Explain to a manager why the cheaper model costs more overall

About this exercise

Dana Whitfield, Spokewise's ops manager, asked why you want to pay 3x per chat for Quill-L. Reply to her by email, plain text, under 120 words, with no jargon: a line "Subject: [under 8 words]", then "Hi Dana,", a first sentence that gives the answer, one math line each for Quill-S and Quill-L (per 1,000 chats, wrong answers included), a line "Saving: ..." for a month, and a last line "Ask: ..." asking for her approval.

Common mistakes when you ask AI for this

• Defending the price per chat. The AI argues Quill-L is 'better quality' without counting what wrong answers cost. Try: Say: "Compare total cost per 1,000 chats, including $2.50 per wrong answer."
• Jargon for a non-technical reader. The AI mentions accuracy benchmarks, tokens and evals. Try: Add: "Dana is not technical: plain words, no jargon."
• Missing the monthly saving. The AI stops at the per-1,000 figures and never scales to a month. Try: Add: "Saving: the monthly difference at 20,000 chats."
• Long email with no ask. The AI writes a polite, long note with no clear request. Try: Say: "Under 120 words; last line Ask: approve the switch."
• Formatted email. The AI uses bold headings and bullet symbols. Try: Say: "Plain text, no bold."

What a strong prompt includes

• Names the reader
• Counts wrong answers
• Asks for the math
• Asks for Saving and Ask lines
• Asks for plain text
• Says no jargon

Practice this now ↑

Related exercises

• Announce a delay to three audiences with different limits (Leadership AI · Communications and PR)
• Announce one feature separately to admins and to end users (Customer Support & Success AI · Customer success managers)
• Answer "we'll build it ourselves" with their own numbers (Sales AI · Sales and customer success)
• Answer "what does a CSM do for us?" in plain spoken words (Customer Support & Success AI · Customer success managers)
• Answer a director's accrual question as a short Q&A (Finance AI · Accountants)