What this guide does
- Moves the user from a rough idea to a decision-ready output through five connected conversations.
- Uses AI to improve the quality of the thinking, not simply generate a polished document.
- Creates a visible trail of assumptions, evidence, options, risks, ownership, and next steps.
- Works for new ideas, existing processes, improvement opportunities, business cases, and operating plans.
- Keeps accountability with the user: AI organizes, challenges, and drafts; a person verifies and decides.
Use the five prompts in the same AI conversation so each stage builds on the prior output.
The five-stage flow
C
Clarify
Define the real problem, desired outcome, affected stakeholders, and constraints.
R
Research
Gather evidence, examples, practices, benchmarks, and information needed to decide.
A
Assess
Compare options, challenge assumptions, test the idea, and identify gaps.
F
Form
Turn the strongest approach into a structured proposal, process, or plan.
T
Translate
Convert the draft into actions, ownership, decisions, measures, and a conclusion.
What to do at every stage
- Answer the questions with specific facts, examples, and constraints.
- Correct the information when there is an issue or unsupported assumption.
- Provide internal documents, data, prior attempts, or policies when appropriate.
- Confirm the output before moving to the next stage. A weak foundation will weaken everything downstream.
- Save the final output and revise it as better information becomes available.
C
Clarify the Problem
Stage 1
Purpose: Establish a shared, accurate definition of the issue before anyone starts designing a solution.
What is expected
- Enter the business issue, idea, opportunity, or current process in plain language.
- Answer follow-up questions with facts, examples, affected groups, constraints, and prior attempts.
- Say “unknown” where information is not available. Do not fill gaps with guesses.
- Review and correct the final Problem Definition Brief before moving to Research.
What AI should return
- A problem statement that distinguishes the core issue from its symptoms.
- A description of the current state, desired state, stakeholders, impact, and constraints.
- A clear list of confirmed facts, assumptions, unknowns, and information still needed.
- No solution recommendation yet.
✓ Stage checkpoint
Proceed only after you have reviewed the stage output, corrected misunderstandings, and confirmed the information that will carry forward.
Keep unresolved items visible. The next stage should not silently convert an assumption into a fact.
You are my business problem-framing partner. Help me define the issue before proposing a solution. [ENTER THE PROBLEM, IDEA, OPPORTUNITY, OR CURRENT PROCESS HERE] BEFORE YOU BEGIN • Restate what you believe the issue is in one or two sentences. • Separate the underlying problem from symptoms, frustrations, and any solution I may already be assuming. • Ask up to seven focused questions, one at a time. Adapt each new question to my prior answers. • Prioritize questions about the current state, affected stakeholders, business impact, desired outcome, constraints, prior attempts, and available evidence. • If I do not know an answer, record it as an unknown rather than filling the gap. • Do not recommend a solution during this stage. After I answer the questions, create a "Problem Definition Brief" with these sections: 1. Core problem statement 2. Current state and observable symptoms 3. Likely root causes or cause hypotheses 4. Stakeholders and groups affected 5. Business impact and why the issue matters 6. Desired future state and definition of success 7. Constraints, non-negotiables, and decision boundaries 8. Confirmed facts, assumptions, unknowns, and information needed End by asking me to confirm or correct the brief before we move to research. Ask me any clarifying questions before completing the analysis.
R
Research the Issue
Stage 2
Purpose: Build an evidence base so the solution is informed by facts, proven practices, and relevant examples rather than instinct alone.
What is expected
- Confirm the Problem Definition Brief from the Clarify stage.
- Provide internal reports, policies, metrics, customer feedback, prior proposals, or process documentation when appropriate.
- Specify the industry, geography, regulatory environment, or organizational context that affects the issue.
- Tell the AI whether it may use web research, only the materials you provide, or both.
What AI should return
- A concise research brief tied directly to the defined problem.
- Relevant facts, benchmarks, examples, practices, risks, and failure patterns.
- A clear distinction between sourced evidence, internal facts, general knowledge, and assumptions.
- A prioritized list of evidence still needed before a decision is made.
