How to write better prompts when working with AI

After years of using AI in business analysis and product management work, I've realized a good prompt doesn't need to be long — it needs 4 components. I call it the R-C-T-F framework: Role, Context, Task, Format.

Role

Tell the AI what "persona" it should adopt. For example: "You are a Business Analyst experienced in writing user stories for digital banking systems."

This shapes the tone and level of expertise in the answer.

Context

This is the most important part beginners tend to skip. Context includes: what you're doing, for whom, why, and what constraints exist (deadline, budget, audience).

Task

Be clear and specific, use action verbs: "compare," "summarize in 5 bullet points," "list the risks," "rewrite in a friendlier tone." Avoid vague requests like "help me with this."

Format

What do you want back — a table, a list, a paragraph, or code? State it upfront so you don't have to ask twice.

Applying it

Weak prompt:

"Write a roadmap for a new product."

R-C-T-F prompt:

"You are an experienced Product Owner (Role). I'm building a personal expense-tracking app for the Vietnamese market, with a team of 4 engineers and a 3-month timeline for the MVP (Context). Propose a roadmap in 3 phases (MVP, expansion, optimization), stating the goal and 3-4 key features for each phase (Task). Present it as a table (Format)."

The difference in answer quality is huge — not because the AI got "smarter," but because you gave it enough information that it no longer has to guess.

One small but useful habit

When the result isn't quite right, don't rewrite from scratch — edit parts of the previous prompt instead: "Keep the structure above but cut it in half" or "Redo this in a more formal tone." This saves time and helps you learn which part of the prompt is actually driving the result.