35 ChatGPT Prompts for Business Strategy and Operations

A model is genuinely useful for business decisions in one specific way: it will tell you what is wrong with your reasoning faster than anybody on your team will. It is useless at the thing people ask it for most, which is producing a plan from nothing.

So the thirty-five below assume you already know something. Nearly all of them ask you to paste your actual numbers, your actual positioning or your actual plan, and then ask the model to find the flaw. That is where the value is, and it is why almost none of these will work if you describe your business in two lines.

They cover strategy, customers, pricing, operations and decision review. Channel work sits elsewhere: our ChatGPT prompts for digital marketing cover campaigns, ChatGPT prompts for SEO cover search, and ChatGPT prompts for data analysis cover the numbers themselves. All of it works in ChatGPT and any comparable model.

• It does not have your numbers. Ask for a financial projection and you will get plausible fiction with a convincing structure. Paste your actual figures and ask it to analyse them, and the same prompt becomes useful. Never let it supply a number you could have looked up.
• It has no current market data. Competitor analysis without pasted material is invention. Market sizing without a source is a guess presented as a figure. Paste the pages, the pricing and the reports, or ask it what to go and find rather than what the answer is.
• Legal, tax and employment questions need a professional. A model can help you prepare the questions, understand the vocabulary and spot what you have not considered. It cannot advise on your obligations, and acting on a confident answer about contracts, tax structure or employment terms is a genuinely expensive way to find that out.
• Think about what you paste. Financials, customer lists, contracts and unreleased plans are the things people paste without pausing. Check what your tools retain, and redact what the prompt does not actually need - most of these work perfectly well with anonymised figures.
• It does not know your history. The thing that was tried in 2022 and failed for reasons nobody wrote down is invisible to it. Supply that context or you will get advice that is sound in general and wrong here.

Seven prompts about what the business is for and who it is against. None of them writes a strategy - they interrogate the one you have, which is the version worth doing.

1. What we are actually selling

Here is how we describe what we do: [paste your current positioning]. Here is what customers say when they explain us to somebody else: [paste actual quotes if you have them]. Tell me where those two diverge, which version is doing better work, and what we would have to change about the business if the customer version is the accurate one.

2. The alternative we actually lose to

When somebody decides not to buy from us, here is what they do instead: [describe, including doing nothing and building it themselves]. For each alternative, tell me what it offers that we do not, what it costs them that we do not, and which of these we are losing to most often based on what I have described. Do not assume our named competitors are the real answer.

3. Where the money actually comes from

Here is our revenue broken down by product, customer type and channel: [paste the figures]. Identify where the concentration risk is, which segment is quietly subsidising another, and which line of revenue we would notice least if it disappeared. Then tell me which one we spend the most attention on relative to what it returns.

4. The strategy stated as a bet

Here is our plan for the next twelve months: [describe it]. Restate it as a set of explicit bets, each in the form we believe X, therefore we are doing Y, and we will know we were wrong if Z. Then tell me which bet the whole plan depends on, and how long before we would find out it was wrong.

5. What we should stop

Here is everything we currently do: [list products, channels, services and internal projects with rough effort and return]. Identify the two we should stop entirely. For each, state what it frees up, what we would lose, and what we are likely to tell ourselves in order to keep doing it anyway.

6. Growth that does not scale

Here is how we currently acquire customers and roughly what each channel costs us: [describe]. Tell me which of these will stop working if we double, and why. For each, say whether the ceiling is money, time, my personal involvement, or market size, because those need completely different responses.

7. Read our own materials as a stranger

Here is our homepage and our main sales page: [paste the text]. Read them as somebody who has never heard of us and has a problem we might solve. Tell me what you understand us to do, what you are unsure about, and at what point you would leave. Do not be generous.

Prompt 4 is the one worth doing annually. A strategy written as prose can absorb any outcome; a strategy written as bets with falsification conditions cannot, which is exactly why it is uncomfortable.

Seven prompts about the people who pay you. Every one of them works from something real - transcripts, tickets, reviews, churn reasons - because a persona invented from nothing is a description of a customer who does not exist.

8. Their words, not ours

Here are reviews, support tickets and sales call notes: [paste them]. Extract the exact phrases customers use to describe the problem we solve. Show me where their language differs from ours, and tell me which of our terms nobody outside the company uses. Do not paraphrase - I want their words.

