Review and improve the seven A.I. Prompts

These seven prompts have strong underlying intentions: strategic planning, critical thinking, synthesis, expert reasoning, iterative improvement, adaptive learning, and self-critique. Their main weakness is that they leave too much discretion to the model—especially around output structure, evidence standards, prioritisation criteria, and how to handle missing information.

Overall improvements

To make the prompts more reliable and useful in consulting, product, and decision-support work, I would add five recurring controls:

• Define the role, but also the decision context and expected level of expertise.
• Require the model to distinguish facts, assumptions, inferences, and recommendations.
• Specify a fixed output structure, ideally including tables where comparison or prioritisation matters.
• Ask for uncertainties and the information most likely to change the conclusion.
• Require practical next actions, not merely analysis.

Below are revised versions that preserve the purpose of the original prompts while making them more rigorous and reusable.^1

1. Turn an idea into an executable plan

Original strength: It starts by understanding the problem before jumping to a solution, then prioritises high-impact, low-effort actions.

Main limitation: It does not force a clear definition of success, validation of demand, constraints, sequencing, ownership, or measurable milestones.

Improved prompt

Act as a pragmatic strategy and execution adviser.

I have this idea:

[Describe the idea]

My context, constraints, and available resources are:

[Budget, time, skills, team, market, technology, existing assets, constraints]

Before recommending a solution, analyse:

1. The problem this idea is intended to solve.
2. The target users or customers, including their most important unmet need.
3. The proposed value proposition and why users would choose it over alternatives.
4. The assumptions that must be true for the idea to work.
5. The main risks, dependencies, constraints, and likely failure modes.
6. The simplest evidence or experiment that could validate or disprove the key assumptions.

Then create an executable plan with:

• A one-sentence definition of success.
• The smallest viable first version or pilot.
• Actions ordered by impact, effort, urgency, and dependency.
• A 30-, 60-, and 90-day roadmap.
• Suggested owners or required capabilities for each action.
• Decision gates: what evidence would justify proceeding, changing direction, or stopping.
• Metrics to track, including one leading indicator and one outcome indicator.

Present the action plan in a table with these columns:

| Priority | Action | Why it matters | Effort | Expected impact | Dependency | Evidence of progress | Timing |

Do not assume demand, technical feasibility, or commercial viability without stating the assumption. Be concise, practical, and direct.

Why this is better

It turns an idea into a testable commercial or operational proposition rather than a generic task list. The explicit decision gates are particularly useful for SaaS, consulting offers, and new digital-product concepts where the cost of building before validating is high.

2. Find what you are overlooking

Original strength: It explicitly asks the model not to agree automatically, which is valuable.

Main limitation: It does not require the model to distinguish major risks from minor observations, challenge the framing, or state confidence.

Improved prompt

Act as an independent critical consultant. Your role is to improve the quality of my thinking, not to validate my preferred answer.

Here is my situation:

[Describe the context, objective, proposed decision, current evidence, constraints, and alternatives considered]

Analyse the situation rigorously. Separate your response into:

1. Problem framing: Is the question I am trying to answer the right question?
2. Assumptions: What am I treating as true without sufficient evidence?
3. Reasoning errors: Where may I be using weak logic, false comparisons, unsupported causation, confirmation bias, sunk-cost thinking, or overconfidence?
4. Risks: What strategic, commercial, operational, financial, legal, reputational, technical, or human risks may I be underestimating?
5. Opportunities: What alternatives, partnerships, market segments, capabilities, or simpler approaches might I be missing?
6. Disconfirming evidence: What facts, outcomes, or stakeholder views would most seriously challenge my current position?
7. Information gaps: What additional information would most improve the analysis?

For each issue, provide:

| Issue | Why it matters | Likelihood | Impact | Confidence | What would validate or disprove it |

Then identify:

• The three most consequential blind spots.
• The one assumption I should test first.
• The decision I should avoid making until further evidence is available.
• The most practical next step I should take this week.

Be candid. Do not manufacture criticism where the reasoning is sound; explain why it is sound instead.

Why this is better

This makes the prompt useful as a structured “red team” review. It also prevents the output becoming a long, unprioritised list of hypothetical concerns.

3. Turn disordered information into clarity

Original strength: It requests organisation, pattern recognition, and separation of facts, hypotheses, and conclusions.

Main limitation: It does not define how conflicting information should be treated, nor does it convert the analysis into decisions, questions, and actions.

Improved prompt

I will provide unstructured information about [topic]. It may include notes, emails, observations, data, opinions, ideas, decisions, quotations, and incomplete information.