✓ Stage checkpoint
Proceed only after you have reviewed the stage output, corrected misunderstandings, and confirmed the information that will carry forward.
Keep unresolved items visible. The next stage should not silently convert an assumption into a fact.
Use the confirmed Problem Definition Brief from the previous stage as the research scope. Act as a practical business researcher. Build an evidence base that will help us evaluate the issue and possible responses. Research or analyze the following: • Relevant business practices, operating models, frameworks, or standards • Comparable examples from similar organizations, functions, or situations • Available benchmarks or performance ranges that are truly comparable • Common failure modes, unintended consequences, and adoption barriers • Legal, regulatory, security, privacy, financial, workforce, customer, or operational considerations that may apply • Internal data or documents that would materially improve the analysis • Questions an experienced executive, operator, finance leader, risk leader, or customer would ask IMPORTANT RULES • Use reliable, current sources when research tools are available and cite them. • If you cannot browse or verify a claim, label it clearly as general knowledge, a hypothesis, or an item to validate. • Do not invent benchmarks, costs, policies, examples, or citations. • Explain whether each source or example is directly comparable, partially comparable, or only illustrative. • Keep the research focused on the problem; do not recommend a final solution yet. OUTPUT FORMAT 1. Research summary: five to ten most relevant findings 2. Evidence table: finding, source or basis, relevance, and confidence 3. Comparable examples and what can reasonably be learned from them 4. Risks, constraints, and known failure patterns 5. Internal information still needed 6. Unanswered questions that could materially change the decision End by asking me to add any internal evidence or correct any findings before we assess options. Ask me any clarifying questions before completing the analysis.
A
Assess the Thinking
Stage 3
Purpose: Challenge the idea before commitment by comparing realistic options, surfacing tradeoffs, and exposing weak assumptions or missing information.
What is expected
- State your priorities, non-negotiables, risk tolerance, budget or capacity limits, and decision deadline.
- Confirm or adjust the proposed evaluation criteria before the AI scores options.
- Identify the people or groups whose support is necessary for approval and adoption.
- Challenge the AI’s recommendation when it does not fit the operating reality.
What AI should return
- Three or four realistic options, including improvement of the current state when appropriate.
- A transparent comparison using criteria such as value, feasibility, cost, risk, timing, adoption, and strategic alignment.
- A multi-perspective debate, pre-mortem, dependency analysis, and gap review.
- A conditional recommendation with confidence level and information that could change it.
✓ Stage checkpoint
Proceed only after you have reviewed the stage output, corrected misunderstandings, and confirmed the information that will carry forward.
Keep unresolved items visible. The next stage should not silently convert an assumption into a fact.
Using the confirmed Problem Definition Brief and Research Brief, help me assess the available choices before we build a final solution. First, propose a decision framework with evaluation criteria appropriate to this issue. Include, when relevant: business value, customer or employee impact, feasibility, cost and capacity, time to benefit, operational complexity, risk and compliance, adoption difficulty, reversibility, and strategic alignment. Ask me to confirm or adjust the criteria and their relative importance before continuing. After I confirm the criteria: • Develop three or four realistic options. Include a conservative option, a balanced option, and a more ambitious option. Include improving or maintaining the current approach when that is a legitimate choice. • For each option, describe the approach, expected benefits, limitations, dependencies, resources, risks, adoption needs, and conditions under which it would be appropriate. • Score or rank the options against the agreed criteria. Explain the evidence and rationale behind each rating without presenting assumptions as facts. • Debate each option from the perspectives of: [enter the roles you want to review against -- be specific: CFO, CIO, VP, Associates, business sponsor, operational user, finance or procurement, technology/security/risk, customer or end user, and a constructive skeptic.] • Conduct a pre-mortem: assume the option failed after implementation and identify the most likely reasons. • Identify missing stakeholders, decisions, data, ownership, costs, process exceptions, controls, and change-management needs. • Separate critical gaps from important improvements and optional refinements. OUTPUT FORMAT 1. Agreed decision criteria 2. Options summary 3. Comparison matrix 4. Stakeholder debate and objections 5. Pre-mortem and key failure scenarios 6. Gaps and unanswered questions, ranked by importance 7. Recommended direction, confidence level, and what could change the recommendation End by asking me to select or modify the preferred direction before we form the solution.
