When Your AI Agent Should Ask You Instead of Guessing

An AI agent should ask a question when missing information could materially change the action or its consequences. It should usually proceed when a reversible choice has a sensible default and the user’s intent is already clear. The hard part is separating uncertainty about facts, preferences and authority; asking the same vague question for all three creates friction without improving the result.

A confidence percentage does not solve that design problem by itself. Build an explicit decision rubric around what is missing, what can be checked independently and what happens if the assumption is wrong. The result should be fewer unnecessary interruptions and clearer questions when a person’s answer matters.

The following is a proposed product policy. It is not a measured optimum. Apply it to the concrete action the agent is about to take, rather than to the conversation as a whole.

Just Ask for Calibration found that eliciting verbalized confidence could improve calibration on the question-answering settings it studied. That is a reason to evaluate the signal, not to dismiss it. It is also not a guarantee that a particular score predicts correctness on your business workflow.

An August 2026 paper on certified deferral in small language models examines limits of calibration and conditions for deferring answers. Its setting and statistical assumptions matter. Do not transfer a threshold from a research task into customer actions without testing whether the data and consequences match.

For a local test, retain the model’s proposed answer, confidence and outcome label on representative cases. Compare cases it would execute with cases it would defer. If the confident failures are the expensive ones, an attractive average accuracy does not rescue the policy. Report the types of failure and the resulting review workload.

A factual gap has an evidential answer. If the task needs a publication date, search the source. If two documents disagree, show the conflict. Asking the user to supply a fact already available in the authorized workspace wastes attention and can produce a less reliable answer.

A preference gap concerns the desired result. A founder may prefer a short investor memo over a detailed technical appendix, and the choice changes the work. Ask early when the difference is material; otherwise produce a sensible draft that can be revised.

An authority gap concerns whether the action may happen at all. The agent may understand the recipient and the message perfectly while still lacking permission to send it. Confidence about the content does not create authorization. Prepare the concrete result so the user can decide without imagining what will be sent.

Partition the task by dependency. An unanswered audience question may block the final introduction while leaving source collection and fact-checking available. An ambiguous destination may block sending while leaving the draft reviewable. Continue those independent steps so the user’s answer unlocks a prepared result.

Do not treat silence as approval. A timeout can select a default only when the workflow explicitly permits that default for a reversible choice. It cannot grant missing authority for an external action. Store the pending question and its dependent action together so a later response is applied to the right decision.

Keep a record of recurring questions. If the agent repeatedly asks about the same documented preference, the problem may be retrieval or instruction design. If different users regularly choose different answers, the product may need a persistent setting rather than a longer prompt.

This builds on approval-gate design and where agents read their instructions. The question policy should use those sources of authority, not compete with them.

Review a representative set of completed tasks and classify each interruption: necessary, avoidable or unresolved. Look at the answer the user supplied and whether it changed the outcome. Separately record cases where the agent proceeded under an assumption that should have been clarified.

A low interruption count can hide reckless assumptions. A high completion count can hide work that nobody wanted. Choose acceptance examples that expose both errors: a harmless formatting ambiguity the agent should handle, and a consequential object ambiguity it must resolve.

When changing models or prompts, rerun those cases. Our model-switch testing guide explains the broader discipline of using representative work. For question behavior, the labels should include whether the agent asked the right question at the right point, not just whether its final prose was correct.