When the model should ask questions first
Learn when clarification is safer than confident guessing.
Clarifying is not stalling
If any of these are missing, asking first is usually better than generating immediately:
- The audience is unclear
- Success criteria are unclear
- Risk boundaries are unclear
Especially important for
- High-risk tasks: contracts, pricing, medical or legal wording
- High-rework tasks: long edits, code changes, bulk transformations
What good clarification looks like
- Ask only 2 to 4 questions
- Make every question output-relevant
- Ask enough to proceed, not enough to create an interview
Why it matters
Once you know when to clarify, the next step is learning when the model should stop chatting and start using tools.