When the Best AI Answer Is Another Question

September 6, 2026 · By Michel Valenzuela, Founder of Daryl

“Should I accept the offer?”

It sounds like a question that needs an answer.

But which offer? What would accepting change? What matters most to you? What have you already committed to?

Until those details are clear, a recommendation can arrive before the conversation has earned it.

This is something I keep returning to while building Daryl: a useful assistant needs to recognise when it has enough information to help, and when it should ask.

A clear question can hide an unclear situation

We often express a complicated decision in a short sentence.

Should I change jobs? Launch now? Hire someone? Take on the project?

The wording is simple because we already know the circumstances behind it. The assistant may know almost none of them.

You might ask whether to hire someone when the real issue is an unclear process. You might ask whether to delay a launch when the unresolved question is what belongs in the first release.

An immediate answer can reinforce the way you framed the problem before that framing has been examined.

A useful question creates room to look again.

Ask about the detail that could change the answer

Good clarification has a purpose.

Suppose you are considering a new project. The fee is attractive, but you are already busy.

An assistant could ask you to describe your entire working life. More useful would be a question about the constraint that determines whether the project is feasible:

“What existing commitment would you need to move if you accepted?”

That answer might reveal a firm deadline, an optional activity, or room you had overlooked.

Another question could be:

“Is the client’s start date fixed, or have you assumed it is?”

If the start date is negotiable, the decision changes.

The value comes from identifying the missing information that matters. Asking more questions is useful only when they improve the discussion.

Find out what “better” means

“Which option is better?” requires a standard for comparison.

Better for revenue this month? Better for learning? Better for protecting your time? Better for the direction you want your business to take?

Those priorities can point toward different choices.

Imagine comparing two projects. One pays more and requires familiar work. The other pays less but builds experience in a field you want to enter.

An assistant can describe the trade-off. It cannot establish how much that experience matters to you without understanding your priorities.

A question such as “What are you trying to make possible over the next year?” can give the comparison a meaningful basis.

Leave room for an option you have not considered

Sometimes the most useful question challenges the available choices.

“Should I accept or decline?” may leave out a smaller engagement, a later start, or a referral to someone better suited to the work.

“Which feature should we cut?” may leave out changing the release sequence.

“Should I leave this project?” may leave out renegotiating your role.

A helpful assistant can ask:

“Are these the only workable options, or can part of the arrangement change?”

That question does not guarantee a new solution. It helps you check whether the decision has been narrowed too early.

Uncertainty should remain visible

An AI can generate a convincing explanation while missing a crucial fact.

Fluent language can make that gap difficult to notice. The response sounds complete, so you may assume the underlying understanding is complete too.

I want a personal assistant to make important uncertainty visible.

It should be able to say that a recommendation depends on an assumption, that two options remain difficult to distinguish, or that a missing detail could reverse the comparison.

For example:

“If the deadline is fixed, the smaller scope looks more practical. If the deadline can move, the full project may still be workable.”

That gives you a conditional assessment you can examine. You can confirm the deadline instead of treating the answer as settled.

Knowing when to stop asking matters, too

A conversation can become exhausting if every answer produces another round of questions.

People often need a practical next step with the information currently available. Waiting for complete certainty can keep a decision suspended indefinitely.

A useful assistant should help you distinguish between information you need before acting and information that would merely be nice to have.

You might have enough to propose a revised scope, even if you do not yet know whether the client will accept it.

The next action can resolve the uncertainty.

What this means for Daryl

Daryl is a personal AI for Mac, currently in development, designed around local conversation and support for decisions.

Part of the experience I’m working toward is careful handling of incomplete information: asking relevant questions, making assumptions visible, and helping people recognise when a conclusion is premature.

The examples here describe that design direction. They are not transcripts from a released feature.

I want Daryl to help a conversation move toward clarity at an appropriate pace. Sometimes that means comparing the options. Sometimes it means identifying the one thing you need to find out first.

A question can be the useful outcome

You may begin a conversation wanting a recommendation and leave with a question for your client, colleague, or yourself.

That can be progress.

You have located the uncertainty instead of carrying it invisibly into the decision. You know what to ask and why the answer matters.

For me, that is an important standard for personal AI.

An assistant should help you reach a conclusion you can understand. When the information is incomplete, the next good question may be the most useful thing it can offer.