This article is about thinking with Generative AI i.e Claude, ChatGPT, Gemini etc. We cover the core principles of how generative AI actually reasons, and then we work through specific formulas for specific professional scenarios. Actual templates you can lift and use today.
Most people approach AI the same way they approach a search engine. They type what they want. They hope for the best (LOL).
Sometimes it works. More often, the output is generic, shallow, or just slightly off from what they needed. So they rephrase. Try again. Get a similar result.
The problem is not the tool. The problem is the mental model.
A search engine rewards keywords. A generative AI rewards thinking. The more clearly you have thought through what you need, who needs it, why it matters, and what good looks like, the more useful the output becomes.
Why most prompts underperform
There are three reasons most prompts produce disappointing results.
The first is vagueness. Summarise this document is a request, not a brief. Summarise this document for a board of directors who have not read the underlying research, focusing on the three most important decisions they need to make, in no more than 300 words is a brief.
The second is missing context. AI has no idea who you are, what you work on, who your audience is, or what good output looks like for your specific situation unless you tell it. The default output is calibrated for nobody in particular. Your output should be calibrated for you.
The third is not specifying the output. If you do not describe the format, length, tone, and structure you want, the AI will choose defaults. Sometimes those defaults are fine. Often they are not what you needed.
The universal prompt structure
Before we get to specific scenarios, here is the structure that underpins every effective prompt. Think of it as the skeleton you fill in for any task.
[ROLE] + [TASK] + [CONTEXT] + [CONSTRAINTS] + [OUTPUT FORMAT]
Role. Who is the AI being for this task. A senior lawyer. A plain-language writer. A skeptical stakeholder. A financial analyst. Giving the AI a role sharpens the lens it uses to approach the task.
Task. What specifically you want it to do. Not broadly help with, but specifically produce, analyse, draft, compare, summarise, critique, or explain.
Context. The background it needs to do the task well. Who the audience is. What the document is for. What decisions depend on it. What has already been decided. What it should not repeat.
Constraints. The boundaries. Length. Tone. What to include and what to leave out. What assumptions it should not make. What format it cannot use.
Output format. Exactly how you want the result delivered. Bullet points or prose. Headers or flowing text. Three options or one recommendation. A table or a narrative.
Not every prompt needs all five elements. But the more context and structure you give, the closer the first output lands to what you actually need.
Part One: Writing and communication prompts
Formula 1: The stakeholder update
This is one of the most common professional writing tasks and one of the most frequently underprompted.
You are a [role] writing a [type of update] for [audience].
Here is what happened this week: [bullet points or notes].
The tone should be [confident/transparent/reassuring].
Length: [word count]. Include: what happened, what is at risk, what is next.
Do not use corporate filler phrases. Get to the point.
Applied: You are a product manager writing a weekly status update for a CTO and Head of Engineering. Here is what happened this week: the payment integration is delayed by three days due to a third-party API issue, user testing on the onboarding flow produced strong results with one critical finding around the email verification step, and the sprint is otherwise on track. The tone should be transparent and confident. Maximum 200 words. Include what happened, what is at risk, and what the next steps are. Do not use phrases like it is important to note or as we move forward.
Formula 2: The first-draft email
The briefing that produces a usable email on the first try has four components: the relationship, the ask, the context, and the tone.
Draft an email to [recipient and relationship].
The purpose is to [specific ask or outcome].
Background they need: [relevant context].
Tone: [formal/warm/direct/brief]. Length: [short/medium].
Subject line included. No filler opener or closer.
Applied: Draft an email to a client I have not spoken to in six months. The purpose is to reconnect and suggest a call to discuss how their business has evolved and whether we can help with a new challenge we have seen in their industry. Background: they are a logistics company and the challenge is AI-powered route optimisation, which is now significantly reducing costs for their competitors. Tone: warm and direct. Short. Subject line included. No filler opener.
Formula 3: The executive summary
Executive summaries fail when they summarise everything equally. The formula forces you to decide what matters before you ask for the writing.
Write an executive summary of the following [document type].
Audience: [who will read it and what they care about].
The three most important points are: [list them].
The decision or action this should drive is: [state it].