9. Why they actually left

Here are the stated reasons customers gave for leaving over the past year: [paste them]. Group them, and for each group tell me what the underlying cause might be as distinct from the reason given. Flag which groups we could have prevented and which we could not, and be honest about which ones we probably misdiagnosed.

10. The segment we did not plan for

Here is our customer base broken down by whatever attributes we track: [paste the data]. Identify any segment growing faster than we designed for, any segment that costs disproportionately to serve, and any that we assumed would matter and does not. Give me three hypotheses for the fastest-growing one.

11. Questions before assumptions

I am about to interview ten customers about [describe what you want to learn]. Write the question list. Every question must be about what they actually did rather than what they would do, and none may suggest the answer. Then tell me which two questions will produce the most useful answers and which one I will be tempted to add that would bias the whole thing.

12. Read the competitor properly

Here is a competitor's pricing page, homepage and three recent announcements: [paste them]. Work only from this material. Tell me who they are positioning for, what they have decided not to serve, what their pricing implies about their cost structure, and what they appear to be betting on. Say clearly where you are inferring rather than reading.

13. The objection we keep meeting

Here is the objection we hear most in sales conversations and how we currently answer it: [describe both]. Tell me whether the objection is really about what it appears to be about. Then give me three better responses, and tell me which one requires us to change the product rather than the pitch.

14. Who we should not sell to

Here are our best and worst customer relationships, with what each is worth and what each costs to serve: [describe]. Identify the shared characteristics of the bad ones. Then write the two qualifying questions that would have identified them before we signed, and tell me what revenue we would have turned away.

Prompt 14 is the one that pays for itself. Most businesses know which customers they regret and have never written down what they had in common.

Seven prompts about the numbers. Paste real figures into every one of these. A model asked to model your business without your data will produce a spreadsheet-shaped object with nothing in it, and it will look convincing.

15. What our price signals

Here is our pricing and here is our closest competitors' pricing: [paste both]. Tell me what our price communicates about us relative to them, whether that matches how we describe ourselves, and where a customer comparing the two would be confused. Do not recommend a change yet.

16. Model the increase

Here are our current prices, volumes and margins: [paste the figures]. Model a ten per cent price increase. Show me the arithmetic, including how much volume we could lose before we were worse off. Use only my numbers and state explicitly any assumption you had to make.

17. Where the margin actually is

Here is revenue and direct cost by product or service line: [paste]. Rank them by contribution rather than by revenue. Identify anything that looks successful on revenue and is not, and tell me what a full costing would need to include that I have probably left out.

18. The cash question

Here is our monthly cash position, our payment terms in and out, and our fixed costs: [paste]. Identify the month where we are tightest and what causes it. Then tell me which single change to terms or timing would most improve the position, and what it would cost us in goodwill or price.

19. Restructure the packages

Here are our current packages and how many customers are on each: [paste]. Propose a restructure. For each proposed tier, state who it is for, what moves between tiers, and which existing customers would be worse off. Tell me honestly whether the distribution suggests our current middle tier is doing any work.

20. Discounting, examined

Here is every discount we gave last year and why: [paste]. Calculate what they cost us in total. Group them by reason, tell me which category was genuinely necessary, and write the discount policy that would have prevented the rest without losing the deals we actually needed.

21. Prepare the questions for an accountant

I am meeting an accountant about [describe the situation]. Prepare me: the questions I should ask, the documents I should bring, the vocabulary I should understand beforehand, and the things I might not know to ask about. Do not give me the answers - I am paying somebody qualified for those.

Prompt 21 is the pattern to reuse for anything legal, tax or regulatory. A model that prepares you for a professional conversation is genuinely valuable; one that replaces it is a liability with a confident tone.

Seven prompts about how the work gets done and who does it. The people ones are written to help you think rather than to produce documents you would sign without reading.

22. Where the time actually goes

Here is how my week actually breaks down: [describe honestly, including interruptions and admin]. Identify what only I can do, what somebody else could do at eighty per cent quality, and what should not be done at all. Then tell me which of those is hardest to give up and why that is usually a bad reason.

23. Document a process properly

Here is a process I currently do from memory: [describe it as you actually do it, including the bits you improvise]. Write it as a document somebody could follow without asking me. Mark every step where judgement is required, because those are the steps that cannot be delegated without training.

