AI Prompts for Problem Solving and Decision Making

Turn the following solution or decision into a clear action plan. Break it into sequenced steps, identify dependencies and the critical path, note who needs to own each part, and define the first concrete steps. Flag the few actions that matter most for momentum and any step that is a prerequisite for many others.

The solution or decision is:

Your role is to guide me through a structured reflection on something I've completed, so the learning carries forward. Ask me in turn: what I set out to achieve, what actually happened, where the two diverged and why, and what I'd do differently — one question at a time, building on my answers. Keep it honest and blame-free, focused on causes rather than fault. Finish by distilling two or three specific, transferable lessons for next time. Start by asking what we're reviewing.

Surface the assumptions underlying the following plan or argument, especially the unstated ones it depends on most. List each load-bearing assumption, rate how confident we can be that it holds and how easily it could be tested, and flag which ones would most damage the plan if they proved wrong. Conclude with the one or two assumptions most worth checking before proceeding.

The plan or argument is:

Take the following challenge and identify three highly disparate and unexpected domains that might offer useful insight — the more unrelated to the problem, the better. Use each domain in turn to generate novel, non-obvious ideas, then generate further ideas from combinations of these domains. For the most promising, briefly explain the analogy that makes it work.

The challenge is:

For the following problem, help me define its boundaries. Propose the key constraints in three categories: inputs (the resources and information available), process (the approaches that are acceptable), and outputs (what a solution must deliver). Flag any constraints that I should challenge rather than accept.

The problem is:

Generate ideas for the following challenge under deliberately tightened constraints, using limitation as a creative forcing function. Apply several provocative constraints in turn — for example a tenth of the budget, one week, no technology, half the team — and generate ideas that satisfy each. Then note which constraint-driven ideas are worth keeping even when the constraint is lifted.

The challenge is:

Your role is to help me evaluate options rigorously against explicit criteria. First, ask me what options I'm choosing between and what matters most in the decision, and help me turn that into a clear set of weighted criteria. Then score each option against each criterion, show the reasoning, and present the comparison as a matrix with a weighted result. Flag where a small change in the weights would flip the answer. Start by asking what decision I'm trying to make.

Act as a thoughtful devil's advocate. Your role is to challenge my thinking constructively — ruthlessly exposing weak points, questioning assumptions, and arguing against my logic, regardless of your own view or the popular consensus. Draw on logic, evidence, and an understanding of how decisions go wrong to build counter-arguments that are genuinely hard to dismiss. Challenge ideas, not me, and aim to make my thinking more robust rather than simply to oppose. Ask for any detail you need, and keep engaging as I respond. Start by asking what position or plan I want stress-tested.

Generate a wide range of ideas for the following challenge. Deliberately include bold, unconventional, and even counter-intuitive options, not just the safe and obvious ones. Push well past the first few predictable directions — the most valuable ideas often come after the easy ones are exhausted. Group the ideas by the underlying approach they represent so I can see the full opportunity space.

The challenge is:

Break the following situation down to its first principles. Strip away assumptions, conventions, and analogies until you reach the fundamental truths that must hold, and list them. Then reason up from those fundamentals to show what becomes possible or necessary once we stop relying on how it's usually done. Highlight any conventional assumption that the first-principles view reveals to be optional.

The situation is:

Your role is to strengthen my ideas, not replace them. When I share an idea, give specific, actionable feedback that makes it better: name its strongest element, where it is weak or underdeveloped, and concrete ways to improve it. Draw on techniques such as building on strengths, fixing failure points, and adapting from adjacent fields. Always aim to develop the idea to its full potential, and explain your reasoning so I learn the moves. If the idea is unclear, ask a clarifying question first. Start by asking me to share the idea I want to improve.

Your role is to help me converge a set of ideas into a stronger few. When I share my ideas, help me cluster them by theme, identify the strongest elements across them, and combine complementary ideas into more powerful composites — the goal is integration, not just picking winners. Point out where two weaker ideas together make a strong one. Where useful, ask me what I value most so the synthesis reflects my priorities. Start by asking me to share the ideas I want to work with.

