Case studies

Every number here came out of the call record

Not testimonials, and not a survey. These are the actual figures from the portal call log for businesses running Johnni, recomputed from source on 21 September 2026 across 20,141 calls. Client names are withheld, because a booking record is commercially sensitive and it is not ours to publish.

20,141calls examined
48client lines
Dec 25 to Sep 26the period covered
The fleet, all time

What the whole book looks like.

Every line running Johnni, added together, over roughly nine months.

18,785calls taken, 17,097 of them connected
4,713jobs and appointments booked
345 hrsof phone time nobody on a payroll had to spend
Where the bookings came from

Jobs booked, line by line.

Every bar is one client’s lifetime booking count. Hover any bar for the exact figure.

Electrical and fire compliance: 1,423 jobs bookedElectrical and fire compliance1,423IT support and managed services: 1,018 jobs bookedIT support and managed services1,018Appliance repair, two businesses: 524 jobs bookedAppliance repair, two businesses524Exterior cleaning: 266 jobs bookedExterior cleaning266Plumbing and electrical, NZ: 179 jobs bookedPlumbing and electrical, NZ179Martial arts, outbound: 112 jobs bookedMartial arts, outbound112Tax and accounting: 46 jobs bookedTax and accounting46

These seven records, eight client lines, account for 3,568 of the 4,713 bookings. The remaining 1,145 are spread across the other 40 lines, which are shorter runs or narrower jobs.

The strongest records

Seven records, in full.

Ranked by jobs booked. Each of these has enough months and enough volume that the percentages mean something.

AI receptionist case study photo: commercial electrician testing a switchboard after hours for a fire compliance contractor Electrical and fire compliance, national

National electrical and fire compliance contractor

Electrical and fire compliance, national

4,760Calls
1,423Jobs logged
83%Handled right
74.5 hrsPhone time

The biggest line on the book by a distance. More calls in its best month than most businesses take in a year. There are actually two jobs running on this account, not one.

The day line is recognition, not conversation. Six of their regular site contacts are loaded as known callers, matched on the number they ring from, so those callers are greeted by name and skipped straight past the intake script to site, issue, access and escalate. That feature was built for this line and is now available to any client.

The after-hours line is an escalation chain. When somebody rings overnight needing to actually speak to a person, the receptionist takes the site, the fault and the access, then starts ringing the on-call operators one after another until it finds one who picks up and accepts the job. The moment somebody does, it messages both of them: the operator gets the job details, and the caller gets a text telling them who is coming and that they are on the way. Nobody sits on hold while a human works down a contact list, and nobody is left wondering at 2am whether the message got through.

Why 4,760 calls is not 4,760 enquiries. A line doing both of those jobs touches a lot of calls that were never going to be a booking, and the honest read matters more than a flattering one. Here is the whole 4,760, split three ways.

That is also why we do not publish a booking rate for this line. The scorer counts bookings against every call the line touched, including the calls it made itself, so the percentage would flatter or punish it depending on how busy the escalation chain was that week. The count is the number worth trusting.

AI receptionist case study, electrical and fire compliance: the after-hours escalation chain rings on-call operators one after another until one accepts, then texts the operator and the caller
Where the 4,760 calls went
1,423Emergency call-outs logged and dispatched
2,490Connected, but not a new job
847Never connected at all

The middle block is the one people ask about. It holds the outbound legs the system dials itself while it hunts for an available on-call operator, and plenty of those ring out to voicemail before the next one picks up. It also holds existing customers chasing a job already in the diary, callers who needed an answer rather than a booking, and the usual wrong numbers and sales calls. We have not split it further here because we will not guess at a number. The 1,423 and the 847 are counted, the 2,490 is what is left over.

Best 30 days

In its busiest month it answered 1,591 calls and logged 532 emergency call-outs. Twenty-six hours of phone work, overnight and at weekends, that nobody on the team had to stay up for.

AI receptionist case study photo: IT support technician on a headset at a managed services desk Melbourne

IT support and managed services business

Melbourne

3,095Calls
1,018Jobs booked
95%Booking rate
97%Handled right

The best quality record on the book, and the most consistent: eight full months between 240 and 446 calls, with no quiet patch. A 95% booking rate across 1,072 genuine opportunities is not one good month cherry picked, it is 3,095 calls deep.

Two things got it there. They forward their real business line into a number we own, so every call goes through the receptionist rather than only the overflow, which is why the volume is steady instead of spiky. And the assistant has exactly one tool with a live backend, a new-customer hand-off wired to their existing field names. It is not trying to do five things adequately.

