AI Appointment Booking Setup: Configuring Services, Durations and Buffers
The scheduling logic behind good phone bookings. Get your service catalog, durations, buffers and lead times right, and the double-bookings you dread simply stop happening.

Most double-bookings aren't caused by a bad receptionist, human or otherwise. They're caused by rules nobody wrote down. The assistant offered 3:00, the previous appointment was really going to run until 3:15, and there was no buffer to catch the gap. This is the unglamorous, decisive part of AI appointment booking setup: teaching the system the same scheduling common sense you carry in your head, so it stops offering slots that don't actually exist.
When people picture an AI phone assistant taking bookings, they picture the conversation — the greeting, the friendly voice, the read-back of the date. That part matters, and it's covered elsewhere. But a warm conversation that produces a broken calendar is worse than a gruff human who never double-books. The difference between the two isn't the script. It's the rules underneath: what each service is, how long it really takes, what has to happen before and after it, and when it simply can't be booked at all.
This guide walks through configuring those rules the way an owner actually thinks about them. We'll build a service catalog, set durations you'll regret less, add buffers that protect your day, limit how far ahead and how last-minute people can book, and block off slots that should never fill. Do this once, carefully, and the assistant stops needing you to clean up after it.
Why rules matter more than the script
A booking is a promise about time. When someone calls and books a haircut for Thursday at two, you're promising that a chair, a stylist and roughly forty-five minutes will exist then. The assistant can only keep that promise if it knows those three things. If it thinks a haircut takes twenty minutes, it will cheerfully stack three of them into an hour that only fits two, and you'll discover the problem when a real person is standing in your doorway.
This is why so much of AI appointment booking setup happens away from the phone. The voice is the last five percent. The other ninety-five is a small, boring data model: a list of services, each with a duration, a buffer, and a set of conditions about when it's allowed. Get that model right and almost any script sounds competent. Get it wrong and no amount of charm saves the calendar.

Start with an honest service catalog
Everything begins with the list of things a caller can book. It sounds trivial, and it's where most setups quietly go wrong, because the list in your head is fuzzier than you think. "A cleaning" might mean a routine hygiene visit or a deep scaling that takes twice as long. "A consultation" might be ten minutes on the phone or an hour in the room. The assistant can't guess. It books what's in the catalog, exactly as written.
So the first job is to write down every service a caller might reasonably ask for, in the words they use, not your internal codes. Callers don't ask for a "D1110 prophylaxis"; they ask to "come in for a cleaning." Your catalog should map the plain-language request to a concrete, bookable service with a real duration behind it.
- 1List every bookable serviceWalk through a typical week and note everything callers actually book. Include the awkward ones you'd rather forget — the walk-in tyre change, the emergency squeeze-in, the 'quick question' that becomes a full appointment.
- 2Name them the way callers doUse the caller's language. If people say 'balayage' and 'root touch-up', those are separate services with different lengths — don't hide them both under 'colour'.
- 3Split anything with two durationsIf the same name covers a 20-minute version and a 60-minute version, it's two services. One name, one duration. Otherwise the assistant averages them and is wrong both ways.
- 4Mark what needs a humanSome requests shouldn't auto-book at all — a first-time complex case, a large group, anything you'd want to vet. Flag these so the assistant takes a message instead of committing time.
That last step matters more than it looks. A good assistant knows the difference between what it can confidently book and what it should hand back to you. Vunoon doesn't have to force every call into a slot — for the flagged services, it can take the details and have you call back, which is far better than confidently booking something you'd have priced or scoped differently.
Set durations you'll actually regret less
Ask any owner how long a service takes and they'll give you the best-case number — the time it takes when nothing goes wrong, the client is on time, and you're not also answering the door. That number is a fantasy for scheduling purposes. If you build your calendar on best-case durations, every good day looks fine and every ordinary day runs late by lunchtime.
The duration you configure isn't how long the work takes. It's how long the slot should be blocked. Those are different numbers. A twenty-minute treatment might need a thirty-minute slot once you count the client settling in, the quick note you write afterwards, and the reality that a fifth of your clients arrive five minutes late.
