changes immediately. Here are the commands I use most often, with the reason each one works.

Act as a sceptical reviewer of the following plan. List the three assumptions most likely to be wrong and what evidence would disprove them. This is the single most valuable prompt I know. It converts a machine that wants to agree with you into one that argues, and it surfaces the flaw you have stopped being able to see.

Summarise this document in three layers: one sentence, one paragraph, and ten bullet points with the numbers preserved. Layered summaries are far more useful than a single blob of text because they let you decide how deep to go, and insisting on the numbers stops the vagueness that summaries usually drift into.

Rewrite this to be forty per cent shorter without losing any factual content. Tell me what you removed. The second sentence is the important one. It forces the model to show its edit rather than quietly deleting the point you cared about.

Explain this to me as though I know the field well but not this particular topic. No analogies to sport or cooking. Precision comes from telling the system what level to pitch at. Banning the standard analogies removes about half the padding.

Here are my notes from a meeting. Produce the action list only, each with an owner and a date, and flag anything where the owner was not stated. Ideal admin task. The flagging instruction is what makes it reliable, because it stops the model inventing an owner to complete the pattern.

Give me the five strongest arguments against my position, ranked by how difficult they are to answer. Then answer them. Useful before any negotiation, board paper or difficult conversation. You will find at least one objection you had not prepared for.

Turn this into a table with these exact columns, and write unknown where the source does not say. Structure forces honesty. Without the unknown instruction you will get plausible filler in every empty cell.

Plan this trip as a timed itinerary with realistic travel times between each point, and mark anything that is likely to be closed on the day of the week I am visiting. Generic itinerary requests produce fantasy schedules. Demanding travel times and opening constraints makes the output usable.

Read the following and tell me what a lawyer, an accountant and a customer would each object to. Multiple-lens review catches problems a single pass misses, and it costs you one prompt instead of three meetings.

Ask me five questions before you answer, then answer. This is the prompt most people never try, and it transforms long or complex tasks. Instead of guessing at your context, the system gathers it first.

Draft this in British English, plain and direct, no exclamation marks, no phrases like delve, leverage or in today's fast-paced world. A style ban list is the most effective quality control available. Name the words you never want to see and they stop appearing.

Two habits multiply everything above. The first is to keep a personal instruction set — your role, your business, your tone, your standing rules — so that you are not restating context in every conversation. The second is to always ask for the reasoning behind a recommendation, then check one fact from it. These systems are confident when they are wrong, and the only defence is verification.

Used well, this is not a novelty. It is the difference between spending Sunday evening on admin and spending it with your family.