AI receptionist case study, IT support and managed services: the whole business line forwards to the receptionist, with a single hand-off tool wired into the CRM
Where the 3,095 calls went
1,018Jobs booked
54A genuine opportunity that did not book
2,023Not a booking call in the first place

The scorer judged 1,072 of those calls a genuine chance to book, and 1,018 of them booked. The other 2,023 were never a new job: existing customers with a support question, people chasing something already open, and the ordinary wrong numbers. They still had to be handled properly, and 97% of them were.

Best 30 days

In its best month it answered 361 calls and booked 157 of them, which was 96% of every call that could be booked, and handled 93% of service calls without anyone stepping in.

AI receptionist case study photo: appliance technician testing a washing machine part while the homeowner watches Appliance repair, Perth and Melbourne

Appliance repair, two businesses

Appliance repair, Perth and Melbourne

2,644Calls
524Repairs booked
93 to 97%Handled right
57.9 hrsPhone time

Two appliance repairers, one in Perth and one in Melbourne, running very different setups and landing on the same number of bookings: 262 each. Together they are the second largest source of jobs on the book after the electrical line.

The Perth line takes a message, by design. The office shares the number, so there are no transfer tools at all. Every one of its 262 bookings came from the receptionist holding the whole conversation itself, and in its biggest month 60 of 529 calls came in after 5pm or at the weekend, the ones a message bank would have swallowed.

The Melbourne line books straight into the scheduler. It offers a real two hour arrival window and writes the job, and it runs a service-area check against the client’s 30 km map, so callers out of range are turned away politely instead of becoming a booking somebody has to ring back and cancel. A quarter of its calls arrive after the office has shut, and they book at roughly the same rate as daytime ones.

AI receptionist case study, appliance repair: a 30 km service area check that books in-range callers with a two-hour window and turns out-of-range callers away politely
Where the 2,644 calls went
524Repair jobs booked
579A genuine opportunity that did not book
1,541Not a booking call in the first place

The widest gap on the page, and there is a reason for it. The Perth office shares its number, so a large share of its calls were never repair enquiries, and the Melbourne service-area check turns away out-of-range callers on purpose. Between them, 93 to 97% of the non-booking calls were still handled properly.

Best 30 days

Perth’s biggest month: 529 calls, 60 of them after hours, 93 repairs booked. Melbourne’s best: 151 calls, 44 of them after the office had shut, 49 repairs booked, 20 of those from out-of-hours calls.

AI receptionist case study photo: exterior cleaner pressure washing a Sydney driveway Sydney

Exterior cleaning business

Sydney

999Calls
266Jobs booked
92%Booking rate
88%Handled right

The steadiest line on the book. Nine months running, never below 83 calls in a month and never above 134. That flatness is the selling point: it is what a business looks like when the phone stops being a variable.

Nothing clever explains it. The line answers on the first ring, it knows the service list and the pricing, and it books from the real calendar. Doing that every day for nine months is what turns a spiky enquiry flow into a flat one.

AI receptionist case study, exterior cleaning: nine months of monthly call volume held inside an 83 to 134 band, 266 jobs booked
Where the 999 calls went
266Jobs booked
23A genuine opportunity that did not book
710Not a booking call in the first place

Only 23 real quote enquiries got away across nine months. The 710 were not quote calls at all: customers about a clean already booked, questions about access and timing, and wrong numbers. 88% of those service calls were resolved without anybody stepping in.

Best 30 days

Best month: 128 calls, 45 jobs booked, and every single bookable call converted.

AI receptionist case study photo: plumber and electrician checking a job on a tablet beside their ute at dawn in New Zealand Trades, New Zealand

Plumbing and electrical business

Trades, New Zealand

802Calls
179Jobs booked
68%Booking rate
13.8 hrsPhone time

The first New Zealand client, and the only one running two trades as two independent receptionists on two numbers. Four solid months between 178 and 183 calls each.

Transfers to a human are held back until an emergency callout fee has actually been quoted, so an after-hours job is agreed before anybody gets woken up. The two lines run their own scripts, their own pricing and their own rules.

AI receptionist case study, plumbing and electrical in New Zealand: two trade numbers running two receptionists, with a human transfer only after the emergency callout fee is agreed
Where the 802 calls went
179Jobs booked
84A genuine opportunity that did not book
539Not a booking call in the first place

Two trades on two numbers, so the 539 covers both: existing customers, callers after a price rather than a booking, and the emergency calls that get transferred to a human once a callout fee has been quoted.

Best 30 days

Across two trade lines it has answered 802 calls and booked 179 jobs, with the busiest month running 215 calls through a business that used to let them ring out.

