Your AI receptionist answered the missed call. What's it doing the other 23 hours?

Your AI receptionist picked up the missed call. Tick. You've solved the easy bit.

Most NZ owners I talk to are stuck right there. Is a virtual receptionist worth it? What does it cost?

Same five questions, every meeting.

Sound familiar?

Meanwhile a US dental chain has a voice agent sitting on hold with the insurer for 45 minutes a call. 10,000 times a month.

Five jobs that used to belong to humans. Gone before your competitor figured out the script.

The receptionist is the on-ramp. The rest of the road is where the operators who win get rich.

Same engine. Five jobs.

Here's what nobody selling you a receptionist mentions.

The voice agent engine that picks up your missed call at 9pm? Same engine. Four other playbooks. Already shipping in the rest of the world.

Same agent on your line. Same minute pricing. Same 800ms first-token reply that makes it feel like the kid with a coffee just answered.

Different jobs.

Recognise the pattern? The Hamilton home care agency we wrote about lost $641,000 a year to after-hours voicemails. The maths is in the Hamilton home care write-up.

Sam in Dunedin? 300 calls a month gone. Fixed it in 11 days.

Recovered $231,000 a year. That one's in the Dunedin property manager case study.

Both wins. Both inbound concierge. Both the on-ramp you've already weighed up.

Below are the five playbooks the rest of the world is running. Each backed by a real US, EU, or AU deployment with a name and a number.

Each one's ready to ship in NZ this quarter. Each one's tied to a Waboom build we've already done a variation of.

Pick the one that matches the leak in your business.

Playbook 1: Outbound triggers (the system fires the call)

The receptionist waits for the phone to ring. Outbound triggers don't wait.

Something happens in another system. Xero stamps an invoice 21 days overdue. Your CRM flags a hot lead that's gone cold for 90 days.

A patient gets discharged from a hospital. A renewal price-shock hits your Salesforce instance.

Each one fires a voice agent. The agent calls the human, has a friendly conversation, takes notes. Routes the outcome to your team.

A US sub-prime auto lender called Consumer Portfolio Services puts these on every loan in the early overdue window. Recovery rates north of 80%. The agent auto-generates a compliance audit pack for the regulator after each call.

Want to see what your version looks like?

Want the Auckland version? An electrician's office running Xero. Invoice ticks past day 21.

Voice agent calls the homeowner. Friendly chase. "When can we expect that to land?"

Logs the answer. Emails the AR mailbox a single 5pm digest. Five awkward calls a week neither you nor your office manager wanted to make.

You don't have to take my word for it. We already run the outbound side of this. We covered it in HubSpot dormant leads — 44,000 contacts, 12 meetings booked.

The trigger source is different in your business (CRM dormancy vs Xero overdue). The mechanism is identical.

(Honest moment. First time we ran this on a real client's database, their office manager rang me to ask if we'd hired three new people overnight. We hadn't. Same agent.)

What does it need from you?

A trigger source (Xero, HubSpot, Salesforce, your scheduler). A tight call script. An honest call window so you don't bother people at dinner.

Most clients are live in 7 to 11 days.

Playbook 2: Mid-call functions (the agent fires actions while still on the line)

This is where most operators in your space stop short.

The agent answers. The agent talks. The agent ends the call.

The next system update? 15 minutes later, when a human in your team reads the transcript.

That's the slow version.

The fast version: the agent fires functions during the call. While the human you're trying to convert is still on the line.

A Sydney-style business broker uses this on inbound seller enquiries. AI captures revenue, EBITDA, sector, asking price, owner motivation.

Mid-conversation, the agent triggers a function call. Pulls comparable transactions from a sector database.

Emails the seller a personalised valuation report straight to their inbox. The seller hangs up. They open their email. The report's already there.

Recognise that flow?

Same trick for restaurants. Golden Nugget casino-hotel runs $600k a month through a voice agent that takes the deposit on the same reservation call. The agent captures card details securely. Processes the charge before goodbye.

Sound useful for your business?

Or for solo experts. A 400K-follower author handed his "what do you charge?" calls to a voice agent that books straight to Calendly mid-call. He built it in under an hour.

Want the Wellington version? A solo consultant on LinkedIn.

