“Add AI by Friday.” What do you ask first?
Resolve the risk boundary before choosing the model or writing the prompt.
The question
At fictional Harbour Design in Sydney, your manager wants an AI feature that suggests customer-facing product copy by Friday. You can ship an internal draft tool in two days; automatic publishing needs a new approval flow.
- Request
- “Make it save time. Use the best model. We need a demo Friday.”
- Risk
- Suggested copy can include untested performance claims. The product lead is responsible for approved claims.
- Options
- Internal drafts with human approval; or automatic publishing to the live catalogue. No error threshold has been agreed.
“What would you clarify, and what would you propose for Friday?”
Not saved by this site. Your browser may restore text when you navigate back.
A strong answer, step by step
Open each step in turn. This is one strong answer, not the only one.
Step 1: Clarify the decision the demo must enable
What would you do here? Open the move you’d make.
Pick the strongest model and start building.
Tempting, but it misses something. Model choice does not resolve what the feature is allowed to publish.
Clarify the user, output and approval boundary.
The strong move. The answer changes the scope and the acceptable risk.
See how a strong candidate says it
“Ask who uses the output, whether it is a draft or a published claim, and what Friday’s demo must prove.”
Why it works: The answer changes the scope and the acceptable risk.
Step 2: Set a bounded proposal
“Propose an internal draft tool with approval before publishing. Test it on a fixed set of real use cases with approved evidence.”
Why it works: A useful demo does not need uncontrolled access to customers.
Step 3: Name the quality and approval owners
“Agree which claim errors are unacceptable and who approves both wording and release. Record the open risks.”
Why it works: The engineering team should not silently own a product-claims decision.
Step 4: Return with a clear choice
“Offer the Friday draft demo and a separate plan for automatic publishing after controls pass.”
Why it works: This gives the manager a workable decision with explicit limits.
The answer, in short
- Decision
- Offer an internal draft demo for Friday with human approval; defer automatic publishing until its controls are agreed.
- Because
- The request has no stated publishing boundary, claim standard or approval owner.
- What would change it
- If the output is already constrained to approved claims with tested controls, a broader release could be considered.
Common traps
- Starting with model selection.
- Treating a demo date as release approval.
- Adding a warning while leaving automatic publishing enabled.
Check your answer
- Did I clarify the user and output?
- Did I distinguish drafts from publishing?
- Did I name the quality owner?
- Did I offer a workable Friday scope?
Last reviewed Nov 9, 2026