Service business customer support is dominated by two questions: 'can I move my appointment?' and 'do I owe a fee if I cancel?' Operators commonly estimate these at around a third of the inbox, and none involve a Shopify order. A written playbook lets AI answer the routine ones and route the rest to a person.
Key takeaways
- Reschedule and cancellation emails often make up roughly one-third of a service inbox.
- A playbook needs four parts: notice window, fee rule, rebooking link, deposit handling.
- AI should resolve in-policy requests; humans take disputes, late cancellations, and emergencies.
- Heylio drafts replies but does not replace your booking calendar.
What share of service-business support is rescheduling and cancellation?
For most appointment-based businesses, rescheduling and cancellation requests make up about a third of inbound messages, based on the common experience of clinic, studio, and salon operators. This is an estimate, not a universal benchmark. Count your own: tag 100 recent messages and see how many are appointment changes.
The share matters because these messages are repetitive. The same four facts decide almost every answer: when the appointment is, what the notice window is, whether a deposit was paid, and where the customer rebooks.
A quick audit takes 20 minutes:
- Export the last 100 customer messages from Gmail, Outlook, Instagram DMs, or WhatsApp.
- Label each as reschedule, cancel, fee question, no-show dispute, group change, or other.
- Mark which ones were answered by copying a policy you already had.
The last label is your automation opportunity. If 25 of 100 messages were policy copy-paste, that is the pool an AI can safely take.
What should a reschedule playbook contain?
A reschedule playbook needs four elements: a notice window (for example 24 or 48 hours), a fee rule, a rebooking link, and deposit handling. With those four written down, a reply can be drafted without anyone guessing. Missing any one forces improvisation, which is where AI and humans both go wrong.
- Notice window: state it in hours, such as 24 hours before the start time.
- Fee rule: state the amount or percentage and when it applies.
- Rebooking link: the URL of your existing scheduling tool.
- Deposit handling: whether it transfers, is refunded, or is forfeited, by case.
Add the exceptions list too. Illness, emergencies, and your own cancellations should be named so the AI knows they exist and escalates them.
Which requests should AI resolve alone, and which go to a human?
Let AI resolve requests that fit policy exactly: reschedules inside the notice window, standard cancellations, and deposit questions with a clear rule. Send anything involving judgment to a person: late cancellations, disputed no-show fees, and medical or emergency cases. A useful rule is that AI applies policy and humans make exceptions.
In Heylio, a handover to a human is free. You pay only for conversations the AI closes by itself, so escalating a risky case costs nothing. That removes the temptation to let the AI push through a dispute to save money.
Escalate immediately when a message contains:
- Words like 'hospital', 'emergency', 'funeral', or 'doctor'.
- A challenge to a fee already charged.
- A request for an exception to the written policy.
- An angry tone plus a refund demand.
How do you write a cancellation-fee policy an AI can apply without improvising?
Write the policy as numbered conditions with numbers, not adjectives. 'Cancel at least 24 hours before and pay nothing; under 24 hours, pay 50% of the service price' can be applied by anyone. 'Reasonable notice is appreciated' cannot. Ambiguity is what makes an AI invent rules.
Citation capsule: A no-show fee policy template that an AI can follow has five parts. First, the notice window in hours. Second, the fee as a fixed amount or percentage. Third, the deposit rule: whether it is credited to a new booking or forfeited. Fourth, the exceptions list, naming illness and emergencies as human-review cases. Fifth, the escalation trigger, stating that any dispute goes to a named person. For example: 'Cancel or move 24 hours or more before your start time at no charge. Within 24 hours, 50% of the service price is due. Deposits transfer once to a new date if you move inside the window. Illness and emergencies are reviewed by our team.' Every clause is testable, so the AI quotes it rather than interpreting it. Put this text in your Heylio knowledge base and in the cancellation playbook so replies match it word for word.
How do the five common request types map to action, owner, and reply?
Five request types cover most appointment-change email. Three are usually safe for AI, one is conditional, and one belongs with a human. The table below shows the action, who handles it, and a sample reply for an appointment appearing in the customer's own scheduling tool.
| Request type | Action | AI or human | Sample reply |
|---|---|---|---|
| Reschedule within window | Confirm and send rebooking link | AI | 'Happy to move that. You are inside our 24-hour window, so there is no charge. Pick a new time here: [your booking link].' |
| Reschedule outside window | Apply fee rule, offer link | AI if fee is fixed; human if customer objects | 'Moving within 24 hours carries a 50% fee under our policy. You can rebook here: [your booking link]. Reply if you want our team to review.' |
| Cancel with deposit | State deposit rule, confirm cancellation | AI | 'Your booking is cancelled. Per our policy, the deposit is [credited/forfeited] because you cancelled [inside/outside] the window.' |
| No-show dispute | Pause, gather details, escalate | Human | 'Thanks for explaining. I have passed this to our team, who will reply personally today.' |
| Group booking change | Check size and date rules, escalate if over limit | Human, or AI for simple date moves | 'We can move your group inside the window. Please confirm the headcount and use this link: [your booking link].' |
Replace the bracketed items with your real link and wording in the playbook, so no placeholder reaches a customer.
Where does Heylio fit, and where does it not?
Heylio drafts and closes replies from Gmail, Outlook, Instagram DMs, and WhatsApp, using your knowledge base and playbooks. It does not replace your booking calendar. It cannot move the appointment itself; the reply links the customer to your existing scheduling tool, where they pick the new slot.
That boundary is deliberate. Your calendar stays the source of truth for availability. Heylio handles the conversation around it.
- Does: answer policy questions, quote fees, send the rebooking link, close the thread.
- Hands over, free: cases outside policy, such as disputes or emergencies.
- Learns: it tracks corrections your team makes and suggests knowledge base and playbook updates.
- Does not: book, move, or block calendar slots.
One honest note: Heylio's Shopify tools, like order lookup and refunds, matter little for pure service businesses. The standalone mode, with its own sign-in and billing, suits operators without a Shopify store.
How do you set this up in a week?
A small service business can go live in about five working days if the policy is already written. Spend day one on the audit, day two on the policy text, and the rest on testing against real past messages. Start narrow: automate only reschedules inside the window first.
- Day 1: audit 100 messages and tag them.
- Day 2: write the fee policy with numbers and exceptions.
- Day 3: load the policy and rebooking link into the knowledge base and cancellation playbook.
- Day 4: connect your inbox and test with 20 past messages.
- Day 5: go live on in-window reschedules; review every reply for a week.
When your team edits a draft, that correction becomes a suggested playbook update. Accept the ones that reflect real policy.
What is the next step for your inbox?
Start with your own numbers. Tag 100 recent messages, write the fee policy with a 24-hour window and a fixed fee, and add your rebooking link. Then test the routine cases on the Free plan, which includes 25 AI resolutions. Keep disputes and emergencies with a person from day one.