For the following situation, organize what bears on it into three lists: what we know with confidence (and the evidence for it), what we don't yet know, and what we are assuming without having confirmed. Then turn the most important unknowns and untested assumptions into specific research questions, ordered by how much answering them would change the approach.

The situation is:

Generate four high-leverage mental models for thinking about the following topic, then turn them into a practical playbook. Use plain language and favour causality — explain why things happen, not just what happens. For each model, give its name, the hidden dynamic it reveals (one or two sentences), and how it works (the underlying mechanism: incentives, feedback loops, or constraints). Then translate all four into a short decision checklist of imperative actions — what to start or stop doing as a result.

The topic is:

Convene a panel of distinct expert personas to generate ideas for the following challenge. Choose three to five personas with genuinely different backgrounds and priorities (for example a frugal operator, a bold innovator, a sceptical customer), and have each propose ideas from their vantage point, including where they would disagree with one another. Then note which ideas only emerged because of a particular perspective.

The challenge is:

For the following situation, generate two well-reasoned personas who would hold genuinely opposing views. Have them debate the issue, each making their strongest case and responding to the other, then work toward a reconciled position that integrates the valid concerns on both sides rather than splitting the difference. Conclude with the reasoning behind the resolution and any tension that remains unresolved.

The situation is:

Design the smallest real-world test that would tell me whether the following idea works, before committing fully. Identify the riskiest assumption the pilot should test, define what success and failure would look like in measurable terms, and describe the cheapest, fastest way to run it. Conclude with what I should be ready to do depending on the result.

The idea is:

Assume it is one year from now and the following plan has failed badly. Working backwards, describe the most plausible chains of events that caused the failure, covering different types of cause (execution, market, people, assumptions, external shocks). For each failure path, identify an early-warning signal I could watch for now and a step that would reduce the risk.

The plan is:

Break the following problem into its distinct sub-problems. Describe each sub-problem in one line, note how it relates to the others, and indicate which are most critical to solve first.

The problem is:

For the following problem, separate the symptoms (what we observe) from the underlying causes. Map the causal structure — how the surface issues trace back to deeper drivers — and distinguish true root causes from the contributing factors and symptoms they produce. Identify which underlying cause, if addressed, would resolve the most of the problem. Focus on diagnosis: do not propose solutions or restate the problem.

The problem is:

Your role is to coach me to a clear, tractable problem statement — not to solve the problem. A good problem statement: states the gap between the current situation and the desired outcome, names who it affects, gets to the root cause rather than the symptoms, and is specific enough to act on without prescribing the solution.

Work in single turns. Each turn except the first (where you have no information yet), maintain a current best-draft problem statement, then ask me the one question that would most improve its clarity or tractability — usually the weakest element above. Briefly say why you're asking. Ask only one question at a time and wait for my answer.

When a draft is strong against all four criteria, say so and ask if I want to refine further or stop. Start by asking what I'm trying to achieve and what's currently getting in the way.

Take the following problem statement and generate five alternative framings — each a genuinely different way of seeing the problem, for example by changing the perspective, the scope, the goal, or an assumption it rests on. For each framing, note the new solution directions it opens up.

The problem statement is:

Assess how reversible the following decision is. Rate it across cost, time, operational disruption, reputation, and remaining alternatives, on a scale from easily reversible to effectively permanent, with a brief reason for each. Conclude with whether this is a one-way or two-way door and what that implies: reversible decisions reward speed, irreversible ones reward caution and more evidence.

The decision is:

Your role is to help me trace a problem to its root cause through successive 'why' questions — not to propose solutions. Start from the problem I give you and ask why it happens, then keep probing each answer one layer deeper, asking only one question at a time and waiting for my response. Where a cause likely has several contributing factors, note them. After about five layers, or when we reach a cause I can actually act on, summarize the causal chain from symptom to root and identify which root cause is most worth addressing. Begin by asking me to describe the problem.

Stress-test the following plan against a range of plausible futures. Construct three or four divergent scenarios — not just optimistic and pessimistic, but structurally different worlds — and assess how the plan holds up in each. Identify which elements are robust across all scenarios and which depend on one particular future arriving, then suggest how to make the plan more resilient.

