Free preview8 min

Complete vs Incomplete Instructions

Reviewed by Human · Updated July 13, 2026

✗ Wrong way

Write something about our new app for the launch.

Four decisions are missing, so the model made all four for you. 'Something' — a tweet? A press release? A blog post? For whom? How long? What tone? The result: a 600-word generic blog post praising features the app doesn't even have. Technically obedient. Completely useless.

✓ Better way

Write a launch announcement tweet for our budgeting app, FinchWallet. Audience: young professionals who've never budgeted before. Max 240 characters, friendly tone, one emoji, end with the link finchwallet.app.

Every decision is now yours: the action (write a tweet), the scope (FinchWallet's launch, for budgeting newcomers), the constraints (240 characters, friendly, one emoji), and the output form (a single tweet ending with a link). Nothing left to guess.

What Makes a Complete AI Instruction?

A complete instruction answers four questions before the model has to ask them. **Action**: The single clear verb that says what to do — write, summarize, rank, rewrite, translate. **Scope**: Exactly what material the action applies to — and what it doesn't. **Constraints**: The limits the output must respect — length, tone, things to include or avoid. **Output form**: The shape of the answer — a tweet, a table, five bullets, one paragraph. Miss one element and the model fills the gap with a statistical default. Later in this curriculum you'll also teach the model by showing it examples — that technique is called few-shot prompting, and it gets a full course of its own.

Action. Scope. Constraints. Output form. Four elements, one habit.

Same request, two instructions

✗ The half-instruction

Summarize this meeting.

The action is clear, but that's it. Summarize for whom? How long? Focused on what? The model will produce a medium-length, everything-included summary — which serves nobody in particular.

✓ The complete instruction

Summarize this product-team meeting transcript in 5 bullet points for our CEO. Focus only on decisions and deadlines. Plain language, no jargon.

All four elements are present, so the output is shaped before it's even generated. The model knows the reader, the length, the focus, and the format.

Live Demo — AI Terminal

Watch what the complete version of that prompt returns. Notice how every element you specified shows up in the shape of the answer.

Your prompt

Field note

On a client onboarding project last spring, I asked the model to 'draft the welcome email for the new users.' I got back a warm, polished email — for a SaaS product. My client ran a physical gym. Nothing in my prompt said so. The action was there, the output form was there, but the scope lived entirely in my head. Now I never send a prompt without what I call the stranger line: one sentence naming who this is for and what business it belongs to. Ten seconds of typing, and it ended an entire category of confident, useless drafts.

Your turn

Complete this incomplete instruction: 'Fix my resume.' Add all four elements, one line at a time, then combine them into a single prompt.

Reflect

Compare your combined prompt to 'Fix my resume.' Which of the four elements changed the outcome most?

end of module

You've finished this module.

Mark it complete to earn your XP and keep your streak alive.

Progress saved locally · Sign up to earn XP