Tool

Forcing Deeper Reasoning & Structured Thought

A confident paragraph hides its logic. This makes AI show its work in a structure you can read and check, so you catch a bad step before you act on it.

1. What this does for you

By default AI gives you a smooth answer and buries the reasoning. For anything with moving parts, that is a trap: you cannot tell a good conclusion from a confident guess. This prompt forces AI to think step by step and answer in a structure, a numbered list or a table, with its assumptions and confidence on the table. You read the logic, not just the verdict, and you spot the weak link yourself.

2. What to type into your AI

Copy the box below. Replace the words in [BRACKETS] with your own info. Then paste it into Claude or ChatGPT.

Think about this problem step by step before you answer. Do not jump to a conclusion.

PROBLEM:
"""
[DESCRIBE THE DECISION OR QUESTION IN ONE OR TWO SENTENCES]
"""

WHAT I KNOW:
[PASTE THE FACTS, CONSTRAINTS, NUMBERS, OR OPTIONS]

Work in this order:
1. Restate the problem in your own words so I can confirm you understood it.
2. Lay out your reasoning as a numbered list or a table, one step per line.
3. Show any assumption you had to make, and rate your confidence in each as high, medium, or low.
4. Only then give your recommendation, in two sentences.

If a step depends on information I did not give you, stop and ask me instead of guessing.

Try it now in Claude →

See a full worked example ▾

Deciding whether to move a monthly report to a dashboard

Think about this problem step by step before you answer. Do not jump to a conclusion.

PROBLEM:
"""
Should we move our monthly report from a Word document to a live dashboard?
"""

WHAT I KNOW:
- The report goes to 9 people once a month; three only read the summary.
- It takes one staff member about 4 hours to assemble.
- We already pay for a tool that can publish a simple dashboard.
- Two readers are not comfortable with new software.

Work in this order:
1. Restate the problem in your own words.
2. Lay out your reasoning as a numbered list or table, one step per line.
3. Show each assumption and rate your confidence high / medium / low.
4. Only then give your recommendation, in two sentences.

If a step depends on information I did not give you, stop and ask.

Deciding whether to switch our K-2 reading program or keep the one we have

Think about this problem step by step before you answer. Do not jump to a conclusion.

PROBLEM:
"""
Our K-2 reading scores have been flat for two years. I need to decide whether to adopt a new phonics-based reading program next year or keep our current one and add coaching.
"""

WHAT I KNOW:
- New program costs 48,000 dollars plus 3 PD days for 14 teachers.
- Current program is paid off, but only 41 percent of 2nd graders hit the winter benchmark.
- Two nearby districts using the new program saw benchmark gains of about 12 points in year two, not year one.
- I have 60,000 dollars left in the literacy budget and teachers are already stretched.

Work in this order:
1. Restate the problem in your own words.
2. Lay out your reasoning as a numbered list or table, one step per line.
3. Show each assumption and rate your confidence high / medium / low.
4. Only then give your recommendation, in two sentences.

If a step depends on information I did not give you, stop and ask.

Deciding whether to run our snow-removal service in-house or contract it out

Think about this problem step by step before you answer. Do not jump to a conclusion.

PROBLEM:
"""
Our public works snow-removal costs keep rising. I need to decide whether to keep the service fully in-house or contract part of the routes out to a private vendor for next winter.
"""

WHAT I KNOW:
- In-house cost last winter was 1.9 million dollars for 220 lane-miles, including overtime and equipment.
- A vendor bid covers the 70 outer lane-miles for 480,000 dollars per season, fixed.
- Three of our seven plow trucks are past useful life and need 300,000 dollars in replacements within two years.
- We have a union agreement and a public expectation that main roads stay city-cleared.

Work in this order:
1. Restate the problem in your own words.
2. Lay out your reasoning as a numbered list or table, one step per line.
3. Show each assumption and rate your confidence high / medium / low.
4. Only then give your recommendation, in two sentences.

If a step depends on information I did not give you, stop and ask.

Deciding whether to self-perform the sitework or subcontract it on a 4-month build

Think about this problem step by step before you answer. Do not jump to a conclusion.

PROBLEM:
"""
As the GC on a 4-month commercial build, I need to decide whether to self-perform the sitework and grading or hand it to a subcontractor.
"""

WHAT I KNOW:
- Sub bid for full sitework scope is 210,000 dollars, fixed, with a 3-week duration.
- Self-perform would cost roughly 165,000 dollars in labor and rented equipment, but ties up my two best operators for 4 weeks.
- Those same two operators are needed on the concrete pours starting in week 5, and the Owner will not move that date.
- Weather delays this spring have already pushed our schedule 6 working days behind.

