AI automation vs hiring: which should come first?

Automate the repetitive share of the role first, because a new hire inherits your broken process and spends their salary doing work a system should be doing. Then hire when the work genuinely needs judgement, relationships, or creativity. The two are not rivals. Get the order right and the person you eventually bring in does far more valuable work from day one.

What does the honest cost comparison look like?

Put the two options side by side and the gap is bigger than most people expect. A junior admin or coordinator hire in the UK runs roughly £25,000 to £35,000 a year in salary alone, before you add employer’s National Insurance, pension, equipment, software seats, and the recruitment fee to get them through the door.

Then there is the cost nobody puts on the spreadsheet: your time. A new starter needs onboarding, training, and management, and that management does not stop after the first month. You are paying a salary every year, forever, for as long as the role exists.

A one-off automation build is a very different shape. A single well-scoped workflow tends to land between £5,000 and £20,000 as a one-time cost, in line with what we set out in our cost guide, plus modest upkeep. It does not need onboarding, it does not phone in sick, and it does the same task the same way at 2am as it does at 2pm. For the repetitive slice of a role, the annual salary and the one-off build are simply not in the same league.

When does hiring clearly win?

Plenty of the time, and we would rather say so plainly than pretend a script can replace a person. Automation is brilliant at work you can describe with rules. It is useless at work you cannot. Hire, do not automate, when the job is built on any of these:

• Judgement calls. Weighing messy trade-offs where the right answer is “it depends” and the context changes every time.
• Relationships. Winning a client’s trust, managing a difficult supplier, or holding a team together. People buy from people.
• Novel problems. Work that is different every time and has never been written down, so there is no pattern to hand a machine.
• Work you cannot specify. If you cannot explain the task clearly enough for another human to follow, you certainly cannot automate it. That is a sign it needs a person, not a system.

If a role is mostly these things, hire, and hire well. Trying to automate genuine human work produces something worse than doing nothing.

What is the hybrid answer most businesses actually need?

Here is the version that fits most small businesses. Very few roles are purely repetitive or purely human. They are a blend. The trap is treating the whole job as one lump and asking “person or robot?” when the smarter question is which parts of the job are which.

Automate the roughly 80% that is grunt work, the copying between systems, the chasing, the data entry, the first-draft replies, so that the human you do hire spends their day on the 20% that only a human can do. You often end up hiring anyway, just for a better job. Instead of a coordinator drowning in admin, you get someone doing the thinking, the relationships, and the exceptions, which is where the real value sits. That is a role people actually want, and it is far cheaper to run.

What is the trap of hiring to absorb a broken process?

This is the mistake we see most often. A process is painful, so the instinct is to throw a person at it. But if the process is broken, all a new hire does is absorb the pain and make it invisible. The chaos is still there, it is just being manually held together by someone whose salary you are now paying for the privilege.

Worse, the process gets harder to fix, because now it lives in a person’s daily routine and everyone is afraid to disturb it. You have made a bad system permanent. The honest move is to fix or automate the process first, then decide what human work is genuinely left. Sometimes that still means hiring. Often it means hiring for something much better than the job you were about to advertise.