Maximum [word count]. No background section. Lead with the conclusion.
Applied: Write an executive summary of the following market research report. Audience: a board of directors deciding whether to expand into the Nigerian market. The three most important points are: the market is growing at 23% annually, three competitors have already entered with mixed results, and the primary barrier to entry is distribution rather than demand. The decision this should drive is whether to approve a six-month pilot. Maximum 250 words. No background section. Lead with the recommendation.
Formula 4: Adapting tone for different audiences
The same information needs to land differently for different audiences. This formula handles the translation.
Rewrite the following [content] for [new audience].
Original audience was: [who].
New audience is: [who, what they know, what they care about].
Keep all the key information. Change the language, analogies, and emphasis.
Tone should be [technical/plain/formal/conversational].
Applied: Rewrite the following technical architecture document for a Chief Financial Officer who has a strong business background but no engineering background. The original was written for engineers. The CFO cares about cost, reliability, risk, and timeline. Keep all the key decisions and their implications. Replace technical terminology with business equivalents where possible. Tone should be clear and professional, not condescending.
Part Two: Analysis and research prompts
Formula 5: The structured analysis
When you need analysis rather than a summary, the formula specifies the analytical frame.
Analyse [subject] from the perspective of [analytical frame].
Context: [what you are trying to decide or understand].
Specifically address: [list the questions you need answered].
Flag assumptions you are making where the information is incomplete.
Structure: [pros and cons / SWOT / numbered points / recommendation with rationale].
Applied: Analyse the decision to migrate our monolithic application to microservices from the perspective of a senior engineer who has done this three times before. Context: we are a 40-person company growing at 60% annually and our current system is causing deployment bottlenecks. Specifically address: the realistic timeline, the cost in engineering time, the risks during migration, and what signs would tell us we are ready. Flag any assumptions you are making. Structure as a recommendation with rationale.
Formula 6: Competitive research synthesis
For market and competitive intelligence, the formula pulls the analysis toward decision-relevance rather than description.
Research [competitor or market topic] and produce a structured briefing.
I am a [role] at a [type of company] focused on [specific area].
What I need to know: [specific questions or areas].
What would change my strategy if I knew it: [state the decision it informs].
Format: short overview, then bullet points for each question, then one key implication.
Applied: Research how Stripe has approached developer experience over the last two years and produce a structured briefing. I am a head of product at a payments company reviewing our own developer experience strategy. What I need to know: what Stripe has changed in their documentation, onboarding flow, and API design, and what reaction developers have had. What would change my strategy: evidence of what developers prioritise most when choosing a payments API. Format: short overview, then bullet points for each area, then one key implication for our roadmap.
Formula 7: The devil’s advocate
One of the most useful ways to use AI is to stress-test a decision before you commit to it. This formula makes AI play the skeptic.
I am about to [decision or plan]. Challenge this.
Identify the assumptions I am making that might not be true.
Surface the risks I have not mentioned.
Ask the questions a skeptical [stakeholder type] would ask.
Do not soften the critique. I need the hard version.
Applied: I am about to launch a new pricing tier at three times our current price, targeting enterprise customers, with the belief that our product is ready for enterprise scale and that our current customers will not churn. Challenge this. Identify the assumptions I am making that might not hold. Surface the risks I have not mentioned. Ask the questions a skeptical board member and a skeptical head of customer success would each ask. Do not soften the critique.
Formula 8: Synthesising from multiple sources
When you have a set of documents, notes, or data and need to pull insights across them, this formula structures the synthesis.
I am going to share [number] [document types] with you.
After reading all of them, identify: [list of specific things].
What I am trying to understand is: [the question you are answering].
Present findings as: [themes and supporting evidence / ranked list / comparison table].
Flag where sources contradict each other.
Applied: I am going to share six user interview transcripts with you. After reading all of them, identify the top three recurring pain points, the moments where users expressed the most frustration, and any needs they described that our product does not currently address. What I am trying to understand is whether there is a pattern in what is causing our activation rate to be low. Present findings as themes with supporting quotes from the transcripts. Flag where users contradict each other.