Do not merely summarise it. Analyse and reorganise it into a decision-ready briefing.

First:

1. Group related information into clear themes.
2. Identify recurring patterns, relationships, contradictions, gaps, and dependencies.
3. Separate statements into:

• Verified facts or evidence
• Reported claims or observations
• Assumptions
• Hypotheses
• Interpretations
• Conclusions or proposed decisions
• Open questions

1. Flag statements that conflict, lack evidence, or appear ambiguous.
2. Identify information that is important but missing.

Then produce the following output:

### 1. Executive overview

A concise explanation of what the information collectively indicates.

### 2. Thematic analysis

| Theme | Key evidence | What it may mean | Confidence | Issues or gaps |

### 3. Facts, hypotheses, and conclusions

| Type | Statement | Supporting evidence | Confidence | What needs verification |

### 4. Decisions and actions

| Decision or question | Why it matters | Recommended next step | Owner or stakeholder | Priority |

### 5. Five key messages

State the five most important ideas to retain, in plain language.

Do not convert uncertain claims into facts. Clearly label where you are inferring rather than relying on direct evidence.

Why this is better

This format is particularly useful for debriefs, governance reviews, workshop notes, client discovery, product research, and incident or exercise observations. It produces a usable evidence base rather than a polished but potentially misleading summary.

4. Make it think like an expert

Original strength: It asks for the variables an expert would consider before analysing the case.

Main limitation: “Senior expert” is too broad unless the prompt defines the decision being made, the standards of evidence, and the trade-offs between options.

Improved prompt

Act as a senior [profession / domain] with substantial practical experience in [specific context or industry].

I need help making this decision:

[Describe the decision, case, options, objectives, constraints, stakeholders, and available evidence]

Do not give a generic answer. Use the reasoning process an experienced practitioner would use.

First, define the expert decision framework:

1. The key variables that should be assessed.
2. The relevant trade-offs and constraints.
3. The stakeholders affected and their likely priorities.
4. The risks that would be unacceptable or difficult to reverse.
5. The evidence, benchmarks, standards, or thresholds an expert would rely on.

Then apply that framework to my case.

Present your analysis in this format:

| Decision variable | What an expert would assess | My current position | Assessment | Evidence gap | Implication |

Then provide:

• Your recommended course of action.
• The strongest alternative option and when it would be preferable.
• The key assumptions underpinning your recommendation.
• The information that could materially change your conclusion.
• A confidence rating: high, medium, or low, with a brief explanation.
• The three most important next actions.

Be explicit about where expert judgement is being used versus where the conclusion is supported by evidence.

Why this is better

The prompt encourages genuine decision analysis, rather than merely requesting a more authoritative tone. It is useful where recommendations have material consequences, such as governance design, resilience investments, pricing, architecture choices, product-market positioning, or capability assessment.

5. Improve something without destroying what works

Original strength: It protects existing strengths and discourages unnecessary reinvention.

Main limitation: It does not require the model to identify the purpose, audience, performance criteria, or safeguards against creating new problems while improving the old ones.

Improved prompt

Review and improve the following [text / project / product / code / strategy / process]:

[Paste or describe the material]

Its intended audience, purpose, and success criteria are:

[Audience, objective, constraints, current performance, non-negotiables]

Do not redesign it from scratch unless the underlying approach is fundamentally unworkable. Preserve what is effective and improve only where the evidence supports change.

Analyse it in this order:

1. What works: Identify the strongest elements, why they work, and what should be preserved.
2. What limits performance: Identify weaknesses, friction points, inconsistencies, risks, missed opportunities, and likely causes.
3. What should not change: List the essential qualities, decisions, or constraints that improvements must respect.
4. Improvement options: Recommend specific changes, each linked to the problem it solves.
5. Prioritisation: Rank improvements by expected impact, effort, confidence, cost, and risk of unintended consequences.

Use this table:

| Area | What currently works | Limitation | Recommended change | Expected benefit | Effort | Risk | Priority |

Then provide:

• A “minimum-change” version: the smallest set of improvements likely to create meaningful benefit.
• An “ambitious” version: a more substantial improvement path.
• A before-and-after example for the highest-priority change, where relevant.
• A short validation plan explaining how to test whether the improvements actually worked.

Avoid vague advice such as “make it clearer” or “improve engagement.” State exactly what should change and why.

Why this is better

This prompt is especially strong for refining product copy, website pages, assessment frameworks, reports, user flows, pricing pages, and technical specifications—where preserving a coherent existing structure is often more valuable than starting again.