F
Form the Solution
Stage 4
Purpose: Convert the selected direction into a complete, usable draft that fits the type of work being created: a proposal, process, or plan.
What is expected
- Choose the preferred option or describe how multiple options should be combined.
- State whether the primary deliverable is a proposal, process, plan, or another format.
- Provide required templates, approval standards, governance requirements, or organizational terminology.
- Review every assumption, placeholder, owner, timeline, cost, and measure before accepting the draft.
What AI should return
- A structured draft tailored to the chosen deliverable type.
- Clear scope, ownership, activities, decisions, risks, measures, and governance.
- Explicit placeholders where facts, dates, costs, or owners are not yet known.
- A draft that can be refined for an executive audience, working team, or approval process.
✓ Stage checkpoint
Proceed only after you have reviewed the stage output, corrected misunderstandings, and confirmed the information that will carry forward.
Keep unresolved items visible. The next stage should not silently convert an assumption into a fact.
Use the selected direction from the Assess stage to create a complete first draft. My primary deliverable is a: [PROPOSAL / PROCESS / PLAN / OTHER] Audience: [ENTER THE PRIMARY AUDIENCE] Required template, standards, or terminology: [PROVIDE TEMPLATES OR WRITE "NONE PROVIDED"] Start by confirming the intended deliverable, audience, and level of detail. Ask only the additional questions needed to prevent major gaps. Build the core document with these sections when relevant: 1. Executive summary 2. Problem statement and business context 3. Objective and desired outcomes 4. Recommended solution or future-state design 5. Scope, exclusions, assumptions, and constraints 6. Stakeholders, roles, accountability, and decision rights 7. Major activities, phases, milestones, and dependencies 8. Resource, capability, technology, vendor, and funding considerations 9. Risks, controls, mitigations, and escalation points 10. Change-management, communication, training, and adoption needs 11. Measures of success, baseline needs, leading indicators, and outcome measures 12. Governance, review cadence, and continuous-improvement approach 13. Decisions required and immediate next steps TAILOR THE DRAFT TO THE DELIVERABLE TYPE • For a proposal: emphasize the business case, alternatives considered, benefits, investment needs, risks, approval request, and recommendation. • For a process: include trigger, inputs, roles, step-by-step flow, decision points, exceptions, controls, outputs, service levels, and review cadence. • For a plan: include workstreams, tasks, sequence, dependencies, owners, milestones, resources, risks, communications, and measures. IMPORTANT RULES • Do not invent facts, costs, dates, owners, policy requirements, or benefits. • Use clear placeholders such as [OWNER TO CONFIRM] or [DATA REQUIRED] where information is missing. • Label assumptions and show where validation is required. • Keep the language direct, business-oriented, and appropriate for the intended audience. End with a "Draft Completeness Review" listing unresolved items that must be addressed before approval or execution.
T
Translate into Action
Stage 5
Purpose: Turn the draft into a practical path forward and a concise conclusion that leaders and teams can use to decide, approve, and act.
What is expected
- Confirm the desired decision, implementation horizon, governance cadence, and likely owners.
- Resolve or assign the remaining placeholders from the Form stage.
- Confirm how success will be measured and where baseline data will come from.
- Choose the final audience: executive approval, working team execution, stakeholder communication, or all three.
What AI should return
- A phased implementation roadmap and action register.
- Clear decisions, owners, dependencies, milestones, measures, and review cadence.
- An executive-ready conclusion that explains the recommendation, rationale, risks, and immediate actions.
- A final validation check showing what is verified, assumed, or still unresolved.
✓ Stage checkpoint
Proceed only after you have reviewed the stage output, corrected misunderstandings, and confirmed the information that will carry forward.
Keep unresolved items visible. The next stage should not silently convert an assumption into a fact.
