ChatGPT Prompts for Strategic Planning (That Actually Do the Work)

Author: Eric Levine, Founder of StratEngine AI | Former Meta Strategy & Operations | UCLA Anderson MBA | CPA

Published: July 9, 2026

Reading time: 9 minutes

Summary

These ChatGPT prompts for strategic planning are organized by planning stage, from situation analysis through the finished plan. ChatGPT is at its best in the exploratory work of strategy: researching a market, generating options you hadn't considered, pressure-testing your own thinking before you commit real time to it. These prompts are built to get the most out of that phase.

Three principles separate a strategy prompt that works from a generic one: assign a real job with real inputs rather than a costume, constrain the output format so the model has to do the work, and give the model a way to say it doesn't know. Exploring is reversible; committing the company to one path is a different decision that rests on constraints and dependencies tracked over time.

There are fifteen prompts here, not five hundred, because a prompt is worth copying only if you understand why it's shaped the way it is, so each comes with a short note on what it's doing and why. The stages mirror the workflow in my guide to how to use AI for business strategy; this post is the prompt layer for it.

Key Takeaways

  • ChatGPT is at its best in the exploratory phase of strategy: research, option generation, and stress-testing your assumptions. These fifteen prompts are engineered for exactly that work.
  • Structure is what separates a prompt that produces real thinking from one that returns fluent filler. The prompts here specify roles, constraints, and output formats instead of asking open questions.
  • Always give real inputs: your numbers, your constraints, your competitors by name. A brilliantly worded prompt aimed at a vague business description still returns fluent nothing.
  • These prompts are for exploring options and sharpening your thinking. Once you're staking the business on a direction, that decision needs a tool that models your constraints over time, which is a different job.

What separates a strategy prompt that works from a generic one?

Most "strategic planning prompt" lists are open questions dressed up with a persona: "Act as a McKinsey consultant and analyze my business." That returns a confident, generic essay every time. The prompts below are built differently, on three principles that consistently produce sharper output.

First, assign a real job and real inputs, not a costume. "Score these three options against the criteria you committed to earlier" beats "act as a strategist." Second, constrain the output format, because a structure the model has to fill (a ranked list with no ties, a contradiction-only report) forces actual work where an open prompt invites an essay. Third, give the model a way to say "I don't know," so an honest gap shows up as a gap instead of a plausible-sounding guess.

Every prompt below applies at least one of those. Load in your real business, and they do real work.

Prompts for situation analysis

1. The sourced-or-unknown market map

You are a skeptical market analyst. Here is my business: [description, real numbers].
Map the market I compete in. Size segments only where you can name a source;
for anything you can't source, write UNKNOWN instead of estimating.
Then list the three claims in my own description you would challenge first,
and what evidence would settle each one.

Why it works: the UNKNOWN instruction gives the model permission to admit gaps, which turns a research pass into something you can actually build on. The follow-up question turns the model into a skeptic of your own framing, which is where the useful surprises live.

2. The contradiction-only data read

Here is how I describe my customers: [description].
Here is last quarter's actual behavior data: [paste].
Report ONLY where the data contradicts my description.
Do not summarize the data back to me. If nothing contradicts, say so and stop.

Why it works: pointing the model at the gap between what you believe and what your data shows is where it earns its keep, and restricting the output to contradictions keeps it from padding the answer with a summary you already know.

3. The wrong-read competitor pass

Here are my main competitors and what I believe about each one: [list].
For each, make the strongest case that my read is wrong.
Where your case and my belief conflict, name the public evidence that would settle it.

Why it works: "analyze my competitors" gets you their marketing paraphrased; asking the model to argue against your current reads produces positions you can actually go test. For the framework version of this stage, I've written a separate walkthrough on how to generate a SWOT analysis with AI.

AI prompts for business strategy: identifying strategic options

4. Options that actually conflict

Generate three genuinely different strategies for [goal].
Different means they sacrifice different things, not that they're worded differently.
For each: what it wins, what it explicitly gives up, and what kind of company
usually fails executing it. If any two could be pursued at once, they are not
different enough. Redo them.

Why it works: the sacrifice requirement and the "could you do both?" test force genuinely divergent options instead of three wordings of the same safe plan, which is exactly the kind of range you want an exploratory pass to give you.

5. The prosecution

Here is my current plan: [paste].
Your only job is to make the strongest case that it fails.
No praise, no balance, no "however, overall." End with the single weakest link.

Why it works: models are trained to be agreeable and lean toward reassurance when you ask for balanced feedback (Sharma et al., ICLR 2024), so assigning a single adversarial job is how you get the objections you actually need to hear before you commit. If the "weakest link" comes back vague like "execution risk," ask it to name the specific week things break.

6. Criteria before options

Do not ask for my options yet. First: for a business with [constraints],
list and rank the criteria any strategy for [goal] must meet.
I'll share my options in the next message, and you'll score them
against the criteria you just committed to.

Why it works: locking the criteria before the model sees your options keeps the scoring anchored to what actually matters rather than bending toward whichever option you seemed to favor. It's a simple sequencing move that meaningfully sharpens the comparison.

Prompts for prioritization and trade-offs

This is the section that separates a real prompt from a generic one. Ask ChatGPT to "prioritize my initiatives" and you get a reshuffled list with everything still on it. The prompts here impose the constraints that make a ranking mean something, so the exploratory pass actually pushes your thinking toward a decision instead of describing the options back to you.

7. The budget-capped ranking

Here are my initiatives and my real constraints: [initiatives with rough costs;
team size, cash, deadlines]. Rank them. The top three must fit inside [budget/headcount].
No ties. For every initiative below the cut, one sentence on what breaks by deferring it.