Voice agent answers booking enquiries in a voice that sounds like him. Captures intent. Books the discovery call straight to his calendar. Sends the invoice link.

He sleeps.

We covered the function-calling stack in how to extract variables and trigger external actions mid-call.

We've solved the technical part. The hard part is you deciding which functions are worth wiring up.

Playbook 3: Spike-load and queue killers

Some of your calls are short and high-volume. Some are long and rare. The third playbook handles the long-and-rare category, plus the volume spikes humans can't.

Three flavours, all shipping in production right now.

The hold-time killer. A US dental chain called West Coast Dental puts AI on the call to the insurer. The voice agent dials the payer. Sits on hold for 45 minutes.

Navigates the IVR. Verifies the patient's cover. Writes the result back into your practice management system.

10,000 of those calls a month. Five staff freed up.

Your Hamilton specialist clinic version? AI does the Southern Cross pre-auth call so your nurse practitioner doesn't.

The handoff catcher. Sometimes your voice agent does need to escalate to a human.

A US auto group called Toma built a "clawback" mechanism. If the human doesn't pick up the escalation in 30 seconds, the agent retrieves the call. Books the appointment anyway. Files the human-pickup ticket as a follow-up.

That mechanism saved 22,000 calls in 90 days at one 16-dealership group.

We've shipped a variation of this. We documented the mechanic in agent-to-agent transfer with full context preservation.

The storm day. A European airline put 100% of its disruption-day rebookings on AI. When 70 flights cancel at once, your humans cap at one call a minute. The agent runs ten thousand simultaneously.

AKL fogging in next winter? Air NZ has already announced its OpenAI partnership. This is the obvious next move for them.

(I'd put a coffee on it shipping before the 2027 winter cycle, not after.)

The pattern: humans are linear. The agent is parallel. Anywhere your phone queue blows out for predictable reasons, the agent eats the queue.

Playbook 4: Localised persona (the agent sounds like the market it sells into)

Hunter Douglas runs Buddy in Australia. Archie in the UK. Same engine, different persona per market. Conversion lifts when your customer hears their own accent.

The pronunciation matters even more in NZ.

We've already written about why most agents butcher Aroha and Whakatāne. And what that costs you on call one. Full breakdown in the AI voice agent pronunciation dictionary.

Want the Tauranga version? A DTC e-commerce operator with a Kiwi-accented voice agent. Call her Tane.

She knows the difference between Bay of Plenty and Bay of Islands. Pronounces Mauao correctly. Sounds like the kid working the till at Mount Hot Pools on a Saturday.

Your customer's nervous system relaxes when they hear their own accent. Operators who deploy a generic American voice to a NZ customer base are quietly leaking trust on every call.

The technical part? A 30-minute setup in the persona panel.

The hard part is most operators don't know it's a knob they can turn.

Playbook 5: Pilot before you scale (the discipline that separates wins from disasters)

The honest one. The one most operators skip.

A builder on X recently posted his rebuild story. First voice agent looked perfect in the demo.

Production was a different story. Three things only break under load. He rebuilt from scratch.

Me: "Why didn't the demo catch it?"

Builder: "Demo runs 10 calls. Production runs 600 a day."

Me: "And the orchestration layer?"

Builder: "Falls over at 80 concurrent."

Pilot before you scale. Run 60 calls against a controlled list before you point the agent at a 1,400-contact database.

The Hamilton mortgage adviser who's about to switch on outbound to her entire book? Test against 60 first.

Not because the technology is unreliable. Because the workflow needs to break in three places before it breaks in fifty.

We baked this in. Batch testing voice agents covers the regression-test discipline.

The one that separates the operators who ship from the ones who blow up their phone reputation in week 2.

You don't have to learn this lesson on your own database.

Pick the one that bleeds the most

Five playbooks. One engine. Same minute pricing. Same NZ compliance under the Privacy Act 2020.

Which one matches your business?

Run the test. Open your phone log for the last 60 days. What do you see most of?

Or:

The operators in your suburb who deploy first get the local market with no entrenched competition.

The ones who wait three years buy it back at a premium when their competitor has already taken the brand recognition.

Far out, that's a long article.

Pick one playbook. Ship one experiment. You've decided the receptionist part. The rest of the road is yours.