Prompting Fundamentals: Five Habits That Beat Any Secret Prompt

Why your prompts matter more than your model

Two people can use the same AI assistant and get wildly different results. The difference is rarely the tool. It is almost always the prompt: what you asked for, what context you gave, and what you did with the first answer. The good news is that prompting is a learnable skill with a small number of fundamentals, not a bag of magic phrases.

Forget the viral "secret prompts" that circulate on social media. Most of them are folklore. What follows is the boring, reliable core that works across every major assistant.

Fundamental one: give context before instructions

Assistants know nothing about you unless you tell them. A prompt like "write an email to a client about a delay" forces the model to guess the industry, the tone, the relationship, and the reason for the delay. It will guess averagely.

Instead, front-load the context:

• Who you are: "I run a small import business in Mauritius."
• Who the reader is: "The client is a long-standing hotel group, formal but friendly relationship."
• The situation: "A shipment is delayed ten days because of a port backlog."
• What you want: "Draft a short apology email that proposes a partial delivery this week."

That takes thirty seconds to type and saves three rounds of correction. A useful habit: if a colleague could not do the task from your prompt alone, the assistant cannot either.

Fundamental two: specify the output, not just the topic

Vague requests get vague answers. Tell the assistant the shape of what you want: length, format, tone, and audience. "Summarise this report" is weak. "Summarise this report in five bullet points for a board that has not read it, plain language, no jargon, under 150 words" is strong.

Formats worth asking for explicitly include tables for comparisons, numbered steps for procedures, bullet points for summaries, and plain paragraphs for anything a human will read as prose. If you will paste the output somewhere, say where: an email, a WhatsApp message to a supplier, a slide. The assistant will adjust register accordingly.

Fundamental three: iterate instead of restarting

The first answer is a draft, not a verdict. Beginners either accept it or delete everything and start over. Experienced users treat the conversation as an editing session:

• "Good, but make the second paragraph firmer."
• "Too formal. This is going to a colleague, not a minister."
• "Keep the structure, cut it to half the length."
• "You assumed the meeting is online. It is in person in Ebene. Adjust."

Each correction teaches the assistant your intent within that conversation. Three quick iterations routinely beat one elaborate mega-prompt, and they are faster to type.

Fundamental four: make it show its reasoning and its doubts

For anything factual or consequential, add one line: "If you are unsure about any of this, say so explicitly" or "List any assumptions you made." Assistants are trained to be helpful and confident, which means they will sometimes fill gaps with plausible fiction. Asking for assumptions and uncertainty surfaces the weak spots before they reach a client.

For decisions, ask for the reasoning, not just the recommendation: "Compare these two options and explain the trade-offs before recommending one." You are not outsourcing the judgment; you are borrowing a structured second opinion. The final call, and the responsibility for it, stays with you.

Build a small library and stop improvising

The final fundamental is not about wording at all. It is about reuse. Whenever a prompt works well, save it: a note on your phone, a document, wherever you will actually look. Most working people need only eight to twelve saved prompts covering their recurring tasks, things like the weekly report summary, the polite payment reminder, the meeting notes cleanup, the translation with tone guidance.

Over a few weeks, refine each saved prompt as you learn what the assistant gets wrong. That library becomes genuinely valuable: it is portable across assistants, it makes your results consistent, and it turns prompting from a daily improvisation into a routine.

None of this requires technical skill. Context first, output specified, iterate freely, demand stated assumptions, and save what works. Master those five habits and you will get more from a mid-range assistant than most people get from the best model on the market.