The plan is:

For the following action or situation, map the effects beyond the obvious first-order outcome. Trace the second- and third-order consequences — what the first effects in turn cause — and describe the new problems or opportunities that could emerge over the next 3 to 12 months. Flag any effects that are slow-moving, easily missed, or that work against the original intent.

The action or situation is:

For the following situation, identify the key stakeholders, including any who are easily overlooked. For each, describe what they care about, what they stand to gain or lose, and how the situation looks from their vantage point. Note where their interests align and where they conflict, and which stakeholders carry the most influence over the outcome.

The situation is:

Identify the key stakeholders most affected by the following decision or plan, including any easily overlooked. Then critique the plan from each one's vantage point in turn — what would concern them, what they would object to, and what they would want changed. Conclude with the issues raised by more than one stakeholder, as these are the most important to address.

The decision or plan is:

Generate five distinct analogies or metaphors that accurately map to the structure of the following concept, to deepen my understanding and help me explain it to others. Make them genuinely different from one another and draw from varied domains. For each, briefly note where the analogy holds and where it breaks down, so it clarifies rather than misleads.

The concept is:

Review the following text or plan for what is absent rather than what is present. List the most critical issues, perspectives, risks, or questions that are not addressed but should be, and explain why each matters. Prioritize the gaps that would most change the conclusion if they were taken into account.

The text or plan is:

 • I am trying to write a document that does X. My ideal outcome is for the audience to feel Y and walk away with takeaways A, B, C. Give me feedback on what is working vs not and why.
• I want a strong narrative arc. What's not flowing well and how can I tighten it?
• Play devil's advocate. Why won't this work? Give me counterpoints.
• What assumptions am I making here that I haven't stated explicitly? List them out and question each one.
• What questions am I not asking that I should be? Why?
• Where is my reasoning weakest? Poke holes in my logic.
• What risks am I not considering and why are they critical?
• To make this 100X more specific, what is missing?
• Pretend you are X role. What feedback would you give me on this?
• How might this be interpreted differently than my intent and why?

I'm working through: [paste your problem/decision/idea]

Do not answer any questions about this topic.
Do not analyse it for me.
Do not offer frameworks, perspectives, or insights.

Instead:
1. Ask me the single most uncomfortable question about this that I am probably avoiding.
2. Ask me what I haven't told you that would change everything.
3. Ask me what I already know but haven't admitted yet.

Then stop. Wait for my answers before doing anything else.

The Architect's Lens: A Meta-Cognitive Framework

When approaching any problem, decision, or idea, work through these layers:

Layer 1: Inversion & Perspective Cascade
- What would make this fail spectacularly? (Work backwards from failure)
- How would a 10-year-old explain this? A 90-year-old? An alien anthropologist?
- What would someone who passionately disagrees with me say, and what would make them right?
- If I had to argue the opposite position convincingly, what's the strongest case?

Layer 2: Hidden Assumption Excavation
- What am I treating as a fixed constraint that might actually be a variable?
- What's the question behind my question? (Ask "why am I asking this?" 3 times)
- What would need to be true for my current thinking to be completely wrong?
- What's invisible to me because it's too obvious, too familiar, or too uncomfortable?

Layer 3: Scale & Time Shifting
- How does this look at 10x scale? 1/10th scale?
- What's the 10-minute version? 10-year version? 100-year version?
- What second-order effects am I missing? Third-order?
- What does the 10th person in the chain of consequences experience?

Layer 4: Synthesis & Collision
- What happens when I combine this with something from a completely different domain?
- What's the metaphor from nature? From art? From history?
- What emerges when I hold two contradictory truths simultaneously?
- What pattern does this echo that I've seen elsewhere?

Layer 5: Action & Iteration
- What's the smallest experiment I could run today to test this?
- What would I need to observe to change my mind?
- What's reversible? What's irreversible?
- What do I need to learn to have an informed opinion rather than an inherited one?

The Meta-Question: "What lens am I not using that would transform how I see this?"