Before you add a salary,
find out if the work can be removed.
Some of the job needs a person. Some of it is the same three tasks, typed twice, every week. This page separates the two, puts a number on each, and gives you the three honest ways forward.
The test is whether the work needs judgement. Work that varies case by case, needs relationships, or handles exceptions is worth a salary. Work that follows the same steps every time, triggered by the same event, is cheaper to automate: it runs outside office hours, needs no training, and costs the same in year two.
Work that needs a person.
Judgement, relationships, the awkward customer, the job nobody wrote a rule for. It changes every time, which is exactly why software is bad at it. If most of the week looks like this, hire, and do not let anyone sell you otherwise.
Work that only looks like it needs a person.
The same details typed into two systems. The quote chased on day two. The report pulled every Monday morning. Same trigger, same steps, same result. This is the block worth removing, and it is usually bigger than anyone guesses before they count it.
Work nobody is doing at all.
The calls after five. Saturday. The enquiry that came in at 11pm and got answered on Tuesday. It never made the job description because there was never anybody there to do it, and a nine to five hire will not cover it either.
What the hire actually costs. In your numbers.
Placeholder figures, deliberately. Swap in your own; the sum is the point.
Not our numbers and not a promise: put your own salary figure in, then tally one week of the tasks you were going to hand them. The share that repeats is the part worth automating. The rest is a real job, and you should go and fill it.
Three ways to fix it. Two of them aren’t us.
Hire the person
Advertise, interview, pay the salary. Honest verdict: the right answer whenever the week is mostly judgement, site visits, or people who need calming down. No system runs a difficult customer conversation, and anyone telling you otherwise has something to sell.
Hire part-time, or share one
A VA, a part-time admin, or a few hours a week through an agency. Honest verdict: the cheapest way to buy real judgement by the hour, and plenty of businesses should stop right here. Falls down on cover and on hours: they still finish at five, and the repeating tasks still get done by hand, just by cheaper hands.
Remove the repeating work first, then decide
Automate the tasks that fire on the same trigger every time, then look at what is actually left. This one is ours. It only wins where the tally shows a real block of repetition; if your week is all judgement, it does not, and we will say so.
Our recommendation: Do the one week tally before anything else. If under a fifth of the hours repeat, hire the person and do not ring us. If a third or more repeat, remove that block first, because otherwise you pay a salary to do it and still need the same person next year. Bring the tally and we will size it honestly against the maths above.
We count before we build: one week of the tasks you were going to hand a new starter, sorted into judgement and repetition, sized with the sums above. Then we automate the repeating block and tell you plainly what is left for a person, which is often still a real job worth filling. Pricing follows the problem: a one-time setup fee plus a monthly fee for optimisation and support, with options to choose between, never a take-it-or-leave-it.
This site runs on the systems we sell. Every article here was researched, drafted and scheduled by our own machine and approved by a human before it published, and the number in the header is answered by the receptionist we build. Neither of those was ever a job we hired for.
It costs nothing, and it decides the question.
- For one week, note every time you do a task you have done before in the same way. Just the task and the minutes. A notes app is fine, no system required.
- On Friday, split the list into two columns: things that needed you to make a judgement, and things that followed the same steps every time. Add up the minutes in each.
- Set a missed-call auto text on your business mobile today and turn on out-of-hours replies. Free, five minutes, and it covers the hours no hire was ever going to cover anyway.
- Take the repeating column to whoever you were about to hire and ask what else they would do with those hours. If the answer is thin, the job was the repetition, and the repetition is the thing to remove.
Should I hire someone or automate the work?
How much does an employee really cost on top of the salary?
Employer National Insurance and the minimum pension contribution are the two that always apply, and on a mid twenties salary they add a few thousand a year between them. Then holiday and sick cover, kit, software seats, and the weeks of training before anyone is useful. The salary is the number people quote; it is not the number to plan against.
Which tasks should I automate first?
The one that fires on a predictable trigger and has never once been done differently. A new enquiry arriving, an invoice landing, a quote going quiet on day seven. Those three shapes cover most of what small businesses hand to a first admin hire, and they are the ones where a system genuinely does the whole job rather than most of it.
Does automating the work mean making someone redundant?
Almost never, in the businesses we work with. The usual pattern is a team already behind, doing repetitive work in the evenings, with a hire being considered because nobody can keep up. Removing the repetition gives existing people their week back before it removes anybody. If your situation is genuinely a headcount reduction, that is an employment decision and not one to take on a sales call.
What if I automate the work and still need to hire?
Then you hire, with a better job description and a smaller salary band, because the role is now the judgement half rather than the judgement half plus the typing. That is a good outcome, not a failed one. It is also why the tally comes first: it tells you which half you are recruiting for.
How long does it take to remove a repeating task?
A single well defined task, of the kind above, is usually days rather than months. The counting and the agreeing on rules take longer than the building. What takes months is trying to automate a process nobody has written down, which is why the first job is always to write it down.
Asked before booking. Answered straight.
We are too small to automate anything.
Small is when it matters most, because there is nobody to absorb the repetition. The question is not headcount, it is whether the same task happens often enough to be worth removing. Twenty minutes a day is roughly two working weeks a year. Do the tally and let the number decide rather than the size of the business.
I would rather have a person I can talk to.
So would we, for the half of the work that needs one. The choice is not a person or a machine, it is what you buy the person for. Paying someone to retype addresses is an expensive way to have company. Buy them for the judgement and give the typing away.
We tried automating before, it broke, and we hired anyway.
Usually because it was built on somebody else’s secrets and nobody owned it after they left. You own what we build: your accounts, your logins, the documentation. And we start with one task rather than a transformation programme, so a failure is one task big and you can see it coming.