When companies cut jobs for AI, what are they really cutting?

When a company cuts a job, it’s not just removing a salary. It may be removing context, judgement, customer memory, exception handling and the invisible connective tissue that keeps the business working.

That’s the part of the AI labour story that rarely fits neatly into a cost reduction slide.

When large companies see others cutting headcount and ask, “Do we need all our people?”, it’s like asking whether the body really needs its pancreas or gallbladder.

You can remove something and see no immediate effect. The system keeps moving. The business still opens on Monday. Customers still call. Reports still get generated. Revenue may not dip straight away.

Then, later, the symptoms appear.

Handovers fail. Customers drift. Decisions slow down. Workarounds multiply. Nobody remembers why a process exists. The person who knew how to fix the exception is gone. The system still functions, but something vital has been removed.

These are the disastrous realities of cutting before understanding.

A job is more than a line item

A job looks simple from a distance.

It has a title, a salary, a manager, a function and a place in the org chart. It can be added to a spreadsheet. It can be benchmarked, rationalised, restructured or removed.

But a job is not one thing.

It’s a bundle of tasks, workflows, relationships, judgement calls and tacit knowledge. Some of that work is visible. Much of it is not.

Inside one role might be the person who updates the system, handles the exception, remembers the customer history, notices when something feels wrong, translates between two teams, fixes the broken handover and quietly prevents the same problem from happening twice.

On paper, they may look like headcount. In practice, they may be part of the organisation’s nervous system.

This is why AI-driven job cuts can be so risky. Not because jobs should never change. They will. Not because every role is sacred. It’s not. But because too many organisations do not fully understand what they are cutting.

They understand the cost of the person.

They may not understand the function of the person.

The output is not always the work

AI is very good at reproducing visible outputs.

It can draft a report, summarise a meeting, answer a support query, generate a campaign, classify a request, write a follow-up and update a record.

That creates a powerful temptation.

If AI can produce the output, why keep the person who produced it?

But the output isn’t always the work.

The report may be the artefact, but the work may be knowing which numbers matter.

The support response may be the artefact, but the work may be recognising that the customer’s real problem is not the one they described.

The campaign may be the artefact, but the work may be understanding timing, taste, market mood and commercial risk.

The system update may be the artefact, but the work may be knowing that this particular case should not follow the normal process.

When leaders confuse the visible output with the whole job, they can make bad decisions quickly.

They remove the role because the output looks automatable, then discover that the value was sitting around the output: in the judgement, context and informal knowledge that made the output useful.

AI exposes the task structure

The smarter way to think about AI and work is not job replacement.

It’s task exposure.

AI forces businesses to look beneath the job title and ask what work is actually being done.

• Which tasks are repetitive?
• Which tasks are judgement heavy?
• Which tasks depend on customer knowledge?
• Which tasks exist because old systems force humans to move information manually?
• Which tasks can be automated safely?
• Which tasks should be assisted, but not handed over completely?
• Which tasks protect value in ways the business has never properly measured?

This is the level of understanding companies need before they start cutting.

Not because every job should be preserved, but because the decision needs to be precise.

Some tasks should be automated. Some should be redesigned. Some should be elevated. Some should disappear. Some should stay human because the value isn’t just speed or cost, but judgement.

The danger comes when the whole job is cut because some of its tasks can now be done by AI.

That’s like removing an organ because one visible function appears redundant.

What companies should ask before cutting

Before cutting a role in the name of AI, leaders should be able to answer a few basic questions:

• What work does this person actually do?
• Which parts of that work are visible in systems and which parts are informal?
• What decisions do they make that are not written into the process?
• What exceptions do they handle?
• What customer, operational or organisational memory do they hold?
• Who relies on them to keep work moving?
• What would break if they disappeared tomorrow?
• What could AI genuinely do better?
• What should AI support, but not own?
• What needs to be codified before the role changes?

These are not sentimental questions. They’re operating questions.

Because the real risk isn’t that AI automates work. The real risk is that companies automate the part they can see and cut the part they didn’t know they depended on.

Cost cutting v job redesign

There’s a difference between reducing headcount and redesigning work.

Reducing headcount is easy to announce. It has a number. It looks decisive. It tells a clean story about efficiency.

Redesigning work is harder.

It means understanding the task architecture of the business. It means separating routine work from judgement work. It means identifying which knowledge should be codified, which processes should be rebuilt and which human capabilities become more valuable because AI can now take on the surrounding labour.

That’s slower than a cost-cutting announcement.

It’s also more likely to produce a business that actually works.

The companies that get this right will not ask only, “How many jobs can AI replace?”

They will ask, “What work are we trying to preserve, improve or redesign?”

That question changes the conversation.

It turns AI from a blunt cost-cutting instrument into a way to understand the business more clearly.

Cut carefully

There will be job cuts. There will be role redesign. There will be work that no longer makes sense in its current form.

Pretending otherwise is not useful.

But cutting first and understanding later is a dangerous bet.

It assumes the job description reflects the real work. It assumes the visible output is the value. It assumes AI can replace the person because it can reproduce some of the tasks.

Sometimes that will be true. Often, it will be incomplete.

When companies cut jobs for AI, they may be cutting cost. They may also be cutting context, judgement, memory and the quiet organisational functions that keep the business alive.

Nothing may happen at first. That doesn’t mean nothing was lost.

AI doesn’t remove the need to understand work.

It makes that understanding urgent.