Work in this order:
1. Restate the problem in your own words.
2. Lay out your reasoning as a numbered list or table, one step per line.
3. Show each assumption and rate your confidence high / medium / low.
4. Only then give your recommendation, in two sentences.

If a step depends on information I did not give you, stop and ask.

Deciding whether to hire a full-time baker or keep contracting the overflow

Think about this problem step by step before you answer. Do not jump to a conclusion.

PROBLEM:
"""
My bakery is turning away large catering orders. I need to decide whether to hire a full-time baker or keep paying a contract baker for overflow.
"""

WHAT I KNOW:
- A full-time baker costs about 52,000 dollars a year plus payroll taxes and benefits.
- The contract baker charges 35 dollars an hour and I used her for about 400 hours last year (14,000 dollars).
- I turned away roughly 30,000 dollars in catering orders last year because I could not staff them.
- My slow season runs January through March, when a full-timer would have thin work.

Work in this order:
1. Restate the problem in your own words.
2. Lay out your reasoning as a numbered list or table, one step per line.
3. Show each assumption and rate your confidence high / medium / low.
4. Only then give your recommendation, in two sentences.

If a step depends on information I did not give you, stop and ask.

Deciding whether to launch a new after-school program or deepen the one we run

Think about this problem step by step before you answer. Do not jump to a conclusion.

PROBLEM:
"""
We have one 40,000 dollar unrestricted gift. I need to decide whether to launch a new job-readiness program or use the money to deepen our existing after-school program.
"""

WHAT I KNOW:
- The new program would need a part-time coordinator (28,000 dollars) and would serve about 25 new teens in year one.
- Deepening the current program would add tutoring and cut our waitlist of 60 kids roughly in half.
- Our lead funder has hinted they want to see "new reach," not just more of the same.
- We are already short one staff person and our program director is at capacity.

Work in this order:
1. Restate the problem in your own words.
2. Lay out your reasoning as a numbered list or table, one step per line.
3. Show each assumption and rate your confidence high / medium / low.
4. Only then give your recommendation, in two sentences.

If a step depends on information I did not give you, stop and ask.

Deciding whether to move a client from hourly billing to a fixed monthly fee

Think about this problem step by step before you answer. Do not jump to a conclusion.

PROBLEM:
"""
One of my bookkeeping clients keeps disputing hours. I need to decide whether to move them to a fixed monthly fee or keep billing hourly.
"""

WHAT I KNOW:
- Their work averaged 9 hours a month last year at 90 dollars an hour (about 810 dollars a month), but ranged from 5 to 16 hours.
- A fixed fee I am considering is 850 dollars a month.
- Year-end and 1099 season reliably spike their hours well above average.
- I want to keep the relationship, but I am eating unbilled time on their messy months.

Work in this order:
1. Restate the problem in your own words.
2. Lay out your reasoning as a numbered list or table, one step per line.
3. Show each assumption and rate your confidence high / medium / low.
4. Only then give your recommendation, in two sentences.

If a step depends on information I did not give you, stop and ask.

Deciding whether to take a low-paying but high-visibility mural commission

Think about this problem step by step before you answer. Do not jump to a conclusion.

PROBLEM:
"""
A well-known downtown venue offered me a mural commission that pays below my usual rate. I need to decide whether to take it or hold out for full-rate work.
"""

WHAT I KNOW:
- The offer is 3,500 dollars for a wall that would normally be a 6,000 dollar job for me.
- The wall faces a busy plaza and the venue tags artists on a page with about 40,000 followers.
- The mural would take about three weeks, which overlaps with two smaller paid jobs I would have to turn down (worth about 2,800 dollars together).
- I have been trying to break into larger public and civic commissions this year.

Work in this order:
1. Restate the problem in your own words.
2. Lay out your reasoning as a numbered list or table, one step per line.
3. Show each assumption and rate your confidence high / medium / low.
4. Only then give your recommendation, in two sentences.

If a step depends on information I did not give you, stop and ask.

3. What to do with the answer

Pick a decision with more than one moving part. Write it as a one-sentence problem.
Paste the prompt into Claude or Gemini. Fill in the problem and everything you already know.
Read the reasoning steps first, not the recommendation. That is where a mistake hides.
Check the assumptions it flagged as low confidence. Those are the things you need to confirm.
If a step is wrong, correct that one input and run it again. Do not argue with the conclusion; fix the premise.

