The choice is usually staged as a salary against a subscription, and both of those figures are wrong. Two things moved this year in New Zealand and they point in opposite directions: on 1 April the statutory cost of employing someone rose again, and on 5 August Stats NZ published the loosest labour market in more than a decade. Neither is a reason on its own. Together they change which of the two answers you can take back.
The Quarterly Employment Survey put average ordinary time hourly earnings at $44.62 in the June 2026 quarter, against $43.39 a year earlier. Take that as the base for one full-time person and the payroll arithmetic is short. Two thousand and eighty paid hours come to $92,810. Employer KiwiSaver, which rose from 3 to 3.5 per cent on 1 April 2026, adds $3,248. The ACC work levy at $0.69 per $100 of liable earnings — the average rate MBIE set for 2026/27, up from $0.66 — adds $640. Call the year $96,700.
That is a loading of 4.2 per cent on top of pay, which is smaller than most owners expect and is not where the useful number sits. The useful number appears when you stop counting hours you pay for and start counting hours the work can be done in.
| Basis | Hours a year | Cost per hour |
|---|---|---|
| Paid hours — forty a week, fifty-two weeks | 2,080 | $46.49 |
| Less four weeks annual holidays | 1,920 | $50.36 |
| Less twelve public holidays | 1,824 | $53.02 |
| Less ten days sick leave, if fully taken | 1,744 | $55.45 |
So a business budgeting from $44.62 is buying availability at $53 to $55 an hour. The gap is 19 to 24 per cent and it is purely statutory: no desk, no software, no overheads, no management time. Run the same ladder on the adult minimum wage, $23.95 since 1 April 2026, and it goes from $23.95 to $28.46.
What actually changed in April is worth stating plainly rather than overselling. KiwiSaver moving half a point and the work levy moving three cents cost about $492 a year for one person at the average wage. That is not a reason to do anything. It is a reason to redo the arithmetic once, because the direction is fixed: the default KiwiSaver rate goes to 4 per cent in April 2028.
Ask what a hire costs in this market and the same set of numbers comes back: $18,982 on average, $9,772 entry level, $17,841 mid, $34,440 for an executive, 47.2 days to fill a role against 39.2 in Australia. All of it traces to a single survey of more than 1,500 HR professionals across Australia and New Zealand, run by ELMO Software and published in November 2019. Self-selected respondents from a payroll vendor’s own audience, collected before the pandemic reordered wages and hiring alike, still quoted in 2026 as current. Same documentary dead end as speed-to-lead statistics: the most-cited figure in a field is the one nobody has remeasured.
Use it for the order of magnitude and nothing finer — recruitment costs weeks of attention, not a line item. What is checkable is the agency fee, quoted here at 15 to 20 per cent of first-year salary, which on a $92,810 role is $13,900 to $18,600.
Automation gets sold as the reversible choice and a hire as the commitment you are stuck with. In New Zealand in 2026 that is close to inverted, and this is the part of the decision that rarely reaches the spreadsheet.
Every employer has been able to use a ninety-day trial period since 23 December 2023, when the Employment Relations (Trial Periods) Amendment Act extended it beyond firms with fewer than twenty staff. It has to be agreed in writing, signed before the first shift and entered in good faith — but within those bounds, a hire that is not working can be ended inside three months. A person who is wrong for a role also announces it within a fortnight, in ways nobody has to be monitoring.
