Recommend the next best action for a customer

Role: Act as a CX manager designing next-best-action guidance for agents. Context: Customer signals available (product usage, tenure, tier, recent tickets, sentiment, lifecycle stage, entitlements): [SIGNALS]. Candidate actions (proactive tips, feature nudges, plan changes, retention offers, check-ins, escalations): [ACTIONS]. Business goals (retention, expansion, deflection, satisfaction) and their priority: [GOALS]. Guardrails (offer limits, do-not-contact, fairness, consent): [GUARDRAILS]. Task: Produce: (1) the signal set and how each is read; (2) for common customer situations, the recommended next best action and the reasoning, balanced across the goals; (3) the decision logic as clear rules or a decision table; (4) the guardrails as hard constraints the recommendation must never breach; (5) how an agent should present the action as a suggestion, not a script; (6) what to log to learn whether it worked. Format: Sections: Signals | Situations -> recommended action & why | Decision table | Guardrails (hard rules) | Agent framing | Measurement. Verification: Check no recommendation can breach a guardrail; flag situations where signals conflict and the logic is ambiguous. Validation: Test on a churn-risk customer and a happy power user: are the actions genuinely helpful to the customer, not just to our metrics? Adjust anything that feels manipulative.