Fourteen Days From Signature to Live
AI reception fails far more often through rushed rollout than through poor technology. This is the sequence that avoids the usual mistakes: gather before you configure, test before you cut over, and start with a slice of traffic rather than all of it.
Work through it in order. Progress is saved in this browser, so you can tick items off across the full fortnight without losing your place.
Fourteen-Day Rollout Checklist
Gather, configure, test, cut over. In that order.
Your ticks are saved in this browser, so you can work through the list over several sessions.
01Days 1-3 · Gather
0/7Write it down before anybody configures anything.
02Days 4-7 · Configure
0/7Build the flows and connect the systems.
03Days 8-11 · Test
0/8Test the awkward calls, not the easy ones.
04Days 12-14 · Cut over and review
0/8Start with a slice. Keep the old routing ready.
A general implementation guide rather than provider-specific instructions. Exact steps vary between platforms, and call recording obligations differ between Australian states and territories.
The Three Rules of a Calm Rollout
Nothing here is technically difficult. The discipline is in resisting the urge to switch everything on at once, which is what turns a straightforward change into a fortnight of firefighting.
Start with a slice, not the whole line
Route after-hours calls first, or overflow only when your line is busy. You keep your existing safety net intact, real callers exercise the system on genuine enquiries, and nothing catastrophic happens if week one reveals a gap in your configuration.
Test with your worst calls, not your best
Anyone can pass a scripted test call. Test the angry caller, the one asking for a person by name, the one with a question you never anticipated, and the one who wants to cancel. The failure modes you find in testing are the ones you will not meet in production.
Tell your team before your customers find out
Staff who discover the AI by accident undermine it, usually with the best intentions. Staff who understand what it handles, what escalates to them, and why it exists become the people who make it work properly.
The Fourteen-Day Sequence
Four phases. Most businesses can compress this into a week if they need to, but the ordering matters far more than the pace.
Days 1 to 3: gather
Collect everything the system needs to know before anyone configures anything: hours, services, pricing rules, staff names, common questions and the answers you actually want given.
Days 4 to 7: configure
Build the greeting, call flows, booking rules and escalation paths. Connect the calendar or practice management system and confirm a real booking writes end to end.
Days 8 to 11: test
Run structured test calls covering the awkward cases, not just the happy path. Fix what breaks, then test again. Brief the team and agree who owns escalations.
Days 12 to 14: cut over and review
Forward a slice of live traffic, monitor the first days closely, review every call in week one, and only then widen coverage.
Where Rollouts Usually Go Wrong
Four recurring mistakes, each of which is trivial to avoid if you know it is coming and expensive to unpick if you do not.
Configuring before gathering
Teams frequently start building call flows before anyone has written down what the business actually charges, which services it declines, or what the after-hours policy really is. The configuration then encodes guesses, and those guesses are discovered by customers rather than by staff.
- Write down your real opening hours including public holidays and lunch closures
- Decide in advance what the AI should say about price, and what it must not say
- List the services you do not offer, so it can decline them cleanly
- Agree the after-hours policy before it is encoded, not afterwards
No agreed escalation owner
Escalation rules that route to “the team” route to nobody. Every escalation path needs a named person, a fallback for when that person is unavailable, and an agreed response time, otherwise urgent calls land in a shared inbox nobody is watching.
- Name a specific person for each escalation type, not a group
- Define a fallback for leave, weekends and after hours
- Agree what counts as urgent in your business, in plain language
- Test the escalation path end to end before go-live, including out of hours
Cutting over on a Monday morning
Monday is the highest-volume, highest-stakes period for most Australian service businesses. Cutting over then guarantees that any configuration gap is discovered at maximum volume with maximum consequence. Mid-week, mid-afternoon is dull and forgettable, which is exactly what you want.
- Cut over on a Tuesday or Wednesday afternoon where possible
- Avoid the week before or after a public holiday
- Have someone available to watch the first two hours of live calls
- Keep the old routing ready to restore for the first week
Not reviewing week one properly
The single highest-value hour in the whole rollout is sitting down at the end of week one and listening to or reading every call. That review typically produces five or six small corrections that lift performance substantially, and almost nobody does it.
- Read or listen to every call from the first week, without exception
- Log every question the system handled poorly and add the correct answer
- Check that bookings actually landed correctly in your calendar
- Only widen coverage once week one has been reviewed and corrected
Next Steps
Phone Answering Script Template
The greeting, qualifying questions and capture fields to hand your provider on day one.
Build the script →AI Receptionist Buyer’s Guide
Not signed yet? Thirty questions worth asking before you commit.
Open the guide →How to Set Up an AI Receptionist
The longer-form walkthrough of what setup involves in practice.
Read the walkthrough →Frequently Asked Questions
Can this really be done in two weeks?
Comfortably, and often faster. The technical configuration for a straightforward single-location business is typically a day or two of work. What consumes the fortnight is the gathering and the testing, both of which depend on your availability rather than the provider’s. Businesses that already have documented hours, services, pricing rules and a clear after-hours policy frequently go live within a week. Businesses discovering that those things have never been written down anywhere take longer, and the delay is genuinely worth it.
Do I need a new phone number?
No, and you should resist any suggestion that you do. Your existing number carries your search rankings, your directory listings, your signage, your vehicle livery and years of customer familiarity. The standard approach is call forwarding from your current number, which you can configure to forward everything, only when busy, only when unanswered, or only outside your hours. Some providers do issue a separate number for testing during setup, which is sensible, but your published number should not change.
What should I route first if I do not want to switch everything at once?
After-hours is almost always the right starting point. Those calls currently go to voicemail, so the downside is essentially zero. The worst case is that the AI performs no better than the voicemail it replaced. It also exercises the system against genuine enquiries with real urgency, which is the traffic you most want it to handle well. Overflow, where calls divert only when your line is busy or unanswered after a set number of rings, is the natural second step. Full-time coverage should be last.
How do I brief my team without making them think they are being replaced?
Be specific and be early. Explain exactly which calls the system takes, which ones come to them, and what problem it is solving. In most deployments the honest framing is straightforward: it answers the calls nobody was available to answer, particularly after hours and during overflow, and it removes the interruptions that stop people doing the work they were actually hired to do. Ambiguity is what generates anxiety. A team that has heard nothing will assume the worst, and a team that quietly resents the system will route around it.
What if it gets something wrong in the first week?
Assume it will, and plan for it rather than treating it as a failure. This is precisely why the recommendation is to start with a slice of traffic and keep your previous routing ready to restore. Log every error, correct the configuration, and re-test. Almost all first-week issues are configuration gaps rather than technology faults, which means they are fixed in minutes once identified. A rollout with zero corrections in week one usually means nobody reviewed the calls.
Do I need to tell callers the call is recorded?
If calls are being recorded, then generally yes. Surveillance devices and listening devices legislation differs between Australian states and territories, but the practical standard adopted by most businesses is a short notification at the start of the call. Most providers include this in the opening greeting by default. Confirm it is enabled rather than assuming, and if your industry carries additional obligations (health, legal and financial services frequently do) check those separately. This is general guidance rather than legal advice.
Want a Second Pair of Eyes on Your Rollout?
Send us your call flows and escalation rules before you go live. We will tell you what we would change, whether or not you are a customer.