Part Three: Strategic thinking prompts
Formula 9: The scenario planner
For decisions with significant uncertainty, AI can help you think through scenarios you have not considered.
I need to make a decision about [topic] under uncertainty.
The key variables I cannot control are: [list them].
Help me think through three scenarios: best case, worst case, most likely.
For each scenario, describe: what happens, what the signals would be early on,
and what I should do differently in each case.
Then recommend the decision that performs best across all three scenarios.
Applied: I need to decide whether to hire three engineers now or wait six months. The key variables I cannot control are: how quickly our pipeline converts to revenue, whether a key competitor launches a similar product in the next quarter, and whether our current two engineers can sustain the current velocity. Help me think through three scenarios: best case, worst case, most likely. For each describe what happens to the business, what early signals would tell me which scenario is unfolding, and what I should do differently. Then recommend the decision that performs best across all three.
Formula 10: The priority framework
When you have too many things to do and need help ordering them, this formula forces the AI to reason about trade-offs rather than just list things.
Here is a list of [tasks or initiatives]: [list them].
My constraints are: [time, team, budget, dependencies].
My goal for the next [time period] is: [specific outcome].
Prioritise this list using [framework or criteria].
Explain the reasoning for the top three, especially where you deprioritised
something that seems important.
Applied: Here is my product backlog for the next quarter: improve onboarding flow, rebuild the reporting module, add API webhooks, fix the mobile navigation bug, launch the enterprise tier, and run a pricing experiment. My constraints are two engineers for three months and a sales team that needs something they can close enterprise deals with. My goal for the next quarter is to close our first three enterprise contracts. Prioritise this list using impact on the goal versus effort. Explain the reasoning for the top three, especially where you deprioritised something that seems important.
Formula 11: The strategy memo
When you need to think through a strategic question before you can write about it, this formula uses AI as a thinking partner first.
Help me think through the following strategic question: [question].
Here is the context: [situation, constraints, what you already know].
I am not ready for a recommendation yet. First, help me identify:
what information I am missing, what assumptions I am making,
and what the strongest argument against my current instinct is.
Then, if the information I have is sufficient, give me your recommendation.
Applied: Help me think through whether to position our product as a tool for individual contributors or a platform for teams. Here is the context: we have 800 users, 70% of whom signed up individually but 40% of whom have invited at least one colleague. Our revenue is higher from individual plans but churn is higher too. I am not ready for a recommendation yet. First help me identify what information I am missing, what assumptions I am making about where the growth will come from, and what the strongest argument is against my current instinct to go platform. Then if what I have told you is sufficient, give me a recommendation.
Part Four: Thinking prompts — using AI as a collaborator
Formula 12: The Socratic brief
Sometimes you do not know what you do not know. This formula asks the AI to ask you questions before it does anything.
I want to [goal or task]. Before you produce anything,
ask me the questions you would need answered to do this well.
Ask a maximum of five questions. Ask the most important ones first.
Once I answer, proceed.
Applied: I want to redesign our customer onboarding email sequence. Before you produce anything, ask me the questions you would need answered to do this well. Maximum five questions. Most important ones first. Once I answer, proceed.
Formula 13: The thinking out loud brief
When you have a half-formed idea and need to develop it, this formula creates space for AI-assisted thinking rather than AI-generated output.
I am going to think out loud about [topic]. Do not interrupt.
When I am done, reflect back: what you heard as my core argument,
what assumptions seem shaky, what I left out,
and what question I should be asking that I have not asked yet.
Applied: I am going to think out loud about why our activation rate dropped last month. Do not interrupt. When I am done, reflect back: what you heard as my working theory, what assumptions seem shaky based on what I said, what I might have left out, and the one question I should be asking that I have not asked yet.
Formula 14: The reframe
When you are stuck, sometimes the most useful thing AI can do is show you the problem differently.
I have been thinking about [problem] as [current framing].
Give me three alternative ways to frame this problem.
For each framing, describe: what becomes visible that was hidden before,
what it implies about where the solution might be,
and what question each framing leads to.
Applied: I have been thinking about our high churn rate as a product quality problem. Give me three alternative ways to frame this problem. For each framing describe: what becomes visible that was hidden before, what it implies about where the solution might be, and what question that framing leads to.
