Most people use AI like an advanced search engine. They ask a question, they get an answer, they're perhaps mildly disappointed by the quality — and the conclusion is that "AI isn't that smart after all." That's the wrong conclusion. The problem isn't the intelligence. It's the communication.
AI models are extremely sensitive to how you formulate your requests. A vague prompt gives a vague answer. A precise, well-structured prompt gives a precise, well-structured answer. And the distance between the two is dramatic — not 10-20% better, but three to five times better, in most cases.
This guide is an introduction to prompt engineering for non-technical teams. No code. No mathematics. Only concrete techniques with direct business application.
Technique 1: Give AI a role, an audience and a context
The simplest improvement you can make is to specify who AI should act as, who the output is for, and what the background context is. Compare these two prompts:
Bad: "Write an email about our new product to a customer."
Good: "You are an experienced B2B sales consultant at a Danish software company selling workflow automation to professional service firms. Write a follow-up email to Thomas Andersen, CFO at a mid-sized accounting firm, who attended our product demo on Tuesday but hasn't replied. During the demo, Thomas was particularly interested in the Microsoft 365 integration and asked about implementation time. The tone should be professional but personal — we've met Thomas twice and know him a little. Maximum 150 words."
That's a dramatically more specific prompt. And it gives a dramatically better output — because AI now knows who it is, who it's writing to, what the context is, and what the constraints are.
Technique 2: Show examples of the output you want
AI is exceptionally good at matching style, tone and format from examples. Use that. Rather than describing what you want, show what you want: "Here are three examples of emails we've sent to clients in similar situations that we're happy with: [Example 1] [Example 2] [Example 3]. Now write a similar email for the following situation: [context]."
This technique — called few-shot prompting — is one of the most effective improvements you can make. It's particularly useful when you have a specific tone or style you want to maintain across communications.
Technique 3: Ask for iterations, not perfection
It's a fundamental misunderstanding to expect the first prompt to give the perfect output. Use AI iteratively: "Give me three variants of this — a formal, a friendly and a concise version." Pick the best element from each and combine. Or: "Revise the previous response and make it 30% shorter while retaining the three most important points."
Iterative prompting isn't inefficient — it's precisely how you truly leverage AI's capacity. Each iteration step refines the output, and you end up with something you would never have achieved on the first attempt.
Technique 4: Be explicit about format and length
AI guesses at the format you want — and the guess is often wrong. Be concrete: "Write this as a bullet-point summary of maximum 150 words, designed to be copied directly into a PowerPoint slide." Or: "Write this as a memo in formal English of maximum one A4 page, with a clear recommendation in the first two sentences."
Format specification is one of the fastest ways to improve output quality. And it saves time — you receive something close to ready for use, rather than something requiring substantial editing.
Technique 5: Use chain-of-thought for complex tasks
For complex analytical tasks — strategic analysis, problem diagnostics, decision support — it's enormously effective to ask AI to think step by step before answering. The technique is called chain-of-thought prompting: "Think step by step about this problem before answering. What are the most important factors to consider? What are the possible answers and their advantages and disadvantages? What would an experienced strategic advisor recommend — and why?"
Chain-of-thought prompting gives markedly better results on complex questions, because it forces AI to structure its reasoning explicitly rather than jumping directly to the conclusion.
Technique 6: Define explicitly what you don't want
Negative instructions are at least as important as positive ones. "Don't write in clichéd business language. Avoid words like 'synergies', 'holistic' and 'proactive'. Write precisely and directly, as an experienced practitioner would phrase it." Or: "Don't include financial figures, legal information or specific product recommendations — we'll add those manually."
Practical templates for immediate use:
For meeting preparation: "You are [your job title] at [your company]. I'm meeting with [name, title, company] about [topic] in [time]. Here's what I know about them: [context]. Prepare: (1) five opening questions designed to surface the real challenges, (2) the three most likely objections and suggestions for how I address them, (3) a brief summary of the most important things to know about their company and situation."
For meeting summary: "Here are my raw notes from a meeting: [raw notes]. Write a structured meeting summary with: (1) concise summary in 3-4 sentences, (2) decisions made, (3) action points with responsible person and deadline. Format: bullet points, formal English."
Prompt engineering is not a technical discipline — it's a communicative discipline. The more clearly you can articulate what you want, for whom and in what format, the better output you get. And that discipline is valuable whether you're working with AI or with people.