Library › Leadership AI › Product managers

Extract launch risks from a pre-mortem into a JSON array

  • For Product managers
  • You practice Catching mistakes and made-up facts
  • Time about 5 minutes
  • Level Advanced

About this exercise

After a pre-mortem session, convert the comments into a JSON array in one code block. Each object has exactly four keys: risk, likelihood (exactly high, medium or low in lowercase), owner, mitigation. Include only risks that someone actually named as a possible failure of the March 12 launch. Put owner as "unassigned" when no one volunteered. Output only the code block with exactly 7 lines inside (one object per line), and no text before or after.

Common mistakes when you ask AI for this

  • Including non-risks. The AI adds the logo comment or pricing copy as risks. Try: Say: "Only comments that name a way the launch could fail."
  • Missing the update. The AI keeps throttling as high from Sana's first comment. Try: Say: "Apply Sana's later note: throttling is now low."
  • Guessing owners. The AI gives unowned risks to the PM. Try: Say: "Use \"unassigned\" when nobody volunteered."
  • Explaining outside the code. The AI adds an intro and notes around the JSON. Try: Say: "Output only the code block, no text before or after."
  • Long pretty-printed JSON. The AI spreads each object over many lines. Try: Say: "One object per line, at most 7 lines inside."

What a strong prompt includes

  • Names the four keys
  • Asks for one code block
  • Filters to real risks
  • Handles unassigned
  • Applies the update

Practice this now ↑

Related exercises