Can AI Make Decisions for You?

October 15, 2026 · By Michel Valenzuela, Founder of Daryl

AI can compare options, recommend a choice, and, in some applications, take actions within a defined workflow.

Those are separate capabilities.

For a personal decision, the useful starting point is assistance you can examine: what the options are, which constraints matter, and what remains uncertain.

You decide how much authority to give the software. A recommendation should not silently become permission to act.

What does “making a decision” mean?

The phrase can describe several different things.

An assistant might organize information, suggest an option, or carry out a choice through a connected tool.

An application can provide one of these capabilities without providing the others.

A chatbot recommending a supplier has not necessarily contacted that supplier. An assistant with messaging access may be able to do more, depending on its tools and permissions.

What can AI contribute to a difficult choice?

It can help make the decision easier to inspect.

You might know the facts but struggle to organize them. You might have two attractive options with different costs. You might be unsure which missing detail deserves attention first.

An assistant can help you express those issues.

For example:

“I am deciding whether to accept a project. The deadline is fixed, my team is already busy, and the client may accept a smaller scope. Help me identify what to clarify before answering.”

A useful response should preserve the fixed deadline, account for capacity, and treat the smaller scope as unconfirmed.

That gives you a clearer basis for your next conversation.

Why does the person’s judgment still matter?

Because the assistant has a partial account of the situation.

It knows what the application makes available to it. That may exclude relationships, commitments, practical experience, or consequences you have not explained.

You might recognize that a technically workable plan would damage trust. You might know that a colleague’s apparent availability does not reflect their actual workload.

Even a detailed conversation cannot guarantee that every relevant factor has been captured.

Your judgment includes deciding whether the assistant’s framing is appropriate.

Can you ask AI to choose between two options?

Yes, but give it criteria.

“Which is better?” leaves the meaning of “better” open.

A clearer request is:

“Compare these options for a project that must launch by Friday. Staying within our current team capacity matters more than adding features.”

Then ask it to explain the tradeoff.

If it recommends the simpler option, you should be able to see why: perhaps it fits the deadline and available capacity while sacrificing some functionality.

You can decide whether that tradeoff is acceptable.

What if neither option works?

Ask the assistant to consider that possibility explicitly.

A request to choose between A and B can encourage a comparison even when both violate a firm requirement.

Try:

“Check whether either option meets the requirements before recommending one. If neither does, explain what would need to change.”

That can lead to a more useful answer.

You may need a revised deadline, a smaller scope, or another approach. Identifying that problem early is better than treating the least unsuitable option as viable.

When does a recommendation become an action?

When the application uses a tool or integration to change something.

Examples include sending a message, submitting a form, purchasing an item, or modifying a file.

A proposed action and a completed action should be distinguishable in the interface.

Before allowing execution, you need to understand the target, the content, and the consequences.

For example, reviewing a draft response to a client is different from authorizing the application to send it.

The permission should match the action you intend.

Does running locally change who has authority?

No.

Local processing describes where the model runs. Authority depends on what tools the application provides and what access you grant.

A local assistant can be limited to conversation. Another local application can have extensive automation capabilities.

Similarly, a cloud assistant may provide advice without having any way to act.

Evaluate processing location and action permissions separately.

Privacy does not establish decision quality, and model intelligence does not establish permission.

What is a practical way to use AI for a decision?

Work through a short sequence.

• State the decision.
• List the realistic options.
• Identify firm constraints.
• Separate known facts from assumptions.
• Ask which missing information could change the answer.
• Review the recommendation and its tradeoff.
• Choose a next step you understand.

Sometimes the outcome will be a choice.

Sometimes it will be a question to ask or a small test to run.

Both can represent useful progress.

How does this relate to Daryl?

Decision assistance is central to what we are building with Daryl.

Daryl is being developed as a personal AI for Mac, with local core conversations, voice and typing, and help examining possible next steps.

The intended experience should let people clarify a situation, compare options, and question a recommendation.

Daryl is still in development. Browser-related capabilities through Omega are also in development and should be described according to what is actually available.

A conversation about an action should not be presented as proof that the action has happened.

Clear authority and clear status are part of making assistance useful.

Frequently asked questions

Can AI tell me what to do?

It can recommend a course of action. Assess whether the recommendation fits your facts, priorities, and constraints.

Should I follow the option with the highest score?

First establish what the score means. A ranking does not necessarily represent the probability of a successful outcome.

Can an assistant act without asking?

That depends on the product and the authority previously granted. Understand the settings before enabling action capabilities.

Can Daryl make the final choice for me?

Daryl is being built for decision assistance. Its purpose is to support an informed choice through an experience you can examine and correct.

Keep the next step clear

A useful recommendation should leave you able to explain why an option fits, what it costs, and what remains uncertain.

Then decide whether to proceed.

That is the relationship we want to build with Daryl: assistance that helps you move forward with a clearer understanding of your choice.