24. The bottleneck

Here is how work moves through our business from enquiry to delivery, with rough timings: [describe]. Identify the constraint. Then tell me what would happen if we removed it, where the constraint would move to next, and whether the second bottleneck is worse than the first.

25. What this hire is really for

I am considering hiring for [role]. Before writing any job description, ask me eight questions that would establish whether this role solves the actual problem or just absorbs the symptom. Then tell me the one thing I said that suggests I have not fully defined the job.

26. Interview questions that reveal something

Here is the role and the three things that would make somebody succeed in it: [describe]. Write six interview questions, each about something the candidate has actually done rather than what they would do. For each, tell me what a strong answer contains and what a rehearsed answer sounds like.

27. A conversation I am avoiding

I need to raise something difficult with somebody on my team: [describe the situation factually]. Help me prepare. What is the specific behaviour and its effect, what am I assuming about their intent that I should check, and what are the two most likely responses? Draft an opening that is direct without being an accusation. If this is a matter with formal or legal implications, say so and tell me to take advice first.

28. What breaks if I disappear

If I were unreachable for a month starting tomorrow, here is what I currently hold: [list relationships, approvals, knowledge and access]. Identify what would break first, what would break quietly, and what nobody would notice until it mattered. Give me the order to address them in.

Prompt 28 is uncomfortable and worth doing once a year. The things that would break quietly are almost always more expensive than the things that would break immediately.

The last seven, and the reason a model is worth having at all for this kind of work. Each one is designed to produce an answer you will not enjoy. Our ChatGPT prompts for problem solving go further on the same principle.

29. Argue against it properly

Here is a decision I have made and my reasoning: [describe both]. Make the strongest case against it - not weak objections I can dismiss, but the argument a well-informed sceptic who wanted us to succeed would make. Then tell me whether any of it should change my mind, and be direct rather than balanced.

30. The assumptions underneath

Here is the plan: [describe]. List every assumption it rests on, ordered from most to least load-bearing. For the top three, tell me what happens if each is wrong, what evidence would settle it, and how long that evidence takes to get. Then tell me which one to test before committing anything.

31. Write the failure in advance

Assume this initiative failed badly twelve months from now: [describe it]. Write the post-mortem as though it had happened. Cover the three most likely causes, the early warning signs each would have produced, and when those signs would first have been visible. Then tell me which sign I am most likely to explain away.

32. Define when to stop

We are about to start [describe]. Before we begin, define the conditions under which we abandon it: a measurable outcome, a threshold and a date. Include one condition that would be genuinely uncomfortable to act on. Then tell me what I will say to myself to avoid honouring these rules.

33. Second-order effects

We are considering [describe the change]. Map the consequences in two layers: what happens directly, then what those consequences cause, including effects on people and teams not involved in the decision. Stop at second order. Tell me which second-order effect is most likely to be the one nobody planned for.

34. Turn the mess into a brief

Here is everything I know about a decision, in the order I happened to write it down: [paste your notes, however messy]. Turn it into a one-page brief: the decision, the two or three real options, what each costs, what I would have to believe for each to be right, and a recommendation. Under 400 words. If the notes do not support a recommendation, say so and list what is missing.

35. Review it afterwards

Here is a decision we made and what actually happened: [describe both]. Separate the quality of the decision from the quality of the outcome - a good decision can have a bad result and the reverse. Tell me which this was, what we could have known at the time, and what this reveals about how we decide things generally.

Run prompt 35 on decisions that went well, not only the ones that went badly. It is the only way to find out which part was judgement and which part was luck.

Match the tier to the job rather than defaulting to whatever is selected. The decision prompts in the last group reward a model that works through a problem in steps rather than answering immediately, which is what GPT-5.6 Sol is built for. The generative ones - twenty ideas, six objections, three versions - run perfectly well on a lighter tier like DeepSeek V4 Flash, and cost you less while you are still exploring. The full model list is worth reading once so you know what you have.

Keep the ones that work. A prompt that produced a useful decision brief in March will do it again in September, and rebuilding it from memory each time is the most common waste in this whole workflow. Save them with the context you had to add, because the context is usually the part you forget. Chat Smith is free to try.

And the habit that decides whether any of this helps: notice when you rephrase a question because you did not like the answer. That is the moment the answer was worth having, and it is the only real test of whether you are using a model to think or to be agreed with.