The one idea
Creativity at work is not a mood. It is a set of prompts you run on a problem until it produces ideas, inspired or not.
Do not try to think of a good idea. Run a prompt that forces a different kind of idea, and do it nine more times.
Your first three ideas are defaults — what you have seen before, arriving fast. The useful material starts around idea seven, when the obvious ones are used up and you have to look at the problem itself.
You get more ideas by using a checklist of ways to change the problem, not by trying harder to be creative.
Separate generating from judging. Produce a wide set first, evaluate second, and never do both in the same pass.
Idea generation fails through premature convergence — evaluating each option as it appears kills the unusual ones early, because unusual options look weak before they are developed.
The nine mechanisms
Each is a question you ask the problem. Take "onboarding takes too long" through them and watch the shape change.
First principles
Strip the problem to what must be true and rebuild without inherited steps. What is onboarding for? If the answer is "so the person can do useful work safely and know who to ask", most of a typical onboarding week serves neither.
Inversion
Ask how to cause the problem deliberately. How would I make onboarding as slow as possible? Scatter information across five systems, require access approvals nobody has authority to give, schedule introductions in week three. Now look at your real process — you have written a list of what to fix, and inversion finds things direct thinking misses.
Subtraction
Remove something essential. What if there were no onboarding documents at all? Maybe a new joiner shadows someone for two days, or gets a small real task on day one and is told who to ask. Subtraction generates ideas that reduce work rather than adding to it, which makes it the most undervalued mechanism here.
Combination
Fuse two existing things. Onboarding plus support: the new joiner answers three real tickets in week one, which teaches the product faster than any document.
Constraint
Impose an impossible limit. Onboarding must take one hour. You cannot do everything, so you must identify what matters. Constraints generate ideas precisely by removing the option of doing it the normal way.
Analogy
How does another field solve this? How does a restaurant train a new server on a Friday night, or a hospital hand over a patient between shifts? Handover checklists from other fields usually beat the ones grown inside a company.
Perspective change
Run the problem as someone else. How would the new joiner design this? The person who quit in month two? A customer dealing with them in week one? Each seat sees a different problem.
SCAMPER
A checklist for when you are empty: Substitute, Combine, Adapt, Modify, Put to another use, Eliminate, Reverse. Deliberately mechanical — run down it and something catches.
Forced association
Take a random object and connect it. Onboarding and a supermarket. Aisles, signage, a map at the entrance, staff who tell you where things are. That gives you a one-page map of who owns what. Feels silly, works reliably, and is what to use when the others run dry.
The exercise
Do this on a real problem, in this order, without skipping to evaluation.
Twenty options in forty minutes
1 of 6Write five normal ideas. The obvious ones. Get them out of your head and onto paper so they stop occupying the space. Do not judge them.
Evaluating what you produced
Four criteria. Score each option low, medium or high — a points system only manufactures false precision.
| Criterion | The question |
|---|---|
| Originality | Is this different from what is already being done? |
| Usefulness | If it worked, would it solve the actual problem? |
| Feasibility | Can it be done with what we have, by people who exist? |
| Clarity | Could someone else execute this from the description? |
Originality alone is a party trick; feasibility alone is why nothing changes. Options worth presenting score at least medium on usefulness and clarity — an idea nobody can act on is a comment, not an option.
Problem: the weekly report takes six hours to produce and nobody reads it.
Normal: a template, a shorter version, automated charts.
Unusual (inversion — how do we make it maximally unread?): send it Friday evening, eleven pages, decision buried in paragraph nine. Which tells you: Monday morning, one page, decision at the top.
Zero budget (subtraction): stop sending it for a month and see who asks. A real option, frequently the correct one — the people who ask tell you what the report is for.
AI: draft the summary paragraph from the raw numbers, with the analyst editing rather than writing from scratch.
Combined: one page, Monday, drafted from the data and edited by a human, sent only to the people who asked when it stopped.
Failure mode: if it exists for compliance, or for one senior person who never comments, "stop sending it" fails badly. Check who mandated it first.
Problem: a project to submit and no idea what to do it on.
Constraint: it must be finishable in a weekend. Subtraction: what if you used no new data and only re-analysed something public? Perspective: what would your professor hate reading for the fortieth time, and what adjacent question does nobody pick?
The constraint pass saves you here. Most stalled projects are not short of ideas; they are full of ideas too big to finish, and more ideation does not fix that.
