Check AI-written FAQ answers against the product spec sheet

About this exercise

An AI tool wrote 5 FAQ answers for your new blender's product page. Before publishing, you want exactly 5 numbered lines, one per answer, labeled Accurate, Wrong, or Not in spec sheet, with the spec sheet's version after any Wrong label. Use only the spec sheet (including its revision note), in plain text with nothing else.

Common mistakes when you ask AI for this

• Missing the revision. The old numbers are in the revision note, and the AI may think they're still valid. Try: Say: "Use the rev. B specs; rev. A numbers are outdated."
• Judging by common sense. Most blenders crush ice, so the AI may call answer 5 accurate even though the spec sheet never says so. Try: Add: "If the spec sheet doesn't say it, label it Not in spec sheet."
• Getting a rewrite. The AI fixes the FAQ for you, and you lose track of what was wrong. Try: Say: "Label only, no rewrite."
• Not saying how many. The AI decides the length itself and usually returns more lines, one per answer than you can use. Try: Be exact: "exactly 5 numbered lines, one per answer."
• Getting markdown back. AI defaults to bold labels and headings, which show up as stray symbols in your approval notes. Try: Ask for "plain text, no bold, no headings."

What a strong prompt includes

• Gives exact labels
• Uses the current revision
• Names the source
• Says exactly how many
• Asks for plain text

Practice this now ↑

Related exercises

• Alert sales to a competitor's price change, facts only (Research AI · Product managers)
• Announce a referral bonus with the steps to claim it (Marketing AI · HR and recruiting)
• Answer an angry 'I unsubscribed' email honestly (Marketing AI · Marketing and content)
• Answer your CMO's campaign question in one Slack message (Marketing AI · Marketing and content)
• Approve or return expense claims against the policy (Productivity AI · Finance and accounting)