Why this works

When you make AI lay its reasoning out as steps, two things happen. It actually reasons more carefully, because the structure forces it to. And you get a map of the logic you can audit, instead of a single confident sentence you have to take on faith. The judgment stays yours; the structure just makes it possible to use.

Section Recap

Core Principle

Make the reasoning visible and you can check it. A structured, step-by-step answer turns a black-box verdict into something you can audit.

Best-Fit Example

A program manager is weighing two vendors with cost, timeline, and staff-comfort tradeoffs. Instead of asking "which is better," she asks the AI to reason step by step in a table with confidence ratings. The table shows the recommendation hinges on an assumption about training time she knows is wrong. She corrects that one input, reruns, and gets the opposite, and correct, answer.

A reading coach or principal is staring at a 48,000 dollar adoption that will not pay off until year two, with tired teachers and one budget line. Instead of going with a gut feeling in a staff meeting, they paste the real numbers in and force the AI to lay out the tradeoffs and flag which assumptions are shaky (like the "12 point gain" that came from a different district). The AI shows its work as a numbered chain, so the principal can see exactly where the logic bends and where it is guessing. She walks into the board meeting with a recommendation she can defend line by line, and she stays the editor because she is the one who knows her teachers' actual capacity.

A public works director has to defend a snow-removal decision to a council and to residents who notice every unplowed street, so the reasoning has to be neutral and visible. She drops in the real per-mile costs, the vendor bid, the aging trucks, and the union constraint, and forces the AI to separate hard facts from assumptions rather than pushing one answer. The numbered chain surfaces the parts that need a labor-relations check before anything is decided, which keeps the memo honest. She uses the output as the skeleton of a council briefing memo, but she stays the editor because only she knows what the union and the public will actually accept.

A GC is tempted by the 45,000 dollar savings of self-performing, but the real risk is his two key operators getting stuck on grading when the Owner-mandated concrete pours start. He forces the AI to reason through the schedule collision instead of just comparing the two dollar figures, and to rate how confident it is that sitework really finishes in the planned window. The numbered breakdown makes the hidden cost (a slipped pour date and a hard Owner deadline) impossible to ignore. He brings a defensible call to the next Owner update, and he stays the editor because he knows his crew and the site better than any model does.

An owner-operator can feel the lost catering revenue but is scared of a 52,000 dollar salary that goes quiet every winter. Instead of guessing, she hands the AI the real numbers (the salary, the contract rate, the turned-away orders, the slow season) and makes it reason through whether the missed revenue actually covers a full-timer year-round. The confidence ratings flag that her "30,000 dollars turned away" is an estimate, which is exactly the number she needs to nail down before she commits. She gets a clear, defensible next step, and she stays the editor because she knows her regulars and her cash flow in a way the model never will.

A program lead has one flexible gift and two honest pulls: a funder who wants new reach and a waitlist of 60 kids she already cannot serve. She forces the AI to reason through the tradeoff and to be clear about which claims are soft, like the funder's hint and her director's real capacity, rather than getting a confident but hollow answer. The numbered logic gives her language she can bring straight into a board memo and a funder conversation without pretending the gaps are not there. She makes the final call herself, staying the editor, because she is the one accountable to both the kids and the funder.

A bookkeeper is tired of arguing over hours but worries a flat fee will bury her during 1099 season and year-end. She feeds the AI the real hour range, the proposed fee, and the seasonal spikes, and makes it reason through whether 850 dollars actually covers the messy months instead of just the average one. The confidence ratings help her see where she is assuming versus where she has hard data, which is the difference between a fee that holds and one she regrets. She treats the output as general guidance, not filed advice, and she stays the editor because she is the one who knows this client's habits and her own capacity.

An artist is being asked to work below rate for "exposure," and the real cost is not just the discount but the two paid jobs she would have to turn down. She hands the AI the numbers and her actual goal (breaking into larger public commissions) and forces it to reason through whether this wall plausibly leads there, rating how much of that is hope versus evidence. Seeing the visibility upside marked "low confidence" next to the very real 2,800 dollars in lost work helps her decide with clear eyes instead of flattery. She makes the call herself, staying the editor, because only she knows what her time and her direction are worth.

Key Takeaways

A one-paragraph answer hides the logic; a structured answer exposes it.
Ask for a numbered list or a table, plus assumptions and confidence levels.
Read the steps before the recommendation; that is where errors live.
Fix the wrong premise and rerun, rather than debating the conclusion.

← All Tools