Skip to main content

Prompt Preparation

"Good prompts are written before you start typing."

Preparation > Engineering​

You've heard of "prompt engineering," the skill of crafting effective prompts. But the most impactful skill is prompt preparation: the work you do before typing a single word.

The difference:

  • Prompt engineering is how to phrase your request
  • Prompt preparation is what to think through before requesting

Most bad AI interactions fail before the prompt is written.

The Preparation Framework​

1. Clarify the Goal​

Before opening any AI tool, answer:

  • What exactly do I want to accomplish?
  • What does "done" look like?
  • How will I know if the output is good?

Bad: "I need help with this code" Prepared: "I need this function to handle null inputs gracefully, returning an empty array instead of throwing"

2. Gather Context​

What information does AI need?

  • Relevant code, documents, or data
  • Constraints and requirements
  • Examples of good outputs
  • Things that have already been tried

Bad: "Fix this bug" Prepared: [paste the function] + [paste the error] + [paste example input that fails] + "This worked before commit abc123"

3. Define Constraints​

What boundaries exist?

  • Style/convention requirements
  • Technology constraints
  • Performance requirements
  • What to avoid

Bad: "Make this faster" Prepared: "Optimize this query. Must work with PostgreSQL 14, can't add indexes, should reduce response time from 2s to under 500ms"

4. Anticipate Follow-ups​

What will you need next?

  • Tests for this code?
  • Documentation?
  • Alternative approaches?

Consider asking for these upfront rather than in follow-up messages.

Common Preparation Failures​

The Vague Goal​

What happens: Generic advice that doesn't solve your specific problem Fix: Spend 30 seconds articulating exactly what you need

The Missing Context​

What happens: AI makes incorrect assumptions, wastes your time Fix: Include the actual code/data/document, not descriptions of it

The Unconstrained Ask​

What happens: Technically correct answer that doesn't fit your situation Fix: State your constraints explicitly

The Premature Request​

What happens: You ask, get an answer, realize you asked the wrong question Fix: Think through what you're really trying to accomplish

Preparation Templates​

For Code Tasks​

GOAL: [specific outcome]

CONTEXT:
[paste relevant code]

CONSTRAINTS:
- [technology requirements]
- [style requirements]
- [what to avoid]

DONE WHEN:
- [success criteria]

For Writing Tasks​

GOAL: [what the writing should accomplish]

AUDIENCE: [who will read this]

TONE: [formal/casual/technical/etc]

LENGTH: [approximate]

INCLUDE:
- [required elements]

AVOID:
- [what not to include]

EXAMPLE OF GOOD OUTPUT:
[if available]

For Research Tasks​

GOAL: [what I'm trying to learn/decide]

WHAT I ALREADY KNOW:
- [current understanding]

SPECIFIC QUESTIONS:
1. [question 1]
2. [question 2]

FORMAT: [how to present findings]

Preparation Time Investment​

Task ComplexityPrep TimePrompt Time
Quick question10 seconds5 seconds
Standard task1-2 minutes30 seconds
Complex task5-10 minutes1-2 minutes
Project-level15-30 minutes5 minutes

The more complex the task, the more preparation pays off.

Signs You Skipped Preparation​

  • AI asks clarifying questions you should have anticipated
  • You immediately need to correct assumptions
  • The output is technically correct but useless for your situation
  • You spend more time in follow-ups than you would have spent preparing
  • You find yourself thinking "that's not what I meant"

Practice​

Before your next AI interaction:

  1. Pause before opening the AI tool
  2. Write down your goal in one sentence
  3. Gather all relevant context
  4. List your constraints
  5. Then write your prompt

After a week of this practice, it becomes automatic, and your AI interactions become dramatically more effective.