Why it works: the budget cap and the no-ties rule force the model to actually trade options off against each other instead of endorsing all of them. The "what breaks by deferring it" line is what makes the exercise useful: it surfaces the cost of each cut so you're reasoning about consequences, not just order.

8. The kill list

I have to kill two of these five initiatives today: [list].
Pick the two. Defend the choice, then tell me exactly what I lose and when I'd feel it.
If you find yourself recommending I keep all five in some form, start over.

Why it works: "deprioritize" is where trade-offs go to hide. Forcing a kill with a felt cost surfaces the model's actual read instead of a diplomatic shuffle.

9. The dependency map with an empty option

Map the dependencies between these initiatives: [list].
Output only lines of the form "If X slips, Y breaks, because Z."
If you cannot find a real dependency, output NONE. Do not invent connections.

Why it works: the strict output form and the NONE option get you a clean first-draft dependency map instead of a hand-wave, and the model is good at surfacing links you'd overlook. Treat it as a draft to verify against your own knowledge of the business, not a certified map; that verification is yours to do.

10. The second-opinion rerun

[In a fresh chat] Here are my options, in this order: [same options, reversed].
Same constraints as before: [constraints]. Rank them from scratch,
and note which calls you're most and least confident about.

Why it works: running the ranking cold in a fresh chat gives you an independent second pass to compare against the first. Where the two agree, you've got a robust read worth acting on; where they diverge, you've found the calls that hinge on judgment you need to make yourself.

Prompts for strategic foresight: scenario planning and stress-testing

11. The premortem

It is 18 months from now and this strategy failed badly: [paste plan].
Write the internal postmortem: what actually went wrong, in what sequence,
and which early signal we ignored. Every cause must be specific to my business;
if a sentence could appear in any company's postmortem, cut it.

Why it works: the specificity rule is the entire prompt. Without it you get "market conditions shifted and execution lagged," which is astrology.

12. What would have to be true

For this plan to hit [target], list everything that has to be true,
sorted by how much control we have over each item.
Flag the items where the honest base rate says it usually is not true.

Why it works: it converts optimism into an audit. The control-sorting matters because teams over-plan the controllable items and under-watch the rest.

13. The shock test

Apply this shock to my plan: [competitor cuts price 30% / the key hire quits /
funding lands 6 months late]. Sort everything in the plan into three buckets:
survives untouched, needs rework, dies. One sentence of reasoning per item.

Why it works: naming one concrete shock beats "what are the risks?" every time, because the model has to trace consequences through your actual plan instead of listing generic hazards.

Prompts for turning strategy into a plan

14. Decisions in, open questions out

Turn these decision notes into a quarterly plan: [notes].
Each milestone gets one named owner and one "done means" sentence.
Anything in my notes that was discussed but not actually decided goes in a
separate OPEN QUESTIONS list. Do not resolve open questions yourself.

Why it works: a model handed vague notes returns a plan that reads finished, and a finished-looking document ends conversations that should still be happening. The OPEN QUESTIONS rule keeps undecided things visibly undecided.

15. One leading metric per bet

For each initiative in this plan, give one leading indicator that would show
it is working by week six. Not a lagging outcome like revenue.
For each, name the number that should trigger a review.

Why it works: "we'll track revenue" is how initiatives run two quarters past their expiration. Week-six leading indicators force the plan to say what early success looks like, in numbers someone can dispute.

Where prompts end and commitment begins

Used with real inputs and the discipline above, these prompts make ChatGPT a strong partner for exploring a strategy: mapping the market, generating options, pressure-testing your assumptions, drafting a first version of the plan. That's real value, and for a lot of the strategy calendar it's all you need.

The line to hold is between exploring and committing. Exploring is reversible; you're widening the set of things you might do. Committing is when you point the company at one path, fund it, staff it, and stake the quarter on it. That decision rests on how your specific constraints and dependencies actually interact. It has to be tracked and updated as they change, not regenerated from whatever you happened to paste into a chat window that afternoon. That's the job a purpose-built tool is built for, and I've mapped the landscape in the guide to AI strategy tools. Use ChatGPT to explore widely and sharpen your thinking; when you're ready to bet the business on a direction, give that decision a tool built to hold it.

Frequently asked questions

What makes a good ChatGPT prompt for strategic planning?

A good strategy prompt assigns a real job with real inputs (not "act as a strategist"), constrains the output format so the model has to do the work (a ranked list with no ties, a contradiction-only report), and gives the model a way to flag what it doesn't know (an UNKNOWN or NONE escape hatch). A prompt with none of those returns fluent text that reads like strategy and moves nothing.

Do these prompts work with Claude and Gemini too?

Yes. They rely on general properties of how large language models respond to structure and constraint, not quirks of one product, so they transfer across ChatGPT, Claude, and Gemini. Model quality shifts month to month; the prompt engineering carries over.

Can ChatGPT prompts replace a strategic planning tool?

For the exploratory phase, they get you a long way: research, options, stress-testing, a first-draft plan. What they don't replace is a tool that holds a persistent model of your business and updates it as your constraints change, which is what the committing phase depends on. A prompt rebuilds context from scratch every session; a purpose-built tool carries it forward.

About the Author

Eric Levine is the founder of StratEngine AI. He spent five years at Meta in Strategy and Operations, where he led global business strategy initiatives across international markets. He holds an MBA from UCLA Anderson and is a CPA. He builds AI-powered strategic planning tools used by operators, consultants, and executives to compress the mechanical work of strategy and reporting while keeping the judgment human.