Formula 15: The post-mortem brief
Learning from what went wrong requires structured reflection. This formula produces analysis rather than a timeline.
Here is what happened: [timeline of events].
Do not summarise the timeline back to me. Instead, identify:
the earliest decision point where a different choice would have changed the outcome,
the assumptions that turned out to be wrong,
the signals that were present but ignored or missed,
and the one systemic change that would make this least likely to happen again.
Applied: Here is what happened with our product launch last month: we announced a launch date six weeks out, hit a critical API bug in week four, delayed the launch by ten days, lost three press opportunities, and had to give several existing customers a credit. Do not summarise this back to me. Identify the earliest decision point where a different choice would have changed the outcome, the assumptions that turned out to be wrong, the signals that were present but we missed, and the one systemic change that makes this least likely to happen again.
Part Five: The principles that apply to every formula
Lead with what you want, not how you want it done
Most prompts over-specify the method and under-specify the outcome. Rewrite this more simply tells the AI how to approach the task. Write this so that a first-year analyst with no context on our company can understand the key decision and what is being asked of them tells the AI what success looks like.
Specify the outcome. Let the AI choose the method.
Iterate in conversation, not in separate prompts
The most common mistake after a mediocre output is starting a new conversation and trying a different prompt. The better move is to reply in the same conversation with specific feedback.
The second section is too abstract. Make it concrete with an example from our industry.
The tone is too formal. Write it as if you are explaining this to a colleague over coffee.
You buried the recommendation. Lead with it.
Each correction teaches the AI what you need. By the third or fourth exchange, the output has been calibrated to your specific situation in a way that a single prompt can never achieve.
Tell it what not to do as much as what to do
Negative constraints are often more useful than positive ones. Do not use bullet points. Do not summarise what I just told you back to me. Do not hedge every statement with it is important to note. Do not give me options when I asked for a recommendation.
AI will default to certain patterns: over-qualification, excessive structure, unnecessary summaries of the input, and a tendency to present options rather than make choices. Explicitly telling it to avoid these defaults produces significantly cleaner output.
Give it permission to be direct
AI defaults toward balance, diplomacy, and even-handedness. For most professional tasks, you want the opposite: a clear recommendation, a direct critique, a confident recommendation rather than a list of considerations.
Phrases that unlock directness: give me the hard version, do not soften the critique, make a recommendation, do not hedge, tell me what you actually think, not what is technically possible.
The AI knows what it thinks. You have to give it permission to say it.
Use personas strategically
The role you assign the AI changes the answer dramatically. A senior lawyer and a startup founder will approach the same contract clause completely differently. A skeptical CFO and an optimistic sales director will evaluate the same market opportunity differently.
When you are trying to anticipate how different stakeholders will respond to something, assign those stakeholders as personas and ask the AI to evaluate from each perspective. You will get a richer set of objections and considerations than any individual review produces.
Reflection for the week
The gap between a mediocre prompt and a great one is not creativity. It is clarity.
Clarity about what you are asking for. Clarity about who needs it. Clarity about what good looks like. Clarity about what the output needs to do.
The formulas in this article are not tricks. They are structures that force you to think clearly before you type. When you have thought clearly about what you need, the AI has the information it needs to produce it.
The professionals who get the most from generative AI are not the ones who have found the best prompts. They are the ones who have become better at thinking about what they actually need before they ask for it.
That skill transfers to everything. Not just AI. Every brief you write. Every meeting you run. Every decision you bring to a team.
The question worth sitting with this week:
Pick the one scenario from this article that maps most closely to something you do regularly. Use the formula once this week on a real task. What did the brief force you to think about that you had not thought about before?
Want to build this skill properly?
Thinking with AI, configuring your tools, and building workflows that actually change how you work is what we teach in our 12-week AI for Business Professionals programme starting next month.
Live classes. Real tasks. A cohort of people building alongside you.
Apply here: amakoragroup.com/apply
Until next week,
Tochii
Founder, Learn with Tochii
Inspire · Educate · Empower
📧 contact@tochukwuachebe.com
🌐 https://www.tochukwuachebe.com