The How to series · Post 18
The maths of a hire vs an automation
For UK business owners with a small team who are one admin backlog away from writing a job advert: a hire costs about £12,000 a year before NI, and much of the work they'd inherit repeats.
It's 4pm on a Tuesday and you're drafting an advert for a part-time admin hire. The retyping, the chasing, the reformatting into somebody else's template, the copying between systems: somebody has to own it, and the team is full.
Before the advert goes live, it's worth knowing how much of that job is actually human work. That's the loop this post is about: the hire-versus-machine decision, which most owners make on feel and then live with for years.
The maths (a sketch, not a client figure)
The hire. 20 hours a week × £13 an hour × 46 weeks = about £12,000 a year in wages. That's before employer's NI, pension, onboarding and the management time that never stops, and the figure recurs every year.
The machine. No weekly wage. Built once for one loop, then it runs the same steps every week without retraining or notice periods. There's no honest single price for a build, because it depends on the loop. What's real is the trade it changes: the repeating share of the hours stops costing hire money.
The measured example. One industry has put a number on this. Totaljobs surveyed UK recruiters in 2025 and found 17.7 hours of admin per vacancy, roughly £17,000 per recruiter per year, lost to formatting, chasing, retyping and copying between tools. That's recruitment's number, not yours. It's here because of the shape: most of those hours are repeating work, and repeating work is exactly what a machine is for.
Swap recruiters for businesses like yours and the same choice appears. The real question isn't hire or machine. It's which hours.
Score it: the three checks
Run them on the work the hire would inherit. Take reformatting documents into a client's template as the example:
- Hours. How much of the week it eats (0 to 5)
- Repeat. Could a new starter follow it from one page? (0 to 5)
- Risk-if-wrong. If it slips once, how bad? (0 to 5. High means the rules are obvious and a slip is recoverable)
Reformatting scores Hours 4 in most teams that do it weekly, Repeat 4 (writable on one page, which is exactly why a machine can follow it), Risk-if-wrong 3 (a wrong draft is caught at approval). Two checks at 4 or more: start narrow, hand one loop to the machine and measure.
The other side matters just as much. Fee judgement, the awkward conversation, deciding which client gets goodwill, and what to do about the client who always pays late: low hours, but Repeat scores 1, because nobody can write it on one page. That work is human-only, and it's usually why the hire is worth making at all.
Try one thing this week (free)
Before the advert goes out, list what the new hire would do in week one. Mark each item "same steps every time" or "different every day". On Friday, count the marks. If most items repeat, you're about to pay hire money for machine work, and the advert deserves a second look before it costs £12,000 a year.
Where a person stays
Nothing in this decision hands over the deciding. Software prepares: the draft, the reformatted document, the reminder that's due, the chase that's overdue. A person approves anything that goes out. Financial work, compliance work and legal work stay approval-only in every setup we build, with no exceptions. The honest version of hire versus machine isn't people against software. The machine takes the repeating steps. The person keeps the judgement.
FAQ
When the work changes every day, hiring is the honest answer. We'll say so plainly. A machine only works where the work repeats and there's a right answer. The three checks exist to tell those kinds of work apart before either cheque is signed.
Totaljobs' 2025 survey of UK recruiters: 17.7 hours of admin per vacancy, roughly £17,000 a year each. It's one industry's measured example and the only outside number in this post. Your own figure comes from timing your own loops for a week, which costs nothing.
In a team of 2 to 50, usually not. The choice in front of most owners isn't automate the work or keep it. It's who does the repeating share: a new salary every year, or software that runs the same steps, while the people do the parts that need a person.
It depends on the loop, so we don't quote from a blog post. The starting point is the three checks above: they tell you whether a loop is even worth building. Nobody should pay for a machine for work a person should keep.
Yes, and it's often the sensible order. The hire owns the judgement from day one. As each repeating step proves itself, it moves over. The mistake isn't choosing a hire. It's hiring someone into a job that's mostly typing.