Output Formats

Recall: Context and Few-Shot Prompting

In Lesson 4, you learned that showing AI examples (few-shot prompting) is often more effective than explaining. This principle is especially powerful when it comes to output formats.

By the end of this lesson, you’ll be able to:

• Request specific output formats confidently
• Choose the right format for different tasks
• Use templates to get consistent, usable outputs

Why Format Matters

Imagine asking AI to summarize a document. It gives you three paragraphs of prose.

But you needed bullet points for a presentation slide.

Now you’re reformatting manually—exactly the kind of work AI should save you from.

The fix: Specify your format upfront.

Common Output Formats

Here are the formats you’ll use most often:

1. Bullet Points / Lists

Best for: Summaries, quick reference, scanning

Summarize this article as 5-7 bullet points,
each one sentence max.

2. Numbered Lists

Best for: Steps, rankings, sequences

Give me a 10-step plan to launch this product,
numbered 1-10 in order of execution.

3. Tables

Best for: Comparisons, structured data, analysis

Compare these 3 options in a table with columns for:
Name, Cost, Pros, Cons, Best For

4. Headers and Sections

Best for: Long documents, reports, reference material

Write this report with clear headers:
## Executive Summary
## Key Findings
## Recommendations
## Next Steps

5. Code

Best for: Programming, automation, technical tasks

Write a Python function that [does X].
Include comments explaining each section.

6. JSON / Structured Data

Best for: When output feeds into other systems

Return the results as JSON with this structure:
{
  "name": "string",
  "score": number,
  "tags": ["array", "of", "strings"]
}

The Format Formula

Here’s a reliable pattern for specifying format:

“Present your response as [format] with [specific structure]. Each [item] should include [requirements].”

Examples:

For a comparison:

“Present your response as a table with columns for Feature, Tool A, Tool B, and Winner. Each row should cover one key feature.”

For action items:

“Present your response as a numbered list. Each item should start with an action verb and include an owner and deadline.”

For analysis:

“Present your response with these sections: Summary (2-3 sentences), Key Points (bullet list), Recommendations (numbered). Use headers for each section.”

Format Templates You Can Copy

Template: Meeting Notes → Action Items

Convert these meeting notes into action items.

Format each action item as:
- [ ] [Action verb + task] | Owner: [name] | Due: [date]

Group by project or topic with headers.
Only include items that require someone to DO something.

Meeting notes:
[paste notes here]

Template: Comparison Analysis

Compare [A] vs [B] vs [C].

Create a table with these columns:
| Criterion | [A] | [B] | [C] | Notes |

Criteria to compare:
1. [criterion 1]
2. [criterion 2]
3. [criterion 3]

After the table, add a "Bottom Line" section with
a 2-sentence recommendation.

Template: Document Summary

Summarize this document in three formats:

1. ONE SENTENCE: The core message in one sentence
2. TWEET: Under 280 characters, conversational
3. EXECUTIVE SUMMARY: 3-5 bullet points with key takeaways

Document:
[paste document here]

Matching Format to Task

Quick Check

You need AI to analyze customer feedback and identify themes. Which format request would work best?

A) “Analyze this feedback”

B) “Analyze this feedback and present as: 1) Top 5 themes as a bullet list with frequency counts, 2) Notable quotes table with columns for Quote, Theme, and Sentiment, 3) Summary of recommended actions”

(Answer: B gives AI a clear, useful structure to fill)

Common Format Mistakes

Mistake 1: No format specified You get whatever AI feels like generating. Fix: Always specify format for anything beyond simple Q&A.

Mistake 2: Conflicting instructions “Keep it brief but comprehensive with lots of detail.” Fix: Be clear about priorities—length OR depth.

Mistake 3: Vague format words “Make it nice” or “format it well” means nothing. Fix: Specify exactly what “nice” means—bullets? Tables? Headers?

Mistake 4: Forgetting who’s using it Technical JSON when you needed plain text. Fix: Think about your downstream use before requesting format.

Practical Exercise

Take this messy output request and add format specifications:

Before:

“Give me ideas for our team offsite”

Your improved version: (Consider: How many ideas? What format? What details per idea?)

Example answer:

“Give me 8 ideas for our team offsite. Present as a numbered list. For each idea include: Activity name, Time required, Team size it works for (use ranges like 5-10), and One-sentence description. Group ideas by type: Indoor vs Outdoor.”

Key Takeaways

• Specify format upfront to avoid reformatting work
• Match format to task: tables for comparison, bullets for scanning, sections for reports
• Use templates for consistent, reusable outputs
• Show examples when the format is complex

Up Next

Lesson 6 covers the common mistakes that derail AI interactions—and how to fix them. You’ll learn to diagnose why prompts fail and what to do about it.