Andrew Bolis Details a 6-Step Framework for Crafting Superior GPT-5 Prompts

In a recent LinkedIn post, Andrew Bolis shares a structured approach to prompt engineering, aiming to help users achieve more expert-level responses from AI models like GPT-5. Bolis outlines a six-step framework designed to enhance clarity, precision, and the overall quality of AI-generated output.

Deconstructing the Prompting Framework

Andrew Bolis emphasizes that effective prompting is crucial for unlocking the full potential of advanced AI. He breaks down the process into distinct, actionable steps, each serving a specific purpose in guiding the AI’s performance. According to Bolis, a well-crafted prompt moves beyond simple questions to become a detailed set of instructions.

“Assign GPT-5 a defined role to shape its perspective. Framing expertise improves clarity and focus in answers.”

This initial step, which Bolis terms ‘Role,’ is about establishing the AI’s persona or area of expertise. By assigning a specific role, users can influence the AI’s tone, knowledge base, and the angle from which it approaches a task. Bolis argues that this framing is fundamental to achieving relevant and focused results.

The Core Components of Effective Prompts

Bolis’s framework continues with the ‘Task’ component, where the desired action is clearly articulated. He stresses the importance of making the outcome specific and measurable. Following this, the ‘Context’ step involves providing necessary background information, limitations, or relevant examples to ensure the AI has the full picture.

“Provide details that help GPT-5 deliver accurate results. Include background, limits, and examples where relevant.”

The ‘Reasoning’ step is highlighted as a method to encourage the AI to ‘think’ before it answers. Bolis suggests instructing the model to outline its thought process or a stepwise approach, which can lead to more robust and logically sound responses.

Specifying Output and Conditions

Further refining the prompt, the ‘Output’ step dictates the desired format of the response. Whether bullets, tables, narrative, or steps, clarity here prevents misinterpretation. Bolis notes that this ensures the information is delivered in a usable and easily digestible manner.

“State clearly how the response should be delivered. Indicate format (bullets, steps, tables, or narrative).”

Finally, the ‘Conditions’ step sets boundaries, defining the scope, length, or specific stopping points for the AI’s response. This helps manage the AI’s output and prevent it from generating excessive or irrelevant information.

“Set boundaries for scope, length, or stopping points. Tell it when to stop or move to the next step.”

By adhering to this 6-step structure – Role, Task, Context, Reasoning, Output, and Conditions – Andrew Bolis asserts that users can significantly improve the quality and utility of their interactions with AI, moving towards more expert-level answers.