Customer Experience
Create a Service-Recovery Plan
Convert verified incident evidence and approved resolution options into an owned, reviewable service-recovery plan without authorizing customer or system action.
Quick facts
- Best for
- Owners · Founder-operators
- Prompt type
- Quality
- Expected result
- A reviewable service-recovery action plan with evidence, four recovery lanes, named ownership, stop conditions, Unknowns, and human approvals.
- Time saved
- 30-45 minutes per recovery plan
- Required inputs
- INCIDENT FACTS · CUSTOMER IMPACT · APPLICABLE POLICY · SEVERITY · CURRENT OWNERSHIP · APPROVED RESOLUTION OPTIONS · RECOVERY CONSTRAINTS
- Works with
- ChatGPT + Claude
What this prompt does
- Builds an internal recovery plan from verified impact, supplied policy, ownership, and approved options.
- Separates immediate containment, customer-response drafting, corrective work, and evidence gaps for named-human review.
When to use
- After an issue has been triaged and named owners need a consistent draft plan for recovery and follow-through.
When not to use
- Do not use it to issue a refund or credit, admit fault, make a retention offer, contact a customer, close a ticket, or change a live system.
- Do not place unauthorized support data or unnecessary personal, sensitive, or confidential data in an AI system.
The prompt
#CONTEXT:
Create a draft service-recovery plan for internal human review. Use only the supplied evidence, approved policy, severity, owner authority, resolution options, and constraints. Do not execute the plan.
#INPUTS:
- Verified incident facts, timeline, and evidence references: [INCIDENT FACTS]
- Customer-reported and verified operational impact: [CUSTOMER IMPACT]
- Current applicable policy with source and effective date: [APPLICABLE POLICY]
- Human-reviewed severity level and supporting reason: [SEVERITY]
- Current owners, responsibilities, and authority: [CURRENT OWNERSHIP]
- Resolution options explicitly approved for consideration: [APPROVED RESOLUTION OPTIONS]
- Capacity, legal, financial, technical, and timing constraints: [RECOVERY CONSTRAINTS]
#INSTRUCTIONS:
1. Build a source-of-truth table separating verified facts, reported statements, conflicts, expired inputs, and Unknown items.
2. Do not invent, infer, or fabricate policy, eligibility, customer facts, outcomes, root cause, dates, capacity, approvals, fault, or liability.
3. Draft a service-recovery action plan in four lanes: containment, fact verification, approved customer-response preparation, and corrective follow-through.
4. For every proposed step, state the evidence basis, named owner, dependency, target proposed for owner approval, stop condition, and verification evidence. Do not convert a proposal into a promise.
5. Include only supplied approved resolution options. Recommend escalation when an option, owner, authority, timing, or eligibility fact is missing or conflicting.
6. Do not issue, apply, process, or promise a refund, reimbursement, credit, compensation, retention offer, discount, concession, exception, or guarantee.
7. Draft customer-response talking points only when supported by verified facts and supplied policy. Do not admit fault or liability, and do not characterize legal responsibility.
8. Do not send, publish, contact, notify, route, assign, close, resolve, or update any live ticket, account, record, or system.
9. Use only authorized support data and authorized customer data in an approved AI workspace or approved AI vendor. Minimize the data and redact or omit unnecessary personal, sensitive, or confidential data. Follow company retention and vendor policy.
10. A named human owner must approve every refund, credit, or compensation decision before action. A named human owner and legal counsel must approve every admission of fault or liability before action. A named human owner must approve every retention offer, discount, or policy exception before action. A named human owner must approve every external communication, sending action, or customer contact before action. A named human owner must approve every ticket, system, account, or record change or mutation before action.
#RESPONSE FORMAT:
## Evidence, conflicts, and Unknowns
| Item | Status | Supplied source | Conflict or missing evidence |
|---|---|---|---|
## Draft service-recovery action plan
| Lane | Proposed step | Evidence basis | Named owner | Dependency | Proposed target | Stop condition | Verification |
|---|---|---|---|---|---|---|---|
## Customer-response talking points for review
- Verified acknowledgment
- Supported next step
- Prohibited or unresolved statement
## Human decisions and approvals
- Decision — named owner — evidence required — approval stateInput checklist
- INCIDENT FACTS
- CUSTOMER IMPACT
- APPLICABLE POLICY
- SEVERITY
- CURRENT OWNERSHIP
- APPROVED RESOLUTION OPTIONS
- RECOVERY CONSTRAINTS
Example input
Fictional example — INCIDENT FACTS: Northstar's onboarding host link failed at 10:02 ET; the team created a replacement link at 10:18 ET. CUSTOMER IMPACT: Juniper Facilities reports six attendees waited; exact attendance is Unknown. APPLICABLE POLICY: Delivery incident policy v3 permits a manager follow-up and rescheduling review. SEVERITY: Operations owner assigned Level 2 after review. CURRENT OWNERSHIP: Omar owns recovery; Maya owns response drafting; Finance owns any credit decision. APPROVED RESOLUTION OPTIONS: Manager follow-up and priority rescheduling review; no refund, credit, or compensation approved. RECOVERY CONSTRAINTS: Trainer availability and customer schedule Unknown; no legal conclusion supplied.Expected output structure
- A reviewable service-recovery action plan with evidence, four recovery lanes, named ownership, stop conditions, Unknowns, and human approvals.
Customize this prompt
- Add your internal incident and corrective-action identifiers so evidence can be traced without copying excess customer data.
- Separate the internal operational target from any customer-facing commitment until the right owner approves it.
Guardrails
- Do not invent, infer, or fabricate policy, eligibility, customer facts, outcomes, root cause, dates, capacity, approval, fault, or liability; unsupported items remain Unknown and conflicts remain unresolved.
- A named human owner must approve every refund, credit, or compensation decision before action.
- A named human owner and legal counsel must approve every admission of fault or liability before action.
- A named human owner must approve every retention offer, discount, or policy exception before action.
- A named human owner must approve every external communication, sending action, or customer contact before action.
- A named human owner must approve every ticket, system, account, or record change or mutation before action.
- A named human owner must approve every payment or financial decision before action.
- A named human owner must approve every deletion, closure, suspension, termination, revocation, deactivation, or archive action before action.
- A named human owner must approve every promise or commitment before action.
- Use only authorized support data and authorized customer data in an approved AI workspace or approved AI vendor; minimize data, redact or omit unnecessary personal, sensitive, or confidential data, and follow company retention and vendor policy.
- The assistant must produce a draft plan only and must not perform customer, financial, legal, ticket, or system actions.
NEXT STEP