Prompting Is a Superpower, But Not for the Reason You Think

There is a lot of excitement around 'prompt engineering', and sometimes it is made to sound like a secret language. It is not. At its most useful, prompting is the ability to express a goal clearly, provide relevant context and tell the AI what a good answer should look like.

Compare two prompts. The first says, 'Tell me about careers.' The second says, 'I am a 17-year-old Kenyan student who enjoys biology and mathematics, prefers practical problem solving and wants to explore careers that could suit me. Compare five possible directions, explain the education pathway for each and give me simple experiments I can try before choosing.' The second prompt works better because the thinking behind it is better.

A simple structure is Goal, Context and Constraints. Goal: what are you trying to achieve? Context: what does the AI need to know? Constraints: what rules, limits or requirements should it follow? You can also add the audience, desired format, examples or evidence you want included.

Good prompting also means knowing that the first answer is rarely the end. Ask follow-up questions. Tell the AI what is weak. Add information it did not have. Ask it to compare alternatives. Ask it to challenge you. The conversation itself becomes part of the thinking process.

One of the most useful prompts is, 'What assumptions have you made?' Another is, 'What would a smart person who disagrees with this say?' You can also ask, 'Which parts of this answer should I verify?' These questions reduce the temptation to treat AI as an authority.

That matters because AI can produce what are often called hallucinations: statements that sound plausible but are inaccurate or unsupported. It may get a date wrong, invent a source, misstate a statistic or create a quotation that does not exist. If the information matters, check it.

There is also an educational reason to prompt carefully. If you always ask AI to 'do this for me', you may become more efficient while learning less. Instead, ask it to teach. Try, 'Do not give me the answer yet. Show me the first step and then ask me what I think comes next.' Or, 'Critique my draft but do not rewrite it until I have tried to improve it myself.'

This turns AI into a tutor, coach or practice partner rather than a substitute for effort. You can use it to simulate interviews, test your understanding, rehearse a debate, role-play a customer conversation or learn a difficult concept at several levels of depth.