Will AI replace my staff?

For most small businesses, no. AI takes over the repetitive parts of jobs, drafting, data entry, chasing, admin, not whole roles. The owners getting this right use it to free good people for work that needs judgement, and they say so out loud, because silence is what spooks a team.

Last updated 11 June 2026

You have got good people and you do not want them thinking the new AI tool is step one of a redundancy plan. You have also watched that same tool draft in thirty seconds an email that takes someone in your office twenty minutes. Pretending you did not notice is not a plan either.

Here is the honest picture.

What does AI actually take over?

Tasks, not jobs. AI is very good at the repeatable chunk of a role: the first draft, the re-keyed data, the chasing email, the meeting notes, the Monday report that is the same every Monday. It is poor at the rest. Judgement, relationships, knowing this customer needs a phone call not an email, spotting that something is off before it costs money.

The best large-scale evidence says the same. The IMF's 2024 analysis found around 60 per cent of jobs in advanced economies are exposed to AI, and that for roughly half of those the likely effect is AI helping with tasks while the person does more of the work only a person can do. For the other half the risk is real, but it sits in roles that are almost entirely routine. Small-business roles rarely are. Your bookkeeper also fields supplier queries. Your admin person also talks down annoyed customers. In a five-to-thirty person firm every role is a bundle, and AI strips out the routine layer of the bundle rather than the person holding it.

I build AI systems for small firms for a living, and no owner has ever asked me to build one that gets rid of a person. Every one of them has asked me to stop good people losing hours to work a machine should be doing.

How do I bring it in without spooking the team?

This is the bit that decides whether it works, and almost nobody writes about it.

Say the quiet part first. Before any tool arrives, tell them straight: this is here to get rid of the boring work, not the people doing it. If your headcount plans have not changed, say that too. Whatever you leave unsaid, the team will fill in with the worst case.

Let the team pick the first target. Ask each person which task they would pay never to do again, then automate that one. The person who hates re-keying invoices becomes your champion instead of your critic.

Put people in charge of the AI, not in competition with it. Whoever did the task should review and approve what the AI produces. Their role gains "runs and improves the system". It does not just lose "types the thing".

Never benchmark someone against the tool behind their back. A quiet pilot to find out whether the software is cheaper than the person is the fastest way to lose the whole room, because that story always gets out.

What should I not promise?

Do not promise that nobody will ever lose their job. You cannot promise that about anything, AI or no AI, and your team knows it. Promise the things you control: nobody gets blindsided, the boring work goes first, and anyone whose role shrinks gets first refusal on the work that grows. That is a promise you can keep, so it is the only one worth making.

If the reason you are looking at AI is that your team is buried in manual routine, start by working out which processes have outgrown the spreadsheet stage. Our guide From spreadsheet to system walks through how to spot them.

Answered by Dean Cookson, Founder and CEO at Operosus.

Now try it yourself

Most answers on this page come with a do-it-yourself route, and we would honestly rather you took it. Get stuck and the weekly Cook-a-Long is free. Decide it is not worth your hours, and that is what we are for.

Book a call. 30 minutes, no pitch deck.