Giving AI the structure it needs to succeed
"Be specific" is good advice — but it understates what's really happening. The quality of what AI produces is almost entirely a function of the quality of what you give it. Structure is the skill.
Think of it this way: AI has enormous capability but no context. It doesn't know who you are, what you're trying to accomplish, who the output is for, or what constraints matter. Every piece of context you provide focuses its capabilities on your actual situation. The more relevant constraints you provide, the closer the output gets to what you actually need.
A well-structured request has five elements. You don't always need all five — but knowing what they are lets you diagnose exactly why a response missed the mark and fix it precisely.
The five elements
Role — who is the AI being?
Telling AI to take on a specific role changes how it frames everything. "You are a skeptical editor" produces very different feedback than "you are an encouraging writing coach." Neither is wrong — they're useful for different purposes. The role shapes the lens.
Context — what is this actually for?
The more AI understands about your situation, the more relevant its output will be. Who is the audience? What do they already know? What decision does this output need to support? Context is not padding — it's the information that transforms a generic response into a useful one.
Task — what exactly do I want?
The most obvious element, but often underspecified. State the task precisely. Not "write something about X" but "write a two-paragraph explanation of X that answers: why does this matter right now?" The more precisely the task is defined, the less guessing the system does.
Format — what should it look like?
AI defaults to whatever format seems most common for a given type of request. That default is often wrong for your situation. Specifying length, structure, prose vs. bullets, formal vs. casual — prevents you from getting output that needs reformatting before it's usable.
Constraints — what should it avoid?
Negative constraints are often more powerful than positive instructions. Telling AI what not to do — what tone to avoid, what assumptions not to make, what topics to stay clear of — directly prevents the most common failure modes. "Don't use jargon" eliminates jargon more reliably than "use simple language."
"When a response disappoints you, the fix is almost always in one of these five elements — something that was missing or underspecified. Diagnosing which one is the skill."
Build a prompt — interactively
Fill in each element to watch your prompt assemble in real time. Leave any field empty to see how the prompt weakens.