Refund Policy Assistant

Evaluate refund eligibility against policy rules with explainable decisions, partial refund calculations, and compliance flags.

refundsupporte-commercebillingpolicy

Refund Policy Assistant

Refund decisions are one of the highest-stakes interactions in customer support. Get them wrong in the customer’s favor and you erode margin; get them wrong against the customer and you risk chargebacks, negative reviews, and in some jurisdictions, regulatory action. The challenge is that most refund policies have enough edge cases — partial use, promotional pricing, subscription timing, regional consumer protection laws — that applying them consistently at scale is genuinely hard.

This skill reads your refund policy, evaluates the ticket context against it, and returns a structured eligibility decision with the reasoning shown. Agents see not just “approved” or “denied” but which policy clause applies and why, so they can explain the decision to the customer and override it when circumstances warrant.

TL;DR

Provide your refund policy document and a ticket (or batch of tickets). The skill evaluates each case against the policy rules, calculates partial refund amounts where applicable, flags edge cases that require human judgment, and outputs a decision with a plain-language explanation the agent can use directly in their reply. Requires read access to tickets; does not process refunds itself.

What it does

Refund evaluation involves more than checking a date. This skill handles the full assessment:

  • Policy rule parsing and application — Reads your refund policy document (plain text, PDF, or structured JSON) and extracts the decision rules: eligibility windows, conditions for full vs. partial refunds, non-refundable categories, and exception clauses. Applies these rules to each ticket’s specific facts.
  • Partial refund calculation — For subscription services, calculates pro-rated refund amounts based on unused days in the billing period. For e-commerce, handles partial returns (some items kept, some returned) and accounts for restocking fees, original shipping costs, and promotional discounts that may affect the refundable amount.
  • Regulatory compliance flagging — Identifies tickets from customers in jurisdictions with statutory consumer protection rights that may exceed your standard policy (EU 14-day cooling-off period, Australian Consumer Law guarantees, UK Consumer Rights Act). Flags these for human review rather than applying your internal policy alone.
  • Chargeback risk assessment — Evaluates whether a denied refund is likely to result in a chargeback based on the customer’s stated reason and account history. Flags high-risk denials so agents can consider whether a goodwill refund is more cost-effective than a chargeback dispute.
  • Decision audit trail — Every decision includes a structured log: which policy clause was applied, what ticket facts were used, what the calculated amount is, and what confidence level the skill assigns to the decision. This log is stored with the ticket for compliance and dispute purposes.
  • Edge case escalation — Automatically routes tickets to senior agents when the situation doesn’t fit cleanly within policy rules: customers citing medical emergencies, cases involving deceased account holders, situations where the product was demonstrably defective, or requests that reference legal action.

Best for

  • E-commerce support teams processing high volumes of return and refund requests where consistency matters — customers compare notes, and inconsistent decisions create trust problems.
  • Subscription services where refund calculations involve pro-rating, trial period rules, and annual vs. monthly billing differences that are easy to get wrong manually.
  • Billing dispute teams at SaaS companies dealing with chargebacks, where having a documented decision trail is essential for winning disputes with payment processors.

This skill carries a High risk rating because refund decisions have direct financial consequences and potential legal implications. It is not appropriate to use this skill in a fully automated mode (decision → refund processed without human review) until you have validated its accuracy against your policy across at least several hundred real cases.

How to use (example)

Scenario: A customer contacts support for a SaaS project management tool. They purchased an annual plan 45 days ago, used the product for the first two weeks, then stopped. They’re now requesting a full refund, citing that the product “didn’t meet their needs.”

Input

Ticket: "I signed up for the annual Pro plan on February 1st. I tried it for a couple of weeks but it's just not the right fit for our team. I'd like a full refund of the $348 I paid."
Customer: Marcus Webb, account ID #71034
  - Plan: Annual Pro ($348/year, billed Feb 1)
  - Usage: Last login Feb 14 (31 days ago), 3 projects created, 2 team members invited
  - Previous refunds: None
  - Location: United Kingdom
Policy document: [attached: refund-policy-v4.md]
  - Standard refund window: 14 days from purchase
  - Pro-rated refunds: Not offered after 14-day window
  - Exceptions: Documented technical failures, billing errors
  - Goodwill refunds: Manager discretion, up to $50
Output: Eligibility decision with reasoning, recommended response, escalation flag if needed

What the skill produces

REFUND ELIGIBILITY ASSESSMENT — Ticket #71034

Decision: INELIGIBLE under standard policy — ESCALATE for regulatory review

Confidence: High (policy application) | Escalation reason: UK Consumer Rights Act

---
POLICY ANALYSIS

Standard eligibility: ✗ Not eligible
  - Purchase date: February 1st
  - Standard refund window: 14 days (expired February 15th)
  - Request received: March 17th (45 days after purchase)
  - Applicable exception: None — no documented technical failure or billing error

Goodwill refund: Possible (manager discretion, up to $50)
  - Customer has no previous refund history
  - Usage was limited (last login 31 days ago)
  - Goodwill refund would not cover full request ($348)

---
REGULATORY FLAG: UK Consumer Rights Act 2015

Customer is located in the United Kingdom. Under the Consumer Rights Act 2015, digital services may be subject to a right to a partial refund if the service was not provided with reasonable care and skill, or if the service did not match its description.

This is distinct from your standard 14-day policy. The customer's stated reason ("didn't meet our needs") does not automatically trigger statutory rights, but this ticket should be reviewed by a team member familiar with UK consumer law before a final denial is issued.