The honest way to set durations is to look backwards, not forwards. Pull a few weeks of your real appointments and notice where you actually ran over. If "initial consult" is booked for thirty minutes but reliably takes forty, the schedule already knows the truth even if you don't. Configure the number the day keeps proving, not the one you wish were true.
| Service | Hands-on time | Realistic slot to configure | Why the gap |
|---|---|---|---|
| Routine cleaning | 30 min | 40 min | Seating, notes, occasional running late |
| New-client consult | 30 min | 45 min | First visits always take longer than repeats |
| Colour + cut | 90 min | 120 min | Processing time, wash, unpredictable length |
| Quick oil change | 20 min | 30 min | Paperwork, key handover, walk-out chat |
Notice that none of these gaps are dramatic. Ten or fifteen minutes each. But an assistant working from the hands-on column will pack a day so tightly that a single slow appointment topples the rest like dominoes. An assistant working from the realistic column simply offers fewer slots — and every slot it offers is one you can actually keep.
“The duration you configure isn't how long the work takes. It's how long the slot should be blocked — and those are almost never the same number.”
Buffers: the gaps that save your day
A buffer is dead time you deliberately reserve around an appointment — before it, after it, or both. It feels like waste when you set it up. It is the single most protective rule you can configure. Buffers are what absorb the slow client, the room that needs turning over, the phone call that runs long, the fact that you are a human who occasionally needs to breathe between jobs.
There are two kinds worth separating. A cleanup buffer is time you need after a service — sterilising instruments, resetting a room, writing up notes, wiping down a chair. A transition buffer is time you need before the next one — travel between sites for a home-service visit, prepping equipment, reviewing the file. Some services need one, some need both, and configuring them per service beats one blanket rule for everything.
- Cleanup buffer — reserved after the appointment. Room turnover, sterilisation, notes. The client is gone but the slot isn't free yet.
- Transition buffer — reserved before the next appointment. Travel time for mobile services, prep, setup. Especially vital when your appointments happen at different addresses.
- Per-service buffers — a deep clean needs more turnover than a quick check; a home visit needs travel time a salon chair doesn't. Set them individually, not as one global number.
- Peak-hour buffers — some owners widen buffers during their busiest block, when running late costs the most, and tighten them in quiet stretches.
When you configure a buffer, the assistant treats it as occupied time. If a service runs 2:00 to 2:45 with a fifteen-minute cleanup buffer, the next bookable slot doesn't start until 3:00 — and the caller never even hears about 2:45. They never feel the buffer. They just experience an assistant that never books you into an impossible corner.

Lead time: not too soon, not too late
Lead-time rules control the two edges of your booking window: how far in advance someone can book, and how last-minute they're allowed to be. Both edges cause real problems when they're left wide open.
The minimum lead time is the near edge. Without it, a caller at 8:52 can book an 8:55 slot — for which you have no chance to prepare, and which may already be spoken for by someone walking through the door. A minimum lead time of, say, two hours means the assistant won't offer anything sooner than that. For some businesses it's fifteen minutes; for a mobile service that has to drive to you, it might be a full day. It depends entirely on how much runway you need.
The maximum lead time is the far edge — how far into the future bookings are allowed. This one surprises owners, because letting people book months ahead sounds generous. In practice, appointments booked ninety days out have the worst no-show rates, and a calendar that fills too far ahead boxes you out of adjusting hours, taking holiday, or repricing. Many small businesses cap it at four to eight weeks and simply invite far-future callers to check back nearer the time.
- 1Decide your minimum runwayHow much notice do you genuinely need to be ready for a booking? Prep, travel, setup. That's your minimum lead time. Be honest — 'zero' almost always causes a scramble.
- 2Decide how far ahead is usefulPick a horizon where a booking is still likely to happen and you're still willing to be committed. Four to eight weeks suits most. Beyond it, the assistant offers a callback instead.
- 3Set different edges per service if neededAn emergency slot might allow near-zero lead time; a big project might require several days. The rules don't have to be uniform across your catalogue.
Blackout slots: when the answer is simply no
Every business has times that look bookable on paper but absolutely aren't. The hour you keep for admin. The Friday you're closing early for a family thing. The lunch break you keep skipping and then resenting. The morning after a trade show when you'll be useless. These are blackout slots — windows the assistant must treat as unavailable no matter how empty the calendar looks.
