You’ve heard enough about AI. This isn’t another article about what it can do, how fast it’s moving, or what it might replace. It’s about one specific thing most people don’t do with it — and why changing your approach is valuable.
Most people use AI to get answers. That’s fine for facts. It’s a serious limitation for decisions, strategies, and complex judgements. When you ask for an answer, you get one shaped entirely by the way you framed the question. The AI inherits your assumptions without questioning any of them. You stay inside the frame you arrived with.
There’s a different way to use AI. Not as an answer machine, but as a thinking partner — one that helps you question the frame itself, surface options you hadn’t considered, and stress-test conclusions before they meet the real world. This is harder than asking a question and accepting a response. It is also considerably more valuable. And it keeps you — not the AI — in charge of the thinking.
The Core Claim
AI won’t think for you — and shouldn’t. But the right kind of back-and-forth can make your thinking more rigorous, your options wider, and your conclusions harder to knock over.
The Problem with Getting Answers
When you ask AI “what should I do about X?” you get a response shaped by the way you framed X. The question contains assumptions — about what the real problem is, what constraints apply, what a good outcome looks like. The AI inherits all of them without questioning any of them.
This is fine for factual retrieval. It’s a serious limitation for decisions, strategies, and complex judgements — exactly the situations where you most need clear thinking.
The psychologist Gary Klein, who studied how experts make decisions under pressure, found that the most common failure in difficult situations wasn’t choosing the wrong option from a known set. It was not seeing the options that were available. You optimise within the frame you’re in. You don’t question the frame.
The most common failure isn’t choosing the wrong option — it’s not seeing the options that were available.
AI, used as an answer engine, reinforces the frame you bring. AI, used as a thinking partner, can break it.
What Thinking with AI Actually Looks Like
The shift is behavioural as much as conceptual. Instead of submitting a question and accepting an answer, you treat the first response as the start of a conversation rather than the end of one. You push back. You ask it to argue the opposite. You request alternatives. You probe the assumptions buried in its own suggestions.
The exchange becomes iterative — a genuine to-and-fro in which your understanding develops across multiple turns, not in one. This is closer to thinking out loud with a knowledgeable colleague than querying a database.
The Reframing Step
One of the most powerful things you can ask an AI to do is reframe. Not “is my analysis right?” but “what would have to be true for the opposite conclusion to be correct?” Not “help me build the case for this decision” but “what are three structurally different ways to think about this situation?”
Reframing is cognitively expensive when done alone. You’re fighting your own anchoring, your own loss aversion, your own vested interest in the conclusion you’ve already formed. AI has none of those — it can generate genuinely alternative framings without effort and without ego. That’s a real and underused capability.
The Sanity Check Step
Once you have options on the table — genuinely different ones, not variations on the same theme — the second critical use is rigorous stress-testing. Not “does this sound reasonable?” but “what are the conditions under which this fails?” Not “what do you think?” but “what would a well-informed sceptic say about the three weakest points in this argument?”
The combination — option generation followed by adversarial testing — is the core of structured thinking. AI makes both faster without making either shallow, provided you drive the process rather than receive it.
Two Ways to Use AI
Answer Mode
- Submit a question, receive a response
- Frame is set by the questioner
- One turn, one output
- Replaces research
- Confidence without examination
Thinking Mode
- Iterative, multi-turn dialogue
- Frame is questioned and expanded
- Options surface across turns
- Augments reasoning
- Clarity earned through challenge
A Practical Structure
This isn’t mystical. It’s a discipline — one that can be learned and applied consistently. The following sequence works across most complex problems.
1. State the problem — then ask AI to restate it differently
Give the situation as you understand it. Then ask: “How else could this problem be framed? What assumptions am I making?” You’re not outsourcing analysis — you’re stress-testing your starting point before you build on it.
2. Generate structurally different options
Ask for three to five responses that are genuinely different in kind, not just in degree. Push for options that would be uncomfortable or surprising. The aim is to expand the possibility space before collapsing it.
3. Apply a consistent evaluation framework
Run each option through the same set of questions: What does this cost? What does it risk? Who does it affect and how? What does it assume? A structured framework prevents you from unconsciously favouring the option you already prefer.
4. Seek the adversarial view
Ask AI to argue against the option you’re leaning towards — not gently, but as a well-informed critic who wants to expose its weakest points. If the option survives this, your confidence is earned. If it doesn’t, you’ve learned something before it costs you.
5. Make the decision yourself
AI has no stake in the outcome. It doesn’t know your context, your risk tolerance, your relationships, or your values with the depth you do. The process is an input to your judgement — it doesn’t replace it. Own the conclusion.
Going Deeper
This sequence is the intuition behind the RESOLVE framework. RESOLVE formalises these five steps into a structured seven-stage method — with defined analytical tools at each stage, a mandatory framing scan before any analysis begins, and a Delta close that forces you to state explicitly what changed in your understanding as a result of the process. If these five steps feel useful, RESOLVE is where they go deeper.
Why This is Harder Than It Sounds
The obstacle isn’t technical. It’s psychological. Getting a confident answer from AI feels productive. Interrogating that answer, asking for its opposite, demanding alternatives — this feels like more work, and it is. It requires you to stay in the driving seat of a process that constantly offers to do the driving for you.
It also requires a degree of intellectual honesty that’s uncomfortable. Using AI to genuinely challenge your thinking means being open to the possibility that your initial framing was wrong, your preferred option is weaker than you thought, or the problem you’re solving isn’t the actual problem. Those are valuable discoveries. They don’t feel like it in the moment.
The process requires intellectual honesty: openness to the possibility that your initial framing was simply wrong.
What AI Cannot Do
AI doesn’t know what actually matters to you. It can model your stated preferences but not your lived ones — the trade-offs you’ve made before and regretted, the relationships that constrain what’s actually possible, the values you hold but rarely articulate. It has no skin in the game.
Used well, this is a feature. A middle-layer with no stake in the outcome and no social awkwardness about saying uncomfortable things is genuinely useful. The risk is treating its absence of values as the presence of objectivity. It’s not objective — it’s differently partial.
The critical skill is knowing which parts of a problem benefit from that detached perspective and which parts require your own judgement to be brought forward, not delegated away.
Better Thinking as a Practice
The argument here isn’t that AI makes you smarter. It’s that the right kind of interaction with AI can make your thinking more rigorous — by surfacing options you’d missed, challenging the frames you brought, and subjecting your conclusions to adversarial scrutiny before they meet the real world.
That’s a discipline. Like any discipline, it requires structure, repetition, and the willingness to do the harder thing when the easier thing is available.
The easier thing is asking for an answer. The harder thing is using the exchange to think.
Most people do the easier thing. That gap is where better thinking lives.