
On this page
Somebody calls your business at 8:14pm. It is not an emergency. They are on the sofa with their laptop open and they want to book something.
By 8:20pm they are on a competitor’s website. That is the whole problem, and it has nothing to do with how good your business is.
Decide what “after hours” means
Start by writing the actual hours, in your customer’s time zone, for each day of the week. Include the awkward ones: half-days, holidays, the Friday afternoon when everyone leaves early.
Most teams discover their real coverage is narrower than the hours on the door. A practice that closes at 5pm but stops answering at 4:40pm has an after-hours problem at 4:45pm.
The three tiers
You do not need the most expensive option. Pick the tier that matches what the calls are worth.
Tier 1 — Do not lose the details (free)
A message that captures what the caller wanted, plus a phone number, and sends them a booking link.
The key detail: it must be a text or a form, not a voicemail. Voicemail is where after-hours enquiries go to be ignored — it takes effort to listen, and nobody does it while opening the shop.
Tier 2 — Answer the routine questions
Hours, address, parking, whether you take new patients, what a typical job costs. This is most of the volume and almost all of it is repeatable.
An AI receptionist handles this well because it is not making decisions — it is reading from a short approved list and booking what it can.
Tier 3 — Book into the calendar
The real version: the caller books at 9pm and the appointment exists in the morning. This needs three things working together:
- A calendar the system can see
- Rules for which slots are bookable
- A confirmation message so the customer knows it is real
Get those right and after-hours stops being a gap. Get them wrong and you will find double-bookings on Monday.
The mistake almost everyone makes
Treating an emergency like a booking. A burst pipe, a lockout, a patient in pain — those are not appointments. They need a person, immediately, on a number that actually rings.
Set this up first, before anything clever:
- The AI never diagnoses, quotes or advises.
- The caller is told the emergency number, and a callback is sent to the on-call person with the details.
- If nobody acknowledges it within a set time, it escalates again.
Write that rule down and test it. An after-hours system that handles routine calls beautifully but mishandles an emergency is worse than a voicemail.
What the message should say
Short, and honest about being automated:
“Thanks for calling Northside Dental. We are closed until 8am. I can book you now, or text you a link — what is easier?”
Two things it should never do: pretend to be a person, and promise a callback time you cannot keep.
The Monday problem
Everything captured over the weekend piles up. If nobody owns the Monday morning review, the leads go stale before anyone reads them.
Assign one person, one time (9am), one list: every after-hours enquiry from the last 72 hours, with what happened to it.
Measure three numbers
- After-hours calls per week — how big the gap is
- How many got a booking or a callback — whether the system works
- How many were emergencies — whether the rules are right
If the first number is high and the second is low, the system is collecting information and not converting it. That is a routing problem, not a technology problem.
What Monday morning looks like
In BittaFlow, every call that came in overnight is already sitting in the shared inbox with a name, a number and a reason. The ones it could book are booked against the calendar. The ones it could not are flagged for a person, with a note saying why.
Nobody starts the week by listening to voicemail.


