You ask for a client update and receive a complete-looking email. It includes a delivery date you never supplied. The assistant filled the gap instead of asking the question that mattered.
An AI assistant should ask for clarification when missing information could change the substance, recipient or permitted next step. For low-consequence presentation choices, a visible assumption may be sufficient. Define that distinction in the task brief rather than asking it to guess what matters.
Photo by Vitaly Gariev on Unsplash.
Separate missing facts from harmless preferences
A missing font preference rarely deserves the same interruption as an unknown price. Start by listing the information the task needs, then ask what would happen if each item were guessed incorrectly.
Use three routes. Proceed where the detail does not affect the result. State an assumption where it affects presentation but remains easy to change. Ask a person where it changes a factual claim, commitment or meaningful decision.
These are editorial working rules, not universal thresholds. The business owner must choose boundaries appropriate to the task.
Build a clarification ladder
Try this structure with a narrow internal drafting task:
Essential: information without which the task would be misleading or unusable.
Conditional: information required only if a particular option is chosen.
Optional: preferences that improve the output without changing its meaning.
Unavailable: facts the requester cannot currently supply and the output must leave unresolved.
Tell the assistant to identify the smallest set of questions needed to proceed. Asking ten questions at once can create more work than the draft saves. A good clarification request explains the decision the missing answer affects.
An illustrative deadline request
Suppose the instruction is: prepare an update about a delayed onboarding. The known facts are the current stage and a missing client document. A revised completion date has not been agreed.
The assistant can draft the confirmed position and leave the date unresolved. It should ask whether an authorised new date exists before including one. It does not need to delay the entire task to ask whether the greeting should say hello or good morning.
If the requester cannot provide the date, the right result is a draft explaining the dependency and next review point. Completeness does not require inventing certainty.
Give assumptions an obvious place
Ask for assumptions and open questions outside the client-facing draft. Otherwise an internal caveat may accidentally become part of a message or disappear during editing.
When the answer arrives, check whether it changes earlier wording. A newly supplied date may also affect dependencies, the next action or the person who needs to review the draft.
Avoid interrogation as the default
More questions do not automatically mean better assistance. Inspect whether the questions resolve a genuine decision or repeat information already supplied.
Keep examples of useful questions with the task instructions. Improve the brief when the same missing fact causes repeated interruptions. The aim is a conversation that moves work forward while keeping important unknowns visible.
For related practical guidance, see defining usable AI output and approving AI-assisted client messages.
Hili Consulting can help specify clarification behaviour for incomplete business requests. Arrange a free 30-minute consultation and bring one example of the work you want to improve.
