How to Write a Great AI Prompt: A Five-Minute Framework for Busy Executives

Effective prompts have four components: Role, Ask, Context, and Instructions. Not every prompt needs all four, but including each where relevant consistently produces better outputs.

Role: tell the AI what expert perspective to adopt. 'Act as an experienced M&A advisor to a FTSE 250 board' produces better strategic analysis than an unconfigured prompt. The role shapes how the AI frames its thinking and what considerations it prioritises.

Ask: be specific about what you want. 'Analyse this proposal' is vague. 'Identify the three most significant risks in this proposal and suggest one mitigation for each' is specific. Specific asks produce specific outputs.

Context: provide relevant background the AI needs to give a useful answer. Paste in the document, the data, or the situation details. The AI cannot draw on information it does not have.

Instructions: specify format, length, or constraints. 'In bullet points, no more than five' or 'as a formal briefing note suitable for a non-executive director' or 'in plain English without jargon.'

For analytical tasks, the most powerful prompts combine a specific analytical question with relevant context and a request for structured reasoning.

Example: 'You are a strategy consultant advising a FTSE 250 retail board. Here is our current strategic plan [paste text]. Identify three assumptions in this plan that are most likely to be wrong, explain why, and suggest what evidence would confirm or deny each assumption. Present your analysis in a table.'

This prompt gives role, specific analytical ask, context, and output format. The output will be significantly more useful than 'Review our strategy document and give feedback.'

For complex analysis, add: 'Think through this step by step before giving me your final answer.' This instruction activates more careful reasoning in most AI models and reduces the likelihood of superficial or overconfident analysis.

For writing tasks, the most common prompt failure is not providing enough context about audience and purpose. The AI needs to know who will read the output and what it needs to achieve.

Example: 'Draft a one-page executive summary of the attached report for a non-executive director who has ten minutes to read it before a board meeting. The summary should cover: the key finding, the recommended decision, the primary risk, and the next steps required. Tone: professional but direct. No jargon.'

This is specific about audience, purpose, time constraint, structure, and tone. The resulting draft will require much less editing than a vague 'summarise this document.'

For communications where tone matters (board communications, sensitive HR situations, customer-facing messages), add a sample of your own writing and ask the AI to match your voice: 'Here is an example of how I write [sample]. Write in a similar style.'

Great prompting is iterative. If the first output is not what you wanted, do not start over: use follow-up prompts to refine.

'Make this more concise' or 'This is too formal; make it more direct' or 'The third point is not quite right; the real issue is [X]. Revise that point' are all effective follow-up prompts that build on the first output rather than discarding it.

Also effective: 'What assumptions have you made in this analysis?' or 'What have you not addressed that would be important to a board?' These meta-questions reveal gaps and uncertainties that improve the final output.

Save your most effective prompts in a notes file or shared document. A library of tested, refined prompts for your most common tasks is more valuable than starting fresh each time.