How to Think Through a Difficult Decision With AI

August 26, 2026 · By Michel Valenzuela, Founder of Daryl

A new opportunity arrives at the wrong time.

The project is interesting. The money would help. You would enjoy working with the client. But your calendar is already full, and accepting means postponing something you have spent months trying to build.

You can make a convincing argument in either direction.

This is where using AI for decision-making can be useful. It gives you a way to lay out the situation, examine competing priorities, and discover what you still need to know.

The quality of that conversation depends on how you approach it.

Start with the situation as it actually is

You do not need a perfect prompt. You need an honest account.

Imagine you are an independent professional considering a six-week client project. You could begin like this:

“I’m considering a project that would bring in useful revenue, but I’m already committed to other work. Accepting it would probably delay my own product launch. I like the opportunity, and I’m worried about overcommitting. Help me work through the decision. Ask me about anything important I haven’t explained.”

That gives the conversation somewhere useful to begin: an opportunity, a constraint, a competing goal, and uncertainty.

Include the details that could change your decision. How much time is genuinely available? How flexible is the deadline? What would you have to postpone? What happens if you decline?

When sharing sensitive information, choose an application whose processing and storage practices you understand. You can often explore the decision without identifying the client or disclosing confidential material.

Decide what matters before comparing the options

“Should I accept?” sounds like one question. Underneath it are several.

Do you need the income now? Is this a strategically valuable relationship? Is your own launch deadline firm? How much additional pressure can you reasonably take on?

An AI can help you articulate those criteria. You might ask:

“Help me identify the priorities that should guide this decision. Separate my firm commitments from preferences I could negotiate.”

That distinction matters.

A preference might be keeping your Fridays free. A firm commitment might be delivering work you have already promised. Both deserve consideration, but they do not carry the same weight.

Once your priorities are explicit, you have a basis for comparing the options. Otherwise, the most persuasive argument in the moment can dominate the conversation.

Give yourself more than two choices

Many difficult decisions begin as a yes-or-no question because those are the options immediately visible.

In our example, there may be several possibilities:

These are possibilities to investigate. The client might have no flexibility. A smaller scope might not solve their problem.

But you cannot evaluate an option you have never considered.

A useful question is:

“What other realistic options should I consider, given these constraints?”

Separate what you know from what you assume

Decisions become harder when estimates quietly turn into facts.

“The project will take fifteen hours a week” might be an estimate based on a brief conversation. “The client won’t accept a later start” might be something you have never asked. “Declining will damage the relationship” might be a fear.

Ask the AI to help you sort the information:

“From what I’ve told you, separate confirmed facts, estimates, and assumptions. Which uncertainties could materially change the decision?”

Then check its interpretation. An AI can misread your account or fill gaps with plausible details.

In this example, confirming the scope and deadline might be more valuable than another hour spent debating whether to accept.

You might discover that the sensible next step is a conversation with the client.

Ask it to challenge your preferred answer

By this point, you may already be leaning toward an option.

That is useful information. It is also a reason to examine your reasoning more closely.

If you want to accept, try:

“Make the strongest reasonable case against accepting. Use the facts I’ve provided and identify where you need more information.”

If you want to decline, reverse the question.

You can also ask:

“What would have to be true for another option to become the better choice?”

Perhaps accepting makes sense only if the scope is fixed and the start date moves. Perhaps declining makes sense only if you have enough financial room to protect the launch.

Those conditions turn a vague preference into something you can test.

Choose a next step you can actually take

A decision conversation should eventually produce movement.

That might mean accepting or declining. It might mean gathering one missing piece of information before committing.

For our example, the next step could be:

“Ask the client whether they would consider a smaller initial engagement beginning after the current delivery deadline.”

You have a concrete action and a reason for taking it. You can then revisit the decision with better information.

For larger choices, write down the reasoning: what you knew, what mattered, what you chose, and what would justify reconsidering. That gives your future self something more useful than a memory of how confident you felt.

Keep responsibility for the judgment

An AI can generate a fluent explanation for a weak recommendation. It can overlook a constraint, accept an inaccurate assumption, or sound more certain than the information allows.

Treat its output as material to examine.

Check the facts. Correct misunderstandings. Bring in people with relevant experience when you need it. Pay attention to consequences that are difficult to reverse.

You are the person who knows what it means to live with the choice.

Why I’m building Daryl around decisions

This is one of the experiences I want Daryl to support.

Daryl is a personal AI for Mac, currently in development, designed for local conversation and help thinking through decisions. You can bring a situation, explore priorities and trade-offs, and work toward a practical next step.

The example in this article illustrates an approach to thinking with AI. It is not a transcript of a released Daryl feature.

My aim is to make that process approachable through voice and typing, with core conversations processed on your own computer and your judgment at the centre.

Sometimes greater clarity leads to a firm answer. Sometimes it reveals the question you need to ask before you can responsibly decide.

Both are progress.