You get:
- linear steps that don’t handle exceptions
- no guidance for decision points
- employees making inconsistent decisions
- escalations to managers for simple decisions
- errors from wrong conditional handling
But a decision tree is not a linear process.
It is a map of possible paths.
- Decision points: where a choice is made
- Conditions: what determines the path
- Actions: what to do based on condition
- Escalation: when to involve a manager
- Default path: what to do if no condition matches
Without decision trees, employees guess.
This framework forces AI to build decision trees that guide conditional processes.
Assume the role of a process designer who creates decision trees for conditional processes. Your task is to create a decision tree SOP. Generate: 1. PROCESS NAME AND GOAL 2. STARTING POINT - What triggers the decision process 3. DECISION TREE (text-based or bulleted) - IF [condition] THEN [action] - ELSE IF [condition] THEN [action] - ELSE [default action] 4. DECISION POINTS MAPPED - Key questions to ask - Possible answers for each 5. ESCALATION PATHS - When to escalate to manager - Who to escalate to 6. DEFAULT PATH - What to do if no condition matches 7. DECISION TREE VISUALIZATION (text-based) - Indented tree structure INPUTS: Process Name: [INSERT] Decision Triggers (what situations require decisions): [LIST] Possible Outcomes (paths): [LIST] Known Exceptions or Edge Cases: [LIST] Authority Levels (who can make which decisions): [DESCRIBE] RULES: - Decision points must be clear (yes/no or multiple choice) - Conditions must be objective (not "if customer seems upset") - Actions must be specific (not "handle appropriately") - Escalation paths: when and to whom - Default path: catch-all for unhandled conditions - Test decision tree with real scenarios before publishing
- Decision trees work best for troubleshooting and customer support.
- Conditions must be objective and verifiable.
- Escalation paths prevent employees from making out-of-authority decisions.
- Default path catches edge cases.
- Test decision tree with real scenarios before publishing.
Process Name: Customer Refund Request Handling
Decision Triggers: Customer requests refund, product defective, wrong item shipped, customer changed mind, outside return window
Possible Outcomes: Approve full refund, Approve partial refund, Offer store credit, Deny refund, Escalate to manager
Known Exceptions: VIP customer (high lifetime value), product damage caused by customer, expired warranty
Authority Levels: Support agent ($0-50), Team lead ($51-200), Manager ($200+ or exceptions)
This framework improves outcomes by forcing:
- decision point identification (clarity)
- condition-action mapping (logic)
- escalation paths (authority boundaries)
- default path (edge case handling)
- visual representation (comprehension)
Great decision trees don’t guess — they guide, condition by condition.
Build Better AI Systems
Subscribe for advanced prompt engineering, AI business strategy tools, SOP frameworks, and practical strategies for leaders and operators.