AI or a new hire? When to automate and when to hire

AI vs hiring in a small business comes down to one test: automate the work that repeats to a pattern, and hire when the job needs judgement or presence.

Should I automate or hire? Write down the week first

You cannot answer this from the feeling of being stretched. Log a normal week in rough half-hours, then read it back. The overflow nearly always splits into two piles: work that arrives in the same shape every time, and work where somebody has to weigh something up. Those piles have different answers, and mixing them is how owners end up hiring for the wrong problem.

Say you run a small plumbing firm and you are under water. Written down, the week turns out to be six hours of quotes typed from the same eight paragraphs, three hours chasing invoices, two hours of scheduling by text, and eleven hours on the tools that only you can do, because you are the one who can look at a job and price it right. That is not one problem. The first eleven hours have an automation answer. The last eleven need a person, or need you to stop doing something else.

Do the log before anything else. It takes a week, costs nothing, and it is the one part of this decision where guessing gets expensive.

What AI is genuinely good at, and where it stops

It is good at work that follows a pattern and ends in text or data. Give it your own examples and a clear instruction and you get a decent first version of anything you have done before. Where it earns its keep in a small business:

  • Drafting the thing you write over and over. Quotes, enquiry replies, listing copy, the same monthly update to a client.
  • Sorting and triage. Turning a full inbox or a list of missed calls into a shortlist of what needs you today.
  • Pulling details out of messages. An email or a photographed docket becoming the four fields your spreadsheet or diary actually needs.
  • Summarising. A fourteen-message thread or a month of reviews, cut down to what was decided and what is still open.
  • Chasing on a schedule. The polite second nudge on a quote or an unpaid invoice, drafted and flagged the day it is due.

Where it stops is anything needing a body in a room or somebody carrying the can. It will not answer the door, hold the other end of a ladder, or calm a customer who has decided to be annoyed at a human. Nor will it notice that this job is unusual and should be priced differently, because it has no stake in whether the work is right.

When to hire rather than automate

Hire when what you are short of is hands, or somebody who owns an outcome. No software puts a second van on the road or a second person behind the counter at eleven on a Saturday.

Two other cases where a person is the right answer. The first is when nobody can describe the process. Automation needs a rule, even a rough one, and if the honest description of a task is “it depends, I just know”, then it is judgement, and judgement gets hired rather than configured. The second is when the work burying you is the work that makes the money. Automating around the wrong bottleneck feels productive and changes nothing. If the constraint is that only you can price a job, the fix is training somebody to price jobs.

A third answer worth naming: neither. Some of what fills the week is work the business should stop doing, like a report nobody reads or a form that exists because one customer asked for it once. Deleting that beats automating it and beats paying somebody to do it.

Can AI work instead of a new hire?

Sometimes, though it more often moves the hire than removes it. That is worth being straight about, because the version sold online (hire nobody, the robots handle it) does not survive a real week.

What automation usually does is take hours off the edges of several jobs at once. An hour here from quoting, forty minutes there from scheduling. Those hours are real, but they are scattered, so they do not add up to a vacancy unless you deliberately gather them into one. If you were about to hire an administrator mainly to handle the inbox, the quotes and the diary, that hire might become part-time, or wait until the autumn. If you were hiring a fitter, it stays a fitter.

The likelier outcome is that the hire you eventually make is a better one. The repetitive part is already handled, so you are recruiting for the judgement you actually needed instead of a job description that is two-thirds copy and paste. Which of those scattered hours are worth going after first is the question I dug into in is AI actually worth it for a small business?

What each one costs, and what it commits you to

A hire is a standing commitment. Past the salary line there is the employer’s share, the recruiting, the weeks before somebody is properly useful, and the management time that never makes it into the sum. It is also the more adaptable of the two: a person can be pointed at a different problem in March, and software mostly cannot.

Automation is front-loaded and small after that: some setup effort, then a modest monthly cost for the tools. It does not ring in sick. It also does not improve on its own, and it needs an owner inside the business, or it quietly rots the day a supplier changes a form. Compare hours against hours rather than a salary against a subscription, because the subscription is never the real cost. Setting it up is. There is a fuller breakdown of that side in how much AI automation costs in Ireland.

A way to decide this month

  1. Log a week in half-hours. Do not tidy it up as you go.
  2. Mark every block that repeats in roughly the same shape. That is your automation pile.
  3. Take the single biggest block in that pile, and only that one. Set it up properly and leave the rest alone.
  4. Read the log again a month later. If the repeating pile has genuinely shrunk, do the next block.
  5. If what is left is still full and all of it needs judgement, you have your answer. Hire, and hire for that.

Most owners I talk to have both problems at once, which is why the question feels harder than it is. It gets easier once the week is on paper and the two piles stop arguing with each other.

If you would rather not spend a month working out which pile is which, that is what the AI Assessment is for: about forty-five minutes on how your week actually runs, then a plain plan for what to automate, what to drop, and what needs a person. Either way, do the week log first. It is the cheapest decision on this page, and it usually answers the question before I do.

Book the AI Assessment

The assessment is credited in full if you go on to a build, and if I can’t find you five hours a week you pay nothing. See how it works.