Should you hire someone or automate?

It always arrives at the same moment: the work has grown, the team cannot keep up, and the two obvious answers are “put someone on it” or “automate it”.

The comparison is usually made on the wrong basis — cost. Automation is not a cheaper employee. It is a different kind of work, with different strengths and different failure modes.

The test that matters

One question: does the work follow rules you could write down?

If you can describe what goes in, what comes out and what happens to the exceptions — it can be automated. If the answer is “it depends, we look at each case”, then not yet.

This is not about complexity. It is about repeatability.

Automate if

The same task happens dozens or hundreds of times in the same way
Inputs and outputs are clear: this file arrives, this record is created
The errors are mechanical — mistyped, skipped step, entered twice
Volume swings sharply: seasons, sales, the tourist period
Nobody on the team wants to do it, and they are right

Hire if

The work needs judgement: negotiation, assessment, an exception nobody planned for
The core of it is a relationship — sales, complex support, partners
You cannot describe the rules yet, because you are still discovering them
The volume is small. Two hours a week is not worth automating
You need someone who will ask why this work exists, not just do it

The step people skip

The most common failure is not the wrong choice. It is the right choice at the wrong time.

You cannot automate what you cannot describe. If the process lives in one person’s head, the first step is not software — it is having someone do it by hand for a while and write down what actually happens, exceptions included.

Two things usually follow: you discover the process is messier than you thought, and that half of it is not needed at all.

The mistake that costs

Automating the mess.

If the process is broken, automation does not fix it — it breaks it faster and at greater scale. Wrong prices reach 500 customers instead of 5. Bad orders ship without anyone seeing them.

Fix the process first. Automate it second.

What automation actually costs

It is not just the build. It is three things people forget:

The exceptions. 80% of cases are easy. The remaining 20% is half the work.
The maintenance. A supplier changes, the invoicing software changes, the law changes. The automation has to change with them.
An owner. Someone has to notice when it stops working. An automation that fails silently is worse than no automation.

If nobody can take on the third, do not start.

Usually the answer is “both”

In practice the right shape is rarely either/or.

Automation takes the repetitive part — entries, checks, notifications, standard replies. People take the exceptions and anything needing judgement. The team does not shrink; its work changes.

That is also the realistic expectation: you rarely “replace a role”. More often you postpone the next hire, or free up people you already have from work that grinds them down.

If what you are describing is not something that answers but something that acts — records, checks, updates other systems — you are talking about an AI agent, and it is worth understanding what that is first.

A quick test before you decide

For one week, note every repeating task and how long it took. Then ask, of the three biggest:

Could I write this as steps on a single page?
How often is there an exception?
What happens if it goes wrong and nobody notices for a day?

If the answers are “yes”, “rarely” and “nothing serious” — you have an automation candidate. Otherwise, hire.

Want to look at your actual numbers? Book a free 30-minute call — if hiring is the right answer, we will say so. See also how we build AI agents or how we work as consultants.