Unit 7AI councils and choosing the right model

The big idea

Not every question deserves the same amount of thinking. Small questions need a quick answer. Big decisions need several points of view. Part of using AI well is knowing which is which.

Why it matters

Knowing which tool fits which decision is exactly the kind of judgment employers pay for. It also keeps you from burning through your free usage on things that didn't need it.

AI councils

For a big decision, I use an AI council:

1. Several AIs from different companies each get the same question and make a plan on their own.
2. They review each other's plans anonymously, so no model favours its own answer.
3. They debate and recommend a decision.
4. A human makes the final call. The council gives you options and arguments. It doesn't decide for you.

Here's how to tell if a question needs a council. Asking a parent for five dollars is a quick yes or no. Asking to sleep over at a friend's on a school night means weighing what time you'll get up, whether they know the family, and what's happening tomorrow. Harder questions need more deliberation.

I use councils for things like reviewing a three-year plan or a big investment. Not for writing an email.

Match the model to the task

Bigger AI models think harder and cost more. Smaller ones are faster and cheaper. A student once told me they use a smaller model because the biggest one "overthinks" simple jobs. They were right.

• Use a strong model to plan and to make important judgment calls.
• Use a smaller, cheaper model to carry out routine steps.
• Don't put entry-level gas in a Ferrari, and you don't need a Ferrari to drive to the corner store.

Staying within free limits

1. Break problems into smaller pieces. Smaller prompts use less.
2. Write tighter prompts. A clear one-shot prompt (Unit 5) wastes less than ten rounds of back-and-forth.
3. Have another AI rate your prompt before you run it. Fix it first, then spend the usage.
4. Use the Socratic method (Unit 2) to think it through yourself when you hit a limit.

Do this

1. List five decisions or tasks from your project or your week.
2. Sort each one: quick answer from one AI, several AIs (a small council), or a human decision.
3. Run a mini council on the biggest one. Ask the same question in two or three different AIs. Paste each answer into the others and ask: "What's wrong with this plan? What did it miss?"
4. Make your decision and write down why.

Questions to ask yourself

• Am I using the most powerful tool for something simple?
• Which decision in my life right now deserves more than one opinion?
• After the council, what did I decide that the AIs didn't?

In my words

"You're not going to put entry-level gas in a Ferrari."

Worksheet

Use Worksheet 7: Model and council chooser in the PDF kit.