Short answer: AI takes over tasks, not jobs. There's almost no job that's a hundred percent repetitive. Typically, 20-40% of one person's day gets automated, and the rest, where it takes judgment, negotiation or handling exceptions, stays with a person. But there's another side I won't hide: if a role is almost entirely moving data from one place to another, the impact on that role is real.
I hear this question on almost every call, more often than the question about price. And it's usually not asked by a worried employee, but by a manager who doesn't want to start with a lie to their team.
Why “will it replace a person” is the wrong question
When we think about work, we think about job titles: office administrator, bookkeeper, account manager. But what gets automated isn't the title, it's a task. And any job's day is made of dozens of different tasks, only some of which are recipes.
Take an office administrator. In a day she answers calls, books appointments, retypes data from emails into the system, hunts for documents, calms down an angry customer, decides what to do with an unusual case, and handles twenty other small things. Retyping from emails into the system might take three hours a day. Those three hours are what we're talking about, not the administrator herself.
What gets automated isn't the job, it's a few hours of it. The only question is what you'll do with those hours.
Where the impact is real
This is where most vendors go quiet. I won't, because you'll see it for yourself anyway.
If someone's job is ninety percent one repetitive action, like entering invoices, moving orders from email into the system, or copying data between two programs, then that role, as it exists today, won't be needed after automation. There's no point lying about it.
But in practice, a company almost never has just one role like that. More often it looks like this: three people, each spending half a day on that kind of work. After automation, none of them goes anywhere. You simply get a full-time job and a half's worth of time that didn't exist before.
What companies actually do when time frees up
From what I see, layoffs are rare. Not out of kindness, but because of simple arithmetic: in Lithuania it's hard to find people, and a trained employee who knows your customers and how things work is worth more than their hourly rate.
Most of the time, the freed-up hours go to three places:
- Work there was never time for. Calling back old customers, cleaning up the price list, following up with people who didn't buy. Work that earns money but always gets pushed back.
- Quality where it used to slip. When one person enters a hundred rows a day, mistakes are inevitable. When they oversee how the system does it and check the exceptions, there are fewer mistakes.
- Growth without new hires. The most common real outcome: orders double and the team stays the same. Formally, nobody was replaced, but no new hires were needed either.
Why layoffs usually don't pay off
If we're going purely by the numbers, let's run them all the way through. Say automation frees up 60 hours a month, about a third of a full-time job. You save about $2,200 a month (the calculator and exact numbers are in the ROI article).
Now the other side. Finding and onboarding a new employee costs months of productivity. A person who knows your customers, knows why one supplier gets treated differently, and can calm down an angry buyer: that value isn't on any invoice, but you see it the day they're gone.
So here's a simple rule: if a project only adds up once you let someone go, the project is badly planned. Automation has to pay for itself from the hours that go back into work, not from the jobs that disappear.
A separate question is not firing but not hiring: when work keeps growing and you're considering a new position. I wrote that comparison up in numbers here: AI agent or new hire.
How to tell your team
The worst option is to stay quiet until people notice the new system on their own. Then they'll make up the story themselves, and it'll be worse than the truth.
- 1. Tell them before they find out. Before the rollout, not after. The message is simple: we're handing off this specific task, because it takes X hours a week.
- 2. Say what isn't changing. Fear comes from uncertainty, not from technology. “We're not cutting jobs”, said out loud, works better than any presentation.
- 3. Involve the people who do the work. They know the exceptions you don't, and without them the system will make mistakes. Being part of building it turns a threat into their own project, and that's the strongest thing you can do.
What not to do
Three mistakes I see most often:
- Calling it “improving efficiency”. Your team translates that phrase correctly: there will soon be fewer people. Talk about the specific task, not about efficiency.
- Promising that nothing will change. When it does, you'll lose their trust. Say what will change and why it's better for them.
- Starting with the most visible person's work. The first project should be the one nobody likes: entering invoices, stitching reports together. How to pick it, I wrote up separately.
Questions about AI and employees
Will AI replace employees?
AI takes over tasks, not jobs. There's almost no job that's 100% repetitive: typically 20-40% of one person's day gets automated, and the rest, which needs judgment, negotiation or handling exceptions, stays with a person. The impact is real where a role is almost entirely moving data from one place to another.
Which jobs does AI affect the most?
Roles made up mostly of work that can be described with rules: data entry, retyping documents, answering the same requests, stitching reports together from several systems. Where it takes negotiation, empathy, accountability for a decision or judging a rare exception, the impact is small.
Do I have to lay people off for automation to pay off?
No. Automation pays off from freed-up hours that go back into work there was never time for: winning back customers, quality, sales. If the numbers only work once you let someone go, the project is badly planned.
How do I tell my team we're bringing in AI?
Tell them before they find out on their own, and be specific: which task is being handed off, why that one, and what isn't changing. Involve the people who do that work: they know the exceptions best, and being part of building it turns a threat into their own project.