Label six messages with intents and flag ambiguous ones

About this exercise

Label each message with one intent from: ORDER_STATUS, REFUND, ACCOUNT_ACCESS, BILLING, OTHER. If a message has two intents, use the first one mentioned and set confidence to low; otherwise high. Return one CSV code block with the header "id,intent,confidence" and six rows. No text outside the block. At most 8 lines.

Common mistakes when you ask AI for this

• Picking the wrong intent for M2. M2 has two intents and the AI picks the second. Try: Say: "Two intents: use the first one mentioned, confidence low."
• Marking everything high. The AI never flags the ambiguous message. Try: Say: "Confidence low only for messages with two intents."
• Extra text. The AI adds an explanation or a table. Try: Say: "One CSV code block, header id,intent,confidence."
• Skipping the code block shape. The AI splits output into several code blocks or none. Try: Say "exactly one code block and nothing else".
• Letting the code sprawl. The AI adds extra lines and comments. Try: Say "at most N lines of code".

What a strong prompt includes

• Asks for one code block
• Caps the code length
• Gives the header
• States the two-intent rule

Practice this now ↑

Related exercises

• Accept the buyer's counteroffer on three stated conditions (Sales AI · Sales and customer success)
• Accrue goods received but not invoiced at PO price (Finance AI · Accountants)
• Add dark mode by overriding your existing CSS variables (Coding AI · Software developers)
• Add input checks to your number guessing game (Coding AI · High school students)
• Add labels to a signup form without rewriting the page (Coding AI · High school students)