AI Prompts for Root Cause Analysis

Root cause analysis is undermined by two habits: stopping at the first plausible cause, and treating hypotheses as findings. A 5 Whys chain that is not tested against evidence at each step is a story; a fishbone diagram that is never pruned is a list. The discipline is to separate what you know from what you suspect, test the suspicions cheaply, and then fix the cause rather than the symptom.

These prompts enforce that discipline. They ask for the evidence behind each 'why', they generate hypotheses by category and then design the tests, and they turn the confirmed cause into corrective and preventive actions with verification. The model can structure the reasoning and challenge weak links; it cannot know what happened — that comes from your data and the people involved.

Before you use these

Have these ready to replace the highlighted [variables]:

• A clear problem statement: what, where, when, how much, how often
• Data and evidence: records, timelines, measurements
• People involved and what they observed
• Previous fixes attempted and their results

The prompts

• 1. Run a 5 Whys with evidence at each step
• 2. Generate and test fishbone hypotheses
• 3. Build the corrective and preventive action plan

1. Run a 5 Whys with evidence at each step

Act as a root cause facilitator running a 5 Whys analysis.

Problem: [what happened, where, when, magnitude, frequency — as a specific statement, not a category]
Evidence available: [records, measurements, timeline, photos, logs]
What people observed: [statements by role]
Previous fixes: [what was tried and what happened]

1. Restate the problem as a precise, measurable statement. If it is vague, say what is missing.
2. Build the causal chain: for each 'why', give the answer, the evidence that supports it (specific — a record, a measurement, an observation), and an evidence rating: confirmed / partial / assumed. Where more than one answer is plausible, branch and follow each.
3. Stop each branch when you reach a cause that, if fixed, would prevent recurrence — or when evidence runs out. Mark the latter clearly and state the test that would extend it.
4. Check the chain backward: does each cause actually produce the effect above it? Flag weak links.
5. Distinguish the root cause from contributing factors and from the trigger.
6. Summarize: the most likely root cause, its evidence rating, the branches not yet excluded, and the two tests that would settle it.

Do not accept 'human error' or 'lack of training' as a root cause — ask why the process allowed the error. Do not fill evidence gaps with plausible narrative.

2. Generate and test fishbone hypotheses

You are facilitating a fishbone (Ishikawa) analysis for [problem statement].

Process context: [steps, equipment, materials, people, methods, measurement systems, environment]
What we know: [facts and data]
What has changed recently: [any changes to materials, people, equipment, methods, volume]

1. Generate hypotheses under each category — Methods, Machines/Equipment, Materials, People, Measurement, Environment (adapt categories for service processes: Process, Systems, Inputs, People, Information, Policy). At least three per category, specific to this problem, not generic.
2. For each hypothesis: the mechanism by which it would cause the problem; current evidence for and against; likelihood (high/medium/low); and a specific, cheap test (data cut, measurement, controlled trial, interview) with the result that would confirm or exclude it.
3. Prune: remove hypotheses inconsistent with the known facts and say why.
4. Prioritize the remaining hypotheses by likelihood × ease of testing. Propose the test sequence for the first week.
5. Note interactions: causes that only produce the problem in combination.
6. Recent changes: highlight any hypothesis linked to a recent change — these are disproportionately likely.

Present the fishbone as a categorized list, then the test plan as a table. Keep hypotheses falsifiable.

3. Build the corrective and preventive action plan

Act as a quality and operations lead building a corrective and preventive action (CAPA) plan.

Confirmed root cause: [statement and evidence]
Contributing factors: [list]
Problem effect: [what customers/operations experienced]
Constraints: [budget, time, resources, systems]

1. Containment: immediate actions to stop the effect reaching customers or spreading while the cause is fixed. Owner, date, and the point at which containment is withdrawn.
2. Corrective actions: changes that remove the root cause. For each: the action, why it addresses the cause (not the symptom), owner, due date, cost, and the risk it introduces.
3. Preventive actions: changes that stop the same cause producing other problems elsewhere — process, standard, training, control, design. Include where else the cause could be present.
4. Verification: for each action, how effectiveness will be measured, the metric and target, the review date, and what happens if it has not worked.
5. Recurrence check: the monitoring that will detect the problem returning, and for how long.
6. Documentation: what is updated (SOP, control plan, FMEA, training records) and the closure criteria.

Distinguish clearly between actions that fix the cause and actions that only make detection faster. Do not close the plan on completion of actions — close it on verified effect.

Worked example

Related prompts

• Process Improvement
• Operational KPI Reviews
• Bottleneck Analysis
• SOP and Operating Routine Drafting
• Supply Chain Disruption Response
• Order Fulfillment Analysis

Logical next step

After this, most operations teams move on to Process Improvement.

All Operations & Supply Chain prompts · Search the full library