Try this
Problem: too many meetings on your team.
Generate three ideas using inversion only — how would you deliberately make meeting load worse?
Your challenge
Level 3 · IndependentTake a real problem you face this month and run the full exercise: five normal, five unusual, five zero-budget, five using AI. Combine the best three. Write the failure mode for each survivor.
Then produce the deliverable that matters: a half-page memo with your recommendation, the two alternatives you rejected, and one line each on why.
Success criteria: twenty distinct options on paper; the recommendation is not one of your first five; each rejection has a specific reason, not "not suitable".
What people usually get wrong
- Judging while generating. The fastest route to three safe ideas. Generate first, evaluate second.
- Stopping at five. Ideas six to fifteen are where the value is.
- Presenting a fake option set. One real proposal plus two obviously bad ones is a well-known move; managers recognise it immediately.
- Confusing an idea with a plan. "Use AI for onboarding" is not an option until you can say what it does, to what input, checked by whom.
- Generating alone when the problem spans teams. You will miss constraints you do not know exist. Generate alone first, then bring it to people who know the other parts.
- Falling in love with the unusual one. Novelty is one criterion of four.
How someone experienced does it
Experienced people generate options against the problem behind the request. "Onboarding takes too long" is a symptom; the real problem may be that nobody owns onboarding, or that hiring is lumpy so it is always someone's fourth priority. Ask what would have to be true for the problem to disappear permanently, and generate against that.
They also keep the discarded options. "We considered X and rejected it because of Y" is worth more than the recommendation itself, because in eight months somebody will propose X again and you will have the reasoning rather than a memory of it.
And they know their own default. Everyone has a shape of solution they reach for — process people add a step, technical people build a tool, managers add a meeting. Knowing yours lets you deliberately generate three options that are not it, which is where most of your new ideas come from.
When not to use this
Do not run this when the answer is known and the work is execution. Twenty options for "which meeting room should we book" is a way of avoiding starting.
Do not run it in an emergency. When something is on fire, run the known procedure. Option generation is for the review afterwards, where inversion — how would we make this happen again? — is unusually productive.
And not when the constraint is absolute. If the decision is already made and unchangeable, generating alternatives produces frustration rather than options. Ask whether it is open before spending the afternoon.
Prove it
Produce one option set on a real problem: twenty options, four criteria scored, one recommendation with its failure mode, two rejected alternatives with reasons.
Keep it. When the problem is revisited — and it will be — you will hold the only document saying what was already considered.
Keep learning this
Paste this into any AI assistant. It turns the assistant into a tutor that tests you instead of just answering you.
Act as an experienced practitioner who is good at teaching. I have just learned running inversion and constraint prompts on a problem I am stuck on. Assume I am intelligent but relatively new to this — treat me as intermediate level. Work through this in order, and wait for my reply at each step: 1. Ask me 5 questions that test whether I actually understood running inversion and constraint prompts on a problem I am stuck on. Do not reveal the answers yet. 2. After I answer, tell me which parts I got right, which I got wrong, and which I only half-understand. Explain only what I misunderstood — do not re-teach what I already know. 3. Give me one practical challenge based on something I could genuinely encounter at work or in daily life. Do not solve it for me. 4. Evaluate my solution the way an experienced person would judge it, including what a professional would have done differently. 5. Tell me what to learn next, and why that comes next. 6. Give me trustworthy sources for deeper study — prefer official documentation, primary research or standards bodies over blogs and videos. Rules for you: no buzzwords. No motivational filler. Say "I'm not certain" when you are not certain, and tell me which parts of your answer I should verify myself. Clearly separate facts from your recommendations and your opinions.
Become independent at this
Use this when you want a path from where you are to actually good, with checkpoints you can test yourself against.
I want to become independently capable at generating options on demand — not permanently dependent on AI, tutorials or step-by-step guides. Design a progression for me with five stages: Beginner, Guided practice, Independent practice, Real-world application, Professional level. For each stage tell me: - what I must know - what I must be able to do without help - the mistakes people make at this stage - one practical challenge - one real project that would prove I reached this stage - one way I can test myself honestly Then tell me the signals that I am ready to move to the next stage, and the signals that I have skipped ahead too early. Keep the theory to the minimum I actually need. Focus on ability I can transfer to situations you and I have not discussed.