Translate the completed draft into an action-ready package. Intended use: [EXECUTIVE DECISION / TEAM EXECUTION / STAKEHOLDER COMMUNICATION / ALL THREE] Target decision or start date: [ENTER DATE OR "TO BE DETERMINED"] Implementation horizon: [ENTER HORIZON OR "RECOMMEND ONE"] Create the following: 1. A phased implementation roadmap appropriate to the issue. Use 30/60/90 days only when that structure fits; otherwise use logical phases. 2. An action register with action, owner, due date or timing, dependency, expected output, and status. 3. A decision register showing the decision required, decision owner, information needed, deadline, and consequence of delay. 4. A risk and issue summary with trigger, mitigation, contingency, and escalation owner. 5. A measurement plan with baseline, leading indicators, outcome measures, data owner, and review frequency. 6. A stakeholder and communication plan showing who needs to know what, when, and through which channel. 7. The first five actions that should happen next, in sequence. FINAL OUTPUT CONCLUSION Create an executive-ready conclusion that includes: • One-paragraph summary of the problem and recommended direction • The decision or approval being requested • Three strongest reasons to proceed • Three most important risks or conditions • Expected business or operational impact, without inventing metrics • What will happen in the first phase • Who is accountable for the next step FINAL QUALITY CHECK • Distinguish verified facts, assumptions, estimates, and unresolved items. • Flag unsupported claims or statements that require a source. • Confirm that no costs, dates, outcomes, policies, or owners were invented. • Identify any contradiction between the problem, research, assessment, solution, and implementation plan. • Provide a final list titled "Before This Is Shared" with the remaining validations or approvals. Present the result in two forms: A. A concise one-page executive summary or in the template format previously provided. B. A detailed working plan for the delivery team.
What you should have at the end
A complete CRAFT conversation should produce more than a polished document. It should create a traceable business package that shows how the problem was defined, what evidence informed the work, which options were considered, why the recommendation was selected, and how it will be put into action.
Final output package
- Problem Definition Brief: A confirmed statement of the core problem, desired outcome, stakeholders, constraints, assumptions, and unknowns.
- Research Brief: Relevant evidence, comparable examples, risks, benchmarks, and information still needed.
- Options and Assessment: A transparent comparison of realistic choices, stakeholder perspectives, tradeoffs, and failure scenarios.
- Structured Proposal, Process, or Plan: A detailed draft with scope, activities, ownership, risks, measures, governance, and decision needs.
- Action-Ready Conclusion: An executive summary, decision request, implementation roadmap, action register, measures, and first steps.
The completed package should answer five questions
| C | Are we solving the right problem? |
| R | What evidence and context should shape the decision? |
| A | Which option is strongest, and what could make it fail? |
| F | What exactly are we proposing, changing, or delivering? |
| T | Who will do what next, and how will we know it worked? |
How to leverage and validate the output
| Audience | Use these outputs | Primary purpose |
|---|---|---|
| Executive or approval body | One-page summary, decision request, options comparison, key risks | Decide, approve, fund, or redirect |
| Delivery team | Detailed plan, action register, roles, milestones, dependencies, measures | Execute and manage the work |
| Stakeholders and users | Problem summary, future-state explanation, communication and adoption plan | Understand, contribute, and adopt |
| Risk, legal, security, finance, or procurement | Assumptions, evidence, risk register, controls, costs, vendor and policy needs | Review and validate |
| Process owner | Process flow, decision points, exceptions, controls, metrics, review cadence | Operate and continuously improve |
Before sharing the result
- Confirm that the problem statement reflects the actual issue, not only the loudest symptom.
- Verify sources, facts, benchmarks, costs, legal requirements, and policy statements.
- Replace every placeholder and assign a real owner to each critical action and decision.
- Ensure the recommendation is feasible within available capacity, funding, timing, and change tolerance.
- Have the accountable reviewer approve the final content and remain responsible for the outcome.
◆ CRAFT principle
Do not just ask AI to write the document. Use AI to improve the thinking behind it — then apply human judgment before the work moves forward.