An automation fails differently. It does not resign. It stops matching three per cent of invoices, or silently drops the enquiries arriving through one form, and the first evidence is a year-end number nobody can account for. The reversal window on a hire is ninety days and visible; on an unwatched automation it lasts however long it takes someone to notice, routinely twelve months. Both belong in the grid.
| Test | How to measure it | Points to automating | Points to hiring |
|---|---|---|---|
| Size of the task | Hours a week, written down in half-hours for a fortnight — not remembered | Two to eight hours: $4,800 to $19,300 a year of someone’s time | Over fifteen hours: $36,300 and up, which is a role rather than a task |
| Variance | Of the last twenty instances, how many followed the same path | Eighteen or more | Fewer than fifteen |
| Judgement | Could a competent stranger work from your written rule | Yes, and the rule fits on one page | No, or the page keeps growing every month |
| Exposure | Does the output carry your name to a customer | Internal work, or a receipt-shaped message | Anything negotiated, disputed or sold |
| Failure mode | If this breaks quietly, when would you find out | Same day — the missing output is obvious | Months later — then a person needs to own it |
| Reversal | What undoing the decision costs in ninety days | The build, already spent and reusable | Ninety-day trial, if agreed in writing before day one |
Stats NZ put unemployment at 5.6 per cent in the June 2026 quarter, up from 5.4 in March and the highest in more than a decade. Underutilisation, the broader measure that also counts people working fewer hours than they want, reached 13.8 per cent. Wage inflation on the Labour Cost Index ran at 2.0 per cent over the year while consumer inflation ran at 4.1, so labour became cheaper in real terms at the same time as it became easier to find.
Read that honestly and it argues for hiring. Candidate availability is the best it has been since 2015, real wage pressure has gone, and the trial period makes the experiment genuinely cheap to run. We build automation for a living and it is still the correct reading of the data: if the work needs judgement, this is a better year to hire into it than 2024 was.
What it has not changed is the other half. A loose market lowers the price of hiring; it does not turn invoice matching into a job worth doing by hand. Work that was not worth a salary at three per cent unemployment is not worth one at 5.6, and the grid exists to keep those two conclusions apart.
Contracting used to sit in the grey between the two. Since the Employment Relations Amendment Act took effect on 20 February 2026 it does not: a gateway test settles the question first. An arrangement meeting all four criteria — among them a written agreement specifying contractor status, and the freedom to work for others — is a contractor, and the older tests fall away. Miss one and the real-nature-of-the-relationship test returns as before. For a small firm the value is not the levy saving; it is that a risk previously carried quietly for three years now has a checkable answer.
At the June 2026 quarter average of $44.62 an hour, one full-time person costs about $96,700 a year: $92,810 in pay, $3,248 in employer KiwiSaver at the 3.5 per cent rate that took effect on 1 April 2026, and $640 in ACC work levy at the average $0.69 per $100 that MBIE set for 2026/27. That is a 4.2 per cent loading on payroll. The figure that matters more is cost per available hour — after four weeks of annual holidays and twelve public holidays there are 1,824 of them, so the real rate is $53.02, or $55.45 if all ten sick days are taken.
For a task of fixed shape at volume, yes, and the margin widens every year. For anything low-volume or judgement-dependent, no — building and maintaining an automation for work done twice a month never pays back. The threshold worth knowing in this market: at $53.02 per available hour, a task consuming eight hours a week costs about $19,300 a year of a person’s time. Below roughly two hours a week, neither answer is usually right and the task should simply stay where it is.
Yes. The Employment Relations (Trial Periods) Amendment Act extended it to employers of any size with effect from 23 December 2023; before that it was limited to firms with fewer than twenty staff. The conditions are strict — the trial clause must be in the written agreement, signed before the employee starts their first shift, and entered in good faith. Handled properly it makes a hire the more reversible of the two decisions, which is the opposite of how the comparison is usually presented.
For judgement work, yes. Unemployment at 5.6 per cent in the June 2026 quarter and underutilisation at 13.8 per cent mean the widest candidate pool in over a decade, and Labour Cost Index growth of 2.0 per cent against 4.1 per cent inflation means real wages are falling. For repetitive work, no. A looser market makes hiring cheaper without making the underlying task worth a person, and treating it as a reason to staff up on routine processing is how firms end up with a role that has to be disestablished two years later.
Kanaky Tech is an AI automation agency working across New Zealand and the Pacific. Start with a free AI opportunity audit: we map how your business actually runs, rank what is worth automating, and give you a clear scope before anything is built — no obligation.