AI receptionist case study photo: martial arts coach welcoming a teenage student and parent before a trial class Outbound lead calling, Queensland

Martial arts centre

Outbound lead calling, Queensland

691Calls made
112Trials booked
87%Booking rate
25%After hours

The proof that this works outbound, not just on the incoming line. The assistant rings web and ad leads back and books them into a trial class.

It also runs in announcement mode with confirm-on-dial, so a lead is only ever called once. That restraint is the difference between an outbound caller and a nuisance.

AI receptionist case study, martial arts centre: outbound calls to web and ad leads, one call per lead, 112 trial classes booked out of 691 calls
Where the 691 outbound calls went
112Trials booked
17Reached the lead, no trial booked
562Never reached the lead

Outbound reads differently from an incoming line. The 562 are calls that never got a person: voicemail, no answer, or a number that no longer works. Of the leads actually reached, 112 out of 129 booked a trial. Each lead is only ever dialled once, which is the difference between an outbound caller and a nuisance.

Best 30 days

In one month it made 220 outbound calls to leads and booked 60 of them into trial classes, which was every single lead worth booking.

AI receptionist case study photo: accountant taking a trades business owner through tax planning Professional services

Tax and accounting firm

Professional services

199Calls
46Jobs booked
92%Booking rate
92%Handled right

The proof that this is not only for trades. The assistant texts the caller the right intake form mid call, choosing between new individual, new business, personal return, general query or a named accountant, and hands the rest to the team.

AI receptionist case study, tax and accounting firm: the receptionist works out the enquiry type and texts the caller the right intake form during the call
Where the 199 calls went
46Appointments booked
4A genuine opportunity that did not book
149Not a booking call in the first place

A practice line is mostly not new business, and that is normal. The 149 are existing clients, people chasing a return, and general queries that get texted the right form and handed to the team. 92% were handled properly without an interruption.

Best 30 days

In its best month it answered 44 calls, converted every single one worth converting, and texted each caller the right form before they had hung up.

Method

How these were counted.

Every figure comes from the portal call table, 20,141 rows, grouped by client. Where a business has run on more than one voice platform over time, those records are counted together, because the client experienced them as one receptionist.

On each client above, where the calls went is split three ways. The booked count is exact. The boundary between a genuine opportunity that did not book and a call that was never a booking in the first place is derived from the scorer’s booking rate for that line, so treat those two as close rather than exact. They always add up to the real total. The one exception is the electrical and fire compliance line, where the count includes the outbound legs the system dials itself, so it is split by what connected instead.

Calls means every call that reached the line. The gap between calls and connected calls is people who hung up while it was still ringing. Booked is the portal scorer’s read that the call produced a job or an appointment. Booking rate is booked calls as a share of the calls the scorer also judged a genuine opportunity, so on a couple of lines it marks more calls booked than bookable. Treat the counts as solid and the percentages as indicative. Handled right is customer service calls resolved properly. After hours is calculated against each client’s real trading hours and public holiday state.

Our own test lines are stripped out of every figure. Test calls made by owners from their own phones are not separable in the data and remain counted, which is roughly a third of calls on a brand new signup, and that is exactly why nothing under 60 days old is quoted here at all.

Every time-of-day figure on this page is computed from corrected timestamps, checked against each client’s real trading hours rather than a server clock.

About these numbers

Frequently asked questions

Why are the business names not shown?

Because a client’s call volume and booking record is commercially sensitive, and it is their information rather than ours. The sector, the location and the numbers are real and unaltered. If you want a reference, ask and we will see which clients are happy to take a call.

Are these your best clients or a fair sample?

The seven records above are the strongest by jobs booked, and we say so plainly. The fleet totals at the top of the page are every line, good and bad, including the ones that book poorly. Forty-eight client lines produced 4,713 bookings between them.

Why is nothing recent on the list?

Anything live for less than 60 days is excluded, because early call logs are full of the owner testing their own line, which inflates the numbers. A business needs a couple of real months before its figures mean anything.

What is a booking rate, exactly?

Booked calls as a share of the calls that were a genuine opportunity to book. A wrong number or a supplier call is not a missed booking, so it is not counted against the rate. On a small number of lines the scorer marks more calls booked than bookable, so the counts are the solid figure and the percentage is indicative.

Can I get these results?

The honest answer is that it depends on your call volume and what you ask it to do. The lines that book best have two things in common: the whole business number forwards through the receptionist rather than just the overflow, and the assistant has a small number of tools wired properly instead of many done loosely. The 7 day trial costs nothing and tells you more than this page can.

Free trial, no contract

See what your own call record looks like.

Seven days free. Every call recorded, transcribed and counted in the portal, so the numbers you get are the same kind as the ones on this page.

Plans from $49 a month plus GST, or $44 paid annually. No lock-in contract.