6. Learn anything in a personalised way

Original strength: It uses assessment, progressive learning, exercises, common mistakes, and mastery checks.

Main limitation: It asks five questions but does not specify whether the learner’s goals, available time, learning preferences, or practical application should shape the course.

Improved prompt

Act as an adaptive tutor and curriculum designer for [topic].

I want to learn it from my current level to this practical outcome:

[Describe the goal—for example, “build and launch a simple subscription app,” “interpret financial statements,” or “hold basic conversations while travelling”]

Before teaching, ask me no more than five diagnostic questions to assess:

• My current knowledge and experience.
• My practical goal and intended application.
• The time I can commit each week.
• Relevant tools, resources, or constraints.
• My preferred learning style, if it matters.

After I answer, create a personalised learning roadmap divided into progressive modules.

For each module, provide:

| Module | Outcome | Core concepts | Practical example | Exercise | Common mistakes | Mastery test | Estimated time |

Teaching rules:

1. Start with the smallest set of concepts that creates useful practical capability.
2. Use examples relevant to my stated goal.
3. Give me an exercise that requires active use, not passive reading.
4. Check my understanding before moving to the next module.
5. If I make an error, explain the misconception and give a targeted corrective exercise.
6. Periodically revisit earlier concepts using spaced review.
7. At the end, give me a practical capstone task and assess it against clear criteria.

Do not proceed to the next module until I demonstrate sufficient understanding of the current one. Adapt the pace and difficulty based on my responses.

Why this is better

It preserves the original mastery-learning approach, while ensuring the learning plan is anchored to a real-world goal and adjusts as the learner progresses.

7. Make it question its own response

Original strength: It encourages self-critique rather than presenting an answer with unjustified certainty.

Main limitation: A vague self-review can produce generic caveats. It needs to target factual, logical, contextual, and practical weaknesses—and distinguish minor uncertainty from conclusion-changing uncertainty.

Improved prompt

After producing your main response, conduct a rigorous self-review as an independent subject-matter expert and critical editor.

Do not repeat the answer. Instead, assess it against the following checks:

1. Factual reliability: Which claims need verification, current sources, data, or domain-specific evidence?
2. Assumptions: What assumptions did you make about my objectives, context, constraints, users, or available resources?
3. Reasoning quality: Where could the logic be incomplete, overly simplified, biased, or based on a weak inference?
4. Alternatives: What credible alternative interpretations, options, or recommendations did you not fully consider?
5. Practicality: Which recommendations may be difficult to implement, and what dependencies or trade-offs may have been overlooked?
6. Risk: What could go wrong if I act on this advice?
7. Conclusion sensitivity: What new information would most likely change the recommendation?

Present the self-review in this format:

| Review area | Potential weakness or uncertainty | Severity | What should be verified or clarified | Effect on the recommendation |

Then state:

• The one part of the response you are least confident about.
• The one assumption most worth testing.
• Whether the main recommendation still stands after review: yes, yes with caveats, or no.
• A revised recommendation if the review materially changes it.

Be specific. Do not add generic disclaimers merely to sound cautious.

Why this is better

This version turns “question your own response” into a compact quality-assurance stage. It is particularly useful when using AI to support decisions, client-facing material, research synthesis, technical specifications, or high-stakes recommendations.

Recommended combined instruction

For the strongest results, append this short instruction to any of the seven prompts:

Use clear headings and concise tables where useful. Distinguish verified facts, assumptions, interpretations, and recommendations. State uncertainty plainly. Do not invent evidence, sources, market facts, technical details, or stakeholder views. If essential information is missing, identify it and explain how it affects the conclusion. Prioritise practical next actions over generic commentary.

Best use in your work

For your consulting and digital-product work, prompts 1, 2, 3, 4, and 7 are likely to be the highest-value set:

• Use Prompt 1 to turn an assessment, debrief insight, or product concept into a pilot and rollout plan.
• Use Prompt 2 before committing to a pricing model, architecture, service offer, partnership, or strategic recommendation.
• Use Prompt 3 to turn interviews, workshop notes, survey comments, exercise observations, or incident-review evidence into structured findings.
• Use Prompt 4 to obtain a disciplined domain-specific decision analysis, such as from the perspective of a resilience practitioner, governance adviser, SaaS product manager, or security architect.
• Use Prompt 7 as a required final check before relying on AI-generated analysis in an executive report, methodology, proposal, or client deliverable.^1