Prompts as Work Briefs

The brief analogy

A prompt is a work brief for an AI assistant. Just as a vague brief to a human contractor produces vague work, a vague prompt produces vague output. "Write me something about marketing" will produce generic filler. "Write a 300-word email to our enterprise clients announcing our new API integration, emphasizing reduced onboarding time, in a professional but approachable tone" will produce something useful.

The single most important insight in prompt engineering is this: the AI does not know your context, your audience, your goals, or your standards unless you include them. Every piece of context you leave out is a gap the AI fills with its own assumptions — which may not match yours.

Think about what you would include if you were briefing a competent freelancer who has never worked with you before. You would explain the audience, the purpose, the tone, the format, any requirements or constraints, and what a good result looks like. The same elements make an effective prompt.

Components of an effective prompt

Role: Tell the AI what perspective to adopt. "Act as a senior financial analyst reviewing quarterly results" produces different output than "Act as a marketing copywriter creating social media posts." The role shapes tone, vocabulary, depth, and approach.

Context: Provide the background information the AI needs. This might include the audience, the situation, relevant facts, the purpose of the output, and any source material. The more relevant context you provide, the more targeted the output.

Task: State exactly what you want the AI to do, using a clear action verb. "Analyze," "Summarize," "Draft," "Compare," "Evaluate," "Create," "Explain." Ambiguous tasks produce ambiguous results.

Constraints: Define boundaries. Word count, tone, topics to avoid, required elements, format requirements, claims that must be verified, and audiences that must not be alienated. Constraints focus the output.

Output format: Specify how you want the result structured. A bulleted list, a table, a narrative paragraph, a formal letter, a casual email, a decision matrix, or a pros-and-cons analysis. Format instructions dramatically change the usability of the output.

Common beginner mistakes

The most common mistake is being too vague. "Help me with my presentation" gives the AI nothing to work with. "Create a 5-slide outline for a board presentation about Q3 revenue shortfall, covering causes, impact, corrective actions, and timeline" is specific enough to produce useful output.

The second most common mistake is not iterating. The first output is rarely the final version. Treat AI interaction as a conversation. Review the output, identify what is wrong or missing, and refine the prompt. Three rounds of iteration usually produce dramatically better results than accepting the first response.

The third mistake is treating AI output as final without review. AI generates plausible text, not verified truth. Every output intended for customer, financial, legal, or public use should be reviewed by a human who understands the subject matter.