Most people use ChatGPT to write something. Far fewer use it to decide something. That’s a missed opportunity. Every week we all make dozens of small and medium decisions: which laptop to buy, which contractor to hire, which plan to pick, which of three job offers to take. Each one involves gathering options, comparing trade-offs and committing. AI is very good at the first two. You remain responsible for the third.
After more than 5,000 hours of hands-on work with generative AI since 2023, this is the method I keep coming back to. It works with ChatGPT, Claude or any capable assistant. It’s simple enough to use on a Tuesday afternoon and structured enough to trust.
The core idea: AI widens the options, you narrow them
Good decisions fail in two predictable ways. We consider too few options, or we compare them on the wrong things. AI helps with both, as long as you treat it as a research partner and not an oracle.
The method has five steps:
- Frame the decision
- Generate alternatives
- Define what matters
- Compare and stress-test
- Decide and record why
Step 1: Frame the decision
Start by telling the AI what you’re actually deciding, in plain language, with the context that a smart friend would need. Include your constraints: budget, timing, location, non-negotiables.
A weak prompt: “What’s the best laptop?”
A strong prompt: “I need a laptop for travel and light video editing. Budget is $1,500. I use a Mac today and don’t want to relearn an operating system. I need it within two weeks. Help me frame this decision before recommending anything: what questions should I answer first?”
Asking the AI to frame the decision before answering it is the single most useful habit in this guide. It surfaces assumptions you didn’t know you were making.
Step 2: Generate alternatives
Next, ask for a wide set of options, wider than you’d come up with alone. Explicitly ask for options you might not have considered, including “do nothing” or “wait.”
- “Give me eight realistic options, including at least two I probably haven’t considered.”
- “Include one option that is cheaper than my budget and one that breaks it, and say why each might be worth it.”
- “What would someone who disagrees with my framing choose?”
This is where AI earns its keep. It removes the narrowing that happens when we only look at the first three search results.
Step 3: Define what matters
Before comparing anything, write down your criteria. Ask the AI to propose criteria based on your framing, then edit them yourself. Price, reliability, time to value, risk and how you’ll feel in six months are common ones.
Then rank or weight them. If you’re not sure how, ask: “Based on what I’ve told you, which of these criteria seem most important to me, and why?” Disagree with it where it’s wrong. That disagreement is information about what you really value.
When the decision is important enough, turn these criteria into a weighted decision matrix. I walk through that in How to Build a Simple Decision Matrix with ChatGPT.
Step 4: Compare and stress-test
Now ask for a side-by-side comparison against your criteria. Then push on it:
- “Which facts in this comparison are you least sure about? How would I verify them?”
- “What would make the top option the wrong choice?”
- “Argue for the second-ranked option as strongly as you can.”
Verify anything that matters: prices, specifications, policies, availability and dates. AI can be confidently out of date. The point of this step isn’t to get a perfect answer from the model. It’s to find out where the comparison is weakest so you can check those things yourself.
Step 5: Decide and record why
Make the decision yourself. Then ask the AI to summarise it in three sentences: what you chose, the two or three reasons that mattered most, and what would make you revisit it. Save that summary.
That short record does two things. It keeps you honest later, when hindsight makes every decision look obvious. And it turns one good decision into a reusable pattern for the next one.
A quick example
Choosing between three internet providers took me about fifteen minutes with this method. The AI framed the question (do I care more about upload speed or price?), surfaced a fourth provider I hadn’t heard of, built a comparison on speed, contract terms, real-world reliability and total two-year cost, and flagged that one provider’s promotional price expired after twelve months. I checked the promotion and the contract terms on the providers’ sites, picked the one that won on two-year cost, and wrote down why.
None of that required trusting the AI blindly. All of it went faster and covered more ground than doing it alone.
Where this goes wrong
AI-assisted decisions fail when you skip the framing, accept unverified facts or let the model make the call. I cover the most common traps in Six Mistakes to Avoid When Using AI for Decisions.
The Human + AI takeaway
This is the pattern behind everything we do at IT Strategists: AI does the wide, fast, tireless part of the work, and people do the judgment. Start with everyday decisions. Once the habit is built, the same method scales to business decisions about vendors, hires, tools and strategy.