Service 03

AI voice agents that answer, book and hand over properly

Not a demo. Voice agents that answer at 6pm, qualify the caller, look the account up, book the job into the real diary and hand a human the context when it matters. Trained on your data, your pricing and your words.

Rate
$150 per hour, ex GST
Start with
$450 remote audit
Builds
Fixed-price quotes
Coverage
Australia wide

What a voice agent is good at, and what it’s not

AI voice agents are good at calls that are frequent, structured and repetitive: after-hours enquiries, overflow when every line is busy, booking and rescheduling, order and job status, and the same five questions every business gets asked.

They’re not the right answer for complaints, negotiations, distressed callers or anything where a person’s judgement is the product. A good agent recognises those calls early and gets them to a human.

What makes one useful rather than a demo

  • It can commit. An agent that can only take a message is a voicemail with better manners. Ours call tools that check live availability, hold the slot and write the booking into your scheduling system.
  • It knows your business. Your service area, your callout fee, your policies and the way you describe your own services, so it answers like your front desk rather than a generic assistant.
  • It knows its limits. A written brief sets what it may do, what it must never do, and exactly when to hand over.
  • It hands over with context. The person who picks up gets a summary, not a caller who has to start again.
  • It leaves a record. Transcript, summary and outcome written into your CRM or job management against the right customer.

An example: after hours

After hours

A voice agent that books the job into the real diary

The agent is only useful if it can commit. This one checks live technician availability, holds the slot, confirms it aloud, writes the booking into the scheduling system and texts the customer. Anything outside its brief goes to the on-call mobile with a summary already written.

Tool callingLive availabilitySMS confirmHuman handover
agents/after-hours.tools.json
[
  {
    "name": "find_available_slot",
    "description": "Next free technician slot for a suburb and job type.",
    "parameters": {
      "suburb": "string",
      "jobType": "string",
      "urgency": "standard|emergency"
    }
  },
  {
    "name": "book_job",
    "description": "Commit the slot. Returns a job number to read back.",
    "parameters": {
      "slotId": "string", "caller": "string", "notes": "string"
    }
  },
  {
    "name": "escalate_to_oncall",
    "description": "Ring the on-call mobile, pass the transcript summary.",
    "parameters": { "reason": "string", "summary": "string" }
  }
]

Where it fits in the call flow

  • After hours. The usual starting point. The agent answers once reception closes, books what it can and sends anything urgent to the on-call mobile.
  • Overflow. During the day it only picks up when the queue is full, so callers get an answer instead of hold music.
  • First line. For high-volume lines, it takes the routine calls and passes everything else to the right person with the details already gathered.

Most clients start with after hours, review the transcripts for a few weeks, and only then decide whether to move the agent forward in the day.

Being upfront with callers

We recommend the agent says it’s an automated assistant, and that callers are told calls may be recorded. Recording and consent rules differ between states, so we check what applies to you. Personal information the agent collects is handled under your privacy policy, and in health settings it’s sensitive information; our page on after-hours answering for clinics covers that case.

How we build and tune one

  1. Pick the calls. From the audit, the call types worth automating first.
  2. Write the brief. Scope, tone, facts, red lines and handover rules, in plain English you sign off.
  3. Connect the tools. Availability, booking, lookup and escalation, wired into your real systems.
  4. Test with your scenarios. Including the awkward ones: wrong suburb, caller who changes their mind, caller who just wants a person.
  5. Soft launch. On overflow or after hours first, with transcripts reviewed closely while it settles in.
  6. Tune. Adjust the brief and tools from what real callers actually say.

For the mechanics of what happens during a call, read how AI voice agents work with a phone system.

Common questions

Will callers know they’re talking to an AI?

They should, and we recommend the agent says so at the start. Callers tolerate an honest, useful assistant far better than one pretending to be a person, and a quick route to a human or a message keeps it respectful.

What happens if the agent doesn’t understand the caller?

It asks once more in different words. If it still can’t help, it takes a message or transfers, depending on the rules in your brief. It should never guess at a booking or a price.

Can it cope with Australian accents and place names?

Speech recognition has come a long way, but suburb names, street names and product names are where errors still show up. We test with your actual vocabulary and build in read-backs, such as confirming the suburb and the booking time aloud, so a mishearing is caught on the call rather than at the door.

What does it cost to run?

Design, build and tuning are charged at $150 per hour plus GST, quoted as a fixed price after the audit. Ongoing usage costs depend on how many calls the agent takes and are set out in the quote before you commit.

Start here

Book the $450 systems audit

We go through what you’re running, what it’s costing you, and where the call is falling out of the process. You get a written plan and a fixed-price quote for the work, and it’s yours to keep whoever you decide to use.

$450 fixed, in AUD, excluding GST. Conducted remotely, Australia wide. On-site audits quoted separately on request.