Output Formats
Design outputs that are easy to review, compare, automate, and reuse.
Learning Objectives
- Match the output format to the job-to-be-done.
- Use tables, JSON, checklists, and memos intentionally.
- Add sorting, length, and column rules to reduce cleanup.
Format is part of the instruction
The output format decides how easy the result is to review, compare, automate, and share. Asking for "a summary" creates cleanup work. Asking for a table with named columns creates something operational.
Use tables when comparing options or extracting repeated fields. Use checklists for execution. Use JSON when another system will consume the output. Use a short memo when a human needs to make a decision.
Format details that matter
- Name the columns or sections.
- Set a length limit.
- Define sorting rules.
- Say what to do when evidence is missing.
- Ask the model to flag uncertainty instead of filling gaps.
The best format is the one that matches the next step. A sales rep may need an email and CRM note. A PM may need themes, quotes, and suggested actions. An engineer may need hypotheses, tests, and fixes.
Examples
Format selection
Use a markdown table for comparison, JSON for automation, a checklist for execution, and a short memo for a decision.
Practice Exercise
Format conversion
Take the same source text and ask for three formats: table, checklist, and executive memo. Compare which format is most useful.
- Each format has explicit fields.
- The output includes no unsupported claims.
- The chosen format matches the downstream use.
Mini Prompt Templates
Table Extractor
You are a product analyst. Context: [CUSTOMER_FEEDBACK] Task: Extract recurring themes. Format: Markdown table with columns Theme, Evidence Quote, Frequency, Suggested Action. Constraints: Sort by frequency and do not invent quotes.