Think out loud with AI
Some of the best work starts before you know what you're asking. Brainstorming with AI isn't about getting an answer. It's about finding the real question.
How to think out loud
1. Say you're exploring
Tell it up front you don't want a final answer yet. Otherwise it rushes to one.
2. Ask for range, not one answer
One idea is a suggestion. Five different ones is a map.
3. Let it interview you
Questions pull out the things you know but didn't think to say.
4. Name the spark when it shows up
When something clicks, say so and push on it. That's where the real idea is hiding.
5. Close the loop
End every brainstorm with a written summary, or it evaporates.
Why this works
It stops the rush to an answer
Ask a question and you get an answer, even when the question was the wrong one. Saying you're exploring buys room to find the right one.
Reacting is easier than describing
You can't always say what you want. You can almost always say what's wrong with what's in front of you. Brainstorming puts things in front of you.
The summary turns talk into work
A good conversation that isn't written down is gone by tomorrow. The closing summary is what you hand over to the doing mode.
A worked example, start to finish
An owner of a small IT shop keeps wondering whether to start offering something new, but can't say what.
We're brainstorming, no final answers yet. Customers keep asking us for things we don't do. I think there's a new service in there but I can't see it.
Let's find it. One question first: what were the last three things customers asked for that you said no to?
Help setting up AI tools for their staff. Someone to check their website isn't broken. And a training session for a new employee.
Two of those are about people, not machines: getting staff comfortable with tools. A different angle: what if the service isn't AI setup, it's onboarding? Every new hire gets their accounts, their tools, and a short session on how to use them, done in a day.
That's it. That's the thing. We already do half of it when we set up their laptop.
Then here's a summary. Decided: explore a new-hire onboarding package. Rejected for now: website monitoring, because it's a different skill. Next step: list what you already do at laptop setup, and what you'd add.
This is an illustrative example. Names and details are made up.
Common mistakes
Asking for the answer in the first message
Fix: Say you're exploring. Ask it to help find the question first.
Taking the first idea
Fix: Ask for range, then react. The first idea is rarely the best one.
Brainstorming forever
Fix: Set an end: "Ten more minutes, then summarize and give me one next step."
Losing the conversation
Fix: Save the summary somewhere you'll see it: a note, a project, a task.
Check yourself
Pick the better request. You'll see why right away.
1. You're not sure how to grow your business.
2. Your brainstorm turned up three good ideas.
Exercise
Pick a decision you've been putting off. Write the first message that sets up a brainstorm, not a request for an answer. Write yours first, then compare.
Show a strong answer
Notice it says the mode, names the tension, asks to be interviewed, and sets an end point.
What Jason actually did
This site started that way. My business partner asked how to catch up on how I work with AI, and I didn't have an answer. I didn't ask an AI to write a course. I talked it through, first with one AI and then another. I rejected a lot: a design that looked like every other AI site, a headline that was too narrow. The spark came when one of them said people don't want a genie, they want the moment something clicks. That line became the site.
The Spark Lab
Write the messy version of something on your mind. Pick where you want to go. You'll get a conversation starter to paste into your AI. Nothing you type leaves this page.
Try it today
Private to you. Saved only in this browser, on this device. Nothing here is sent anywhere, and nobody else can see it. Clearing your browser data erases it.
Your saved prompts
Nothing saved yet. Use "Save this prompt" on any example above.
Go deeper
The sources worth your time if you want the reasoning behind this lesson.
- Let Claude think (Anthropic): why asking a model to think a problem through gives better results.
- One Useful Thing (Ethan Mollick): a professor's ongoing writing on working with AI as a thinking partner.
- Prompting strategies (Google Gemini): practical ways to steer a conversation.