Recommended action: Escalate to senior agent or legal/compliance team for UK regulatory review before responding.

---
IF ESCALATION IS NOT POSSIBLE (agent must respond now):

Suggested response approach:
- Acknowledge the request and the customer's experience
- Explain the standard 14-day refund window has passed
- Offer the goodwill refund option (up to $50, manager approval needed)
- Do NOT issue a final denial — note that the case is under review

DO NOT use the standard refund denial macro for this ticket.

---
PARTIAL REFUND CALCULATION (for reference if approved):
  - Annual plan: $348 / 365 days = $0.953/day
  - Days used (Feb 1 – Feb 14): 14 days × $0.953 = $13.34
  - Unused days: 351 days × $0.953 = $334.66
  - Pro-rated refund if approved: $334.66
  (Note: Standard policy does not offer pro-rated refunds; this calculation is for manager/legal reference only)

Variations

  • Batch mode: Process a CSV of open refund tickets. The skill returns a decision for each, sorted by escalation priority, so your team can work through the queue systematically.
  • Policy comparison mode: Provide two versions of your refund policy (current and proposed) and a sample of recent tickets. The skill shows how decisions would differ under each policy, useful for evaluating policy changes before rolling them out.
  • Chargeback defense mode: For tickets where a refund was denied and a chargeback has been filed, the skill generates a structured chargeback response document citing the relevant policy clauses and transaction evidence.

Permissions & Risks

Required permissions: Tickets (read access to ticket content, customer account data, purchase history)
Risk level: High

The high risk rating reflects three distinct concerns:

Financial accuracy: Partial refund calculations must be correct. An error in pro-rating logic — especially for annual subscriptions with mid-cycle upgrades or downgrades — can result in systematic over- or under-refunding. Validate the calculation logic against your billing system’s records before relying on it at scale.

Regulatory compliance: Consumer protection laws vary significantly by jurisdiction and change over time. The skill’s regulatory flags are based on general knowledge of major consumer protection frameworks (EU, UK, Australia, US state laws) but are not a substitute for legal advice. Any ticket flagged for regulatory review should go to a human with appropriate expertise.

Policy document currency: The skill applies the policy you provide. If your refund policy document is six months out of date, the skill will make decisions based on old rules. Establish a process to update the policy document whenever your terms change, and version-control it so you can audit which policy version was applied to historical decisions.

Recommended guardrails:

  • Never configure the skill to process refunds automatically — always require agent confirmation before any money moves.
  • Log every decision with the policy version used, ticket ID, and decision timestamp.
  • Review a random sample of 10% of decisions weekly during the first month of deployment.
  • Maintain a clear escalation path for regulatory flags — don’t let them sit in a queue.

Troubleshooting

  1. Skill approves refunds that should be denied (or vice versa)
    The most common cause is ambiguous language in your policy document. Phrases like “at our discretion” or “in exceptional circumstances” give the skill too much latitude. Rewrite those clauses with specific criteria: “at manager discretion for accounts with no previous refund requests and less than 30 days of active usage.”

  2. Partial refund calculations don’t match your billing system
    Check how your billing system handles the billing period. If you bill on the anniversary date rather than calendar month, or if your system counts days differently (inclusive vs. exclusive of the purchase date), specify this in the policy document: billing_period_calculation: anniversary_date, days_counted: exclusive_of_purchase_date.

  3. Too many tickets being escalated — team can’t keep up
    Review the escalation criteria. If the skill is escalating every ticket from a particular country, your policy document may not address that jurisdiction’s consumer rights framework. Add a section to your policy document specifying how you handle requests from those regions, even if the answer is “standard policy applies unless the customer cites a specific statutory right.”

  4. Regulatory flags are appearing for customers in jurisdictions where they shouldn’t apply
    The skill infers customer location from the ticket or account data. If location data is missing or incorrect, flags may be misapplied. Ensure your ticket system captures customer country at account creation and passes it to the skill.

  5. Skill misinterprets promotional pricing in refund calculations
    Promotional discounts (first-month-free, annual discount vs. monthly rate) require explicit handling. Add a promotional_pricing_rules section to your policy document specifying how discounts affect refundable amounts: “Annual plan discounts are non-refundable; refund is calculated based on the discounted annual rate, not the equivalent monthly rate.”

  6. Decision audit trail is missing from some tickets
    This usually means the skill ran in a context where the output wasn’t saved back to the ticket system. Ensure your integration writes the full decision log (not just the recommendation) to a ticket note or custom field before the agent sends their reply.

Alternatives

  • Chargebee refund automation — Subscription billing platform with built-in refund workflows. Handles the financial processing side (actually moving money) that this skill doesn’t. Best used alongside this skill: the skill evaluates eligibility, Chargebee processes the approved refund.
  • Stripe refund workflows — Stripe’s native refund API and dashboard tools. Excellent for processing refunds programmatically but doesn’t evaluate policy eligibility — it just executes what you tell it to. Pair with this skill for the decision layer.
  • Manual policy lookup — An agent reads the policy document and makes the decision themselves. Slowest option, highest variability between agents, but maximum human judgment for complex cases. Appropriate for very low-volume billing teams or for the edge cases this skill escalates.

Skills:

  • Support Macros — Draft the customer-facing reply once the refund decision is made
  • Email Triage — Classify and prioritize incoming billing tickets before refund evaluation
  • KB Builder — Maintain the policy documentation that feeds refund eligibility decisions

Guides:

Related Skills

Explore more AI skills in our directory.

Back to Skills