There's a difference between two flavours of these. Recurring blackouts repeat every week — a standing lunch break, no bookings before ten on Mondays, admin time on Wednesday afternoons. You set them once and forget them. One-off blackouts are specific dates — a holiday, a day out, a training course, a supplier visit. You add these as they come up, and the assistant stops offering those windows the moment you do.
- Recurring blackouts — lunch breaks, admin blocks, 'no new-client work on Fridays', early-morning holds. Set once; they protect every week.
- One-off blackouts — holidays, sick days, off-site jobs, a day you'll be at a conference. Added per date, honoured immediately.
- Soft blackouts — times you'd rather not fill but would for the right thing. Some owners leave these bookable but flag them so a human decides.
- Capacity blackouts — you have the hours but not the staff. Block the slots that can't actually be delivered, not just the ones that are physically closed.
The point of blackout slots is that a caller never learns they existed. They ask about Friday afternoon, the assistant simply offers the next real opening, and your family dinner survives. This is the quiet luxury of getting rules right: the assistant defends your time without you having to be there to say no.
“The assistant defends your time without you having to be there to say no. That's what a well-configured blackout actually buys you.”
How the rules combine on a real call
None of these rules works alone. What makes the calendar trustworthy is how they stack. Imagine a two-chair salon. A caller asks for a colour and cut at four on Thursday, and behind that simple request the assistant is quietly running every rule at once.
It checks the service catalog and finds colour-and-cut is a real, bookable service. It reads the duration — a two-hour slot, not the ninety minutes of hands-on work. It adds the cleanup buffer, so the slot really runs until 6:15. It checks lead time: Thursday is inside the eight-week horizon and well past the two-hour minimum, so both edges pass. It checks blackout slots and confirms four o'clock isn't the standing admin block. Only then does it say yes — and offer four o'clock as a slot it can genuinely keep.
If any rule had failed, the caller would never have heard "four o'clock." They'd have heard the next real opening instead. That's the whole game. The assistant sounds effortless precisely because the rules did the hard thinking before it opened its mouth.
Rules are a living thing, not a one-time task
The honest limitation of all this: you'll get some numbers wrong at first, and that's fine. Nobody guesses their true durations perfectly on day one. The trick is to treat your booking rules as something you tune, not something you set and abandon.
The best signal is your own frustration. If you keep running late in the same window, your durations or buffers there are too optimistic — widen them. If you're staring at empty slots the assistant refused to fill, your rules are too strict — loosen them. If a particular service keeps needing you to clean up after a booking, it probably belongs in the human-handoff pile, not the auto-book pile. Every summary and transcript the assistant sends you is data about whether your rules match reality.
- Running late often? Durations or buffers in that window are too tight. Add ten minutes and see.
- Suspiciously empty? A rule is over-blocking — check your lead-time horizon and blackout slots.
- Wrong service booked? Two services sound alike on the phone. Add a clarifying question or split them apart.
- Cleaning up after bookings? That service wants a human. Flag it for a callback instead of auto-booking.
Spend fifteen minutes a month on this in the early weeks and the schedule converges on something low-maintenance. The rules stop fighting reality because you've slowly bent them toward it.

Where to start if this feels like a lot
If you're staring at all of this thinking it's a project, it isn't — it's an afternoon, and you don't have to do it perfectly. Start with the two rules that prevent the most pain: realistic durations and a cleanup buffer on anything that needs turnover. Those two alone kill most double-bookings. Add lead-time limits and blackout slots once the basics are running.
In Vunoon, this is a short wizard, not a spreadsheet. You describe your services, set their durations, and add buffers and blocks in plain language. Then — and this is the part worth doing — you actually call the assistant and try to book something impossible. When it politely refuses and offers a real slot instead, you'll know the rules are working. Then you forward your number, and the calendar starts defending itself.
How do I know what duration to set for each service?
What's the difference between a duration and a buffer?
Should I let people book months in advance?
Can the AI handle services that shouldn't be auto-booked?
What if I set a rule wrong?
Do I have to configure everything before going live?
Set your booking rules once, then trust your calendar
Describe your services, durations and buffers in a short wizard, talk to the assistant to test it, then forward your number when it's booking exactly how you want.
See how phone booking works
Vunoon builds an AI phone assistant that answers your business calls 24/7 — it books appointments, answers common questions and sends you a summary of every conversation.