Will AI replace workers?

The short answer: AI doesn't replace workers, it replaces tasks

What is really at risk is not a whole profession but one kind of job: a job that is nothing but one task, repeated, where it is easy to check the work was done right and the person doing it carries no responsibility for the result. Even that is not really a risk. Take that task away from the person who was doing it, and they are free to do other work. Very few people have a job made up only of tasks like that.

Four examples from a much longer list. These are the professions people ask about most:

• Programmers: writing code gets much faster, and what stays hard is deciding what to build and being accountable when something goes wrong.
• Accountants: classification, reconciliations and drafts move to the machine, but the signature stays with the licensed accountant, and so does the responsibility.
• Lawyers: finding sources, basic comments on a document, even the first draft of a contract are already done differently, but the advice, the negotiation and the professional responsibility stay with the lawyer.
• Doctors: the tools help read scans and data, but the decision and the conversation with the patient stay with the doctor.

Different ways to use AI

When I meet a client, I talk about three different kinds of benefit an organisation can get from AI tools.

1. Efficiency. Using AI to do existing tasks faster or better. For example: editing documents, collecting invoices, building presentations.
2. Augmentation. Giving an existing employee abilities they did not have before. For example: a salesperson who can make their own video, tailored to a client. Or a bookkeeper who can build herself an accounting dashboard she finds easy to use, without bringing in outside help.
3. Innovation. Using AI to change the way the business itself works. For example: apps now cost less to build, so a business can build one that plays like a game and lets customers see and feel the problem instead of reading a PDF. Or a quoting system where the client changes the requirements in real time and sees the price of the quote change straight away.

Sometimes one AI tool is efficiency and innovation at the same time. Say an expert decides to build a system that gradually learns the experience they have built up over the years, and then applies it to raw material as if they were there themselves.

In my AI implementation work, I sit down with clients and sort the processes their business runs into the first two categories, then we brainstorm the third. Those meetings are always fascinating.

What work actually moves to the machine

Three conditions, and all three have to hold:

The work repeats. Not once a quarter, but every day or every week.

It can be defined. Someone could write down on one page what goes in, what comes out and what counts as a good answer.

The result can be checked. There is a way to tell the machine got it wrong before the mistake reaches a customer.

If any one of the three is missing, it is not ready.

The third one above all: work you cannot check is work you cannot hand over (or improve, which is the next level of using AI).

What actually happens in organisations

The pattern I see is not layoffs. It is a shift in the work.

Even with the same number of people, the hours that used to go on gathering data, retyping, first drafts and recurring reports simply stop being wasted. What happens next depends on management: either people do more of the same, or they get to the things they never had time for.

The bottleneck does not disappear. It moves. Once the preparation stops taking three days, it turns out the real delay was always the decision, and the machine does not touch that. How many times have you done the work, emailed the document and then simply waited to hear back?

AI shifts the bottleneck onto people. That is not necessarily a bad thing, and it is happening in a lot of fields.

Will AI replace programmers

This is the question people ask most, and it is where the change is biggest and least understood.

Writing code has become much faster, and once again the bottleneck has moved somewhere else: to deciding what to build, making sure it works, maintaining what has been built, and being accountable when something breaks in production. A developer whose job was turning a spec into code feels this sharply. A developer who owns a system and answers for it feels it less. Concerns like security have moved much higher up the list, because the person overseeing the work no longer sees every line of the code being written.

Will AI replace accountants

The repetitive parts, such as classification, reconciliations, pulling data out of reports and preparing drafts, are moving to the machine fast.

What does not move is the signature. Anything the regulator requires someone to sign, and take professional responsibility for, stays with the person responsible, even when the machine prepared most of the material. What changes is how much grunt work it takes to reach the moment of signing.

Will AI replace lawyers

Much the same. Finding sources, comparing versions, the first draft of a standard contract, summarising long documents: all of this is already done differently.

Advice, negotiation, legal strategy and professional responsibility stay. A lawyer is not judged by how fast they draft, but by what they advise.

Will AI replace doctors

Early in the AI revolution, people were saying that radiology, the profession that reads X-rays, would disappear within a few years. It did not happen. Quite the opposite: more radiologists are needed today. Why? Once again, the basic reading has become simple and very fast, but making the decisions, and being responsible for them, has become the bottleneck. What is clear is that the tools keep getting more effective.

The clinical decision, the conversation with the patient and the responsibility stay human, and not only for regulatory reasons.

Who is really at risk

Not a whole profession. A job that is one repeated task, with output that is easy to check and no responsibility for the result. These are the same jobs that were easy to outsource even before AI.

The bigger risk is a different one, and people talk about it less: not being replaced by a machine, but working next to someone who uses one well and gets three times as much done.

What a manager should do now

Do not lay anyone off on the strength of guesses or promised efficiency gains. Letting people go before the system works is the quickest way to find out that it does not.

Pick one process that everyone in the company complains about, break it down into tasks, and check which of them meet the three conditions above. Start there. Measure how much time it really saved, not how much it was supposed to save.

Then decide what people do with the time. That is a management decision, not a technical one, and it is the one that decides whether the implementation was worth it.

The bottom line

The organisations that get hurt will not be the ones that were late to adopt AI. They will be the ones that adopted it without a clear, sensible plan that sets out what gets done, how the result is measured, and exactly what happens to the time it frees up.

---

Want to apply this to your own processes? A jumpstart meeting, ninety minutes, free: one real process of yours, and which parts of it are worth handing to a machine.