Help me practice structured problem solving on this real situation: [PROBLEM]. The outcome I need is [DESIRED OUTCOME], the constraints are [CONSTRAINTS], and the facts or observations I have are [KNOWN FACTS]. My role, authority, and deadline are [ROLE AND DEADLINE]. Act as a rigorous coach, not as someone who simply hands me an answer. First, restate the problem as a clear problem statement that separates symptoms from the underlying decision. Then guide me through this sequence: 1. Define success: identify 2-4 measurable or observable criteria, including a minimum acceptable outcome and any non-negotiable constraint. 2. Diagnose: create a cause map with plausible hypotheses grouped into people, process, information, incentives, tools, and external factors. Label each hypothesis as supported, uncertain, or contradicted by the facts. 3. Prioritize investigation: recommend the 3 highest-value questions or checks, explaining what each would change and the cheapest reliable way to answer it. 4. Generate options: offer 3 materially different actions, including a low-risk reversible step. For each, list benefits, downsides, resources, dependencies, and likely failure mode. 5. Decide and act: give a decision table against the success criteria, a first 48-hour plan, and a trigger for revisiting the choice. 6. Teach: end with two reasoning habits I should practice in similar situations and one short exercise that asks me to make the next judgment before seeing your example answer. Do not invent facts or assume I control people, budget, or timing that I have not stated. Treat safety, legal, medical, financial, and employment issues as areas where qualified advice may be needed. Ask up to 3 clarifying questions only if a required input is missing. Before answering, check for root-cause claims supported only by correlation and for recommendations that violate a stated constraint. Flag both explicitly.
Fill in
| Placeholder | What to enter | Example |
|---|---|---|
| [PROBLEM] | Describe the situation, what is happening, and why it matters. | Our five-person support team misses its first-response target every Monday, and customer complaints rise by Tuesday |
| [DESIRED OUTCOME] | State the result you are trying to achieve. | Return to a median first response under two hours without adding headcount this quarter |
| [CONSTRAINTS] | List limits on time, money, authority, risk, resources, or rules. | No overtime mandate, no new hires, and the team lead can change schedules but not product priorities |
| [KNOWN FACTS] | List observed facts, relevant data, and what remains uncertain. | Monday ticket volume is 35% higher; billing tickets take longest; two agents are unavailable until noon; the backlog clears by Wednesday |
| [ROLE AND DEADLINE] | State your role, what decisions you can make, and the time available. | I am the support lead and need a plan to trial next Monday |
How to use
- Replace the placeholders with observed facts and constraints; mark guesses as guesses rather than facts.
- Work through the investigation questions before accepting an action plan, especially if the cause is uncertain.
- Choose a reversible test where possible and write down the metric that will determine whether it worked.
- Follow up with: "Here are the results after the trial. Compare them with the success criteria, update the cause map, and recommend the next smallest test."
Variations
Root cause drill
Use when you need to diagnose a recurring operational issue.
Coach me through a root-cause analysis of [RECURRING ISSUE]. The timeline and evidence are [EVIDENCE], affected groups are [AFFECTED GROUPS], and constraints are [CONSTRAINTS]. Build a fault tree that separates proximate causes from contributing conditions. Ask me 5 sequential diagnostic questions, one at a time, and wait for my answer after each. Then summarize the most supported cause paths, disconfirming evidence to seek, and one containment action. Do not treat a repeated association as proof of cause. Ask up to 3 setup questions only if needed. Check that the containment action does not conceal the evidence.
Decision practice
Use when you must choose between viable options under uncertainty.
I need to decide between [OPTION A], [OPTION B], and [OPTION C] for [DECISION]. My priorities are [PRIORITIES], constraints are [CONSTRAINTS], and uncertainty is [UNKNOWN FACTS]. Create a weighted decision framework, but first challenge whether my criteria are independent and complete. Score only what the evidence supports; use ranges or unknown rather than invented numbers. Explain which fact would most change the choice, and give a reversible next move. Ask up to 3 clarifying questions only if necessary. Check for a criterion counted twice under different names.
Incident retrospective
Use after an error or project miss to improve the system rather than assign blame.
Help me run a blameless retrospective for [INCIDENT OR MISS]. The timeline is [TIMELINE], impact was [IMPACT], and existing safeguards are [SAFEGUARDS]. Produce a factual event timeline, contributing-factor analysis, five discussion questions, and 3 corrective actions with owners by role and verification signals. Separate facts, assumptions, and unresolved questions. Do not assign intent or blame without evidence. Ask up to 3 clarifying questions only if required inputs are missing. Check that each action addresses a contributing condition rather than merely reminding people to be careful.
Tips
- State the decision separately from the symptom; “conversion is down” may require a diagnosis, while “choose a retention experiment” is a decision.
- Seek evidence that could disprove your leading explanation before investing in a fix.
- Prefer a time-boxed, reversible experiment when the option is uncertain and the downside is containable.
- Write decision criteria before comparing options so the most vivid anecdote does not become the deciding factor.
FAQ
Can AI teach problem-solving skills rather than solve the problem for me?
Yes. Ask it to reveal assumptions, propose disconfirming checks, and make you choose before it reveals a recommendation.
What if I do not have enough data?
Separate what you know from what you need to learn, then select the cheapest check that could change the decision. Do not fill gaps with confident assumptions.
Should I always find one root cause?
No. Many practical problems result from several conditions interacting. Focus on the cause paths that are both supported and actionable.