AI replaces tasks, not jobs

Meta is reportedly cutting around 8,000 roles as part of a broader efficiency push linked to major AI infrastructure investment. Microsoft has opened its first voluntary buyout in its 51-year history. Amazon announced 16,000 cuts as it pushes further into AI-driven efficiency.

The language around these moves tends to follow a familiar pattern: AI has changed the maths on headcount.

There is some truth in that. Tasks that once took hours can now be completed in minutes. Work that used to require teams of people can increasingly be handled by agentic apps, workflows and decision systems.

But there’s a risk in accepting the headline version too easily.

Calling a structural cost-out an "AI-driven workforce transformation" is cleaner for a board deck than admitting a business got bloated during a cheap-money decade and is now correcting.

(Salesforce’s CEO has been fairly direct about this: blaming AI for layoffs, he suggested, is "the lazy way out.")

The future of work is genuinely changing. Just not in the simple way the layoff headlines suggest.

AI replaces tasks. Businesses replace jobs.

A task is something you do. A job is what you are responsible for.

AI is getting very good at the first. It can draft, summarise, reconcile, classify, analyse, schedule, check, triage and produce the first version of things. It can absorb a growing share of the repetitive production work that fills the day.

But responsibility still belongs to people.

Someone still has to decide what matters. Someone still has to understand the customer, judge the trade-offs, set the standard, catch when something is wrong and own the outcome.

That means job cuts are not a technological inevitability. They are a business choice.

When agents take on time consuming tasks, the best result is not automatically fewer people. It can be better work: people freed from repetitive production so they can focus on the more valuable parts of their roles.

The strategic question is whether businesses use AI to hollow out jobs or redesign them around higher value human contribution.

We’ve been here before, just not at this speed.

This isn’t the first time technology has rewritten what work looks like.

When internet booking arrived in the late 1990s, travel agents were assumed finished. Agents who spent their days searching inventory and booking point-to-point flights found that work vanish almost overnight. Employment in the sector halved.

But travel agents who survived stopped doing what a website could do and started doing what a website couldn't: curating complex trips, managing disruptions, applying judgment. By 2025, their wages had reached parity with the broader private sector, outpacing the wider economy over 25 years.

The task-heavy version of the job died. The judgment-heavy version thrived.

Spreadsheets replaced rooms full of human calculators. Word processors ended the typing pool. In every case, the work that disappeared was most tightly bound to the tool that replaced it. The work that remained required something the tool couldn't supply.

AI is the same pattern, an order of magnitude faster and broader. It can automate work activities absorbing 60 to 70 percent of employees' time. The speed is new. The shape of the shift isn’t.

What’s actually changing?

A job is the set of outcomes someone is responsible for. A task is one of the things they do to get there. AI changes the work inside the job before it changes the job itself.

Take trust and safety. The job is keeping a platform safe. That responsibility has not disappeared. Someone is still accountable for the integrity of the system, the consistency of enforcement and the consequences of getting it wrong.

What has changed is the work underneath.

Tagging content, clearing queues, classifying obvious policy violations and identifying high-volume patterns are increasingly handled by AI.

The human work moves towards the cases that need judgment: ambiguity, context, escalation, edge cases and decisions where the cost of being wrong is high.

No one ever wanted spam-tagging to be their career.

You can see the same shift across knowledge work:

• Engineers write less boilerplate.
• Analysts spend less time producing first drafts.
• Lawyers spend less time on discovery.
• Marketers spend less time generating variants.
• Customer support teams spend less time answering the same question hundreds of times.

The roles do not simply vanish. The work inside them changes. The question is whether the business changes with it.

AI creates capacity. What happens next is a management decision.

When agents absorb the task layer, they create capacity.

That capacity can be used in two very different ways.

A business can treat it as a labour reduction opportunity. Fewer people, more output, lower cost. Sometimes that will be the right commercial decision. Some roles really will shrink. Some jobs were built almost entirely around tasks that AI can now do faster, cheaper and well enough.

But that’s not the only path.

A business can also use the capacity to redesign work. It can ask what people should now spend more time doing because the repeatable production layer no longer consumes so much of the week.

Ace Hardware, a 5,200-store US hardware cooperative whose brand promise is staff who can answer any question a DIYer throws at them, just showed what that looks like in practice.

The company rolled out an AI layer called Hey ARMA across 2,300 stores in February — not to reduce headcount, but to solve a specific human problem: store associates were avoiding customers because they feared being asked something they didn't know.

Hey ARMA gives them instant product knowledge and inventory data mid-conversation. More than 55,000 queries later, the result isn't a leaner workforce. It's a more confident one. As CEO John Venhuizen put it: "We'll never have customers who love us because of great technology. Never. It's got to be people."

That means more time with customers. More time improving quality. More time on product thinking, sales strategy, risk, service design, compliance, relationship management or operational improvement. More time on the parts of work that were always valuable but often squeezed out by the volume of routine tasks.

This is the more interesting version of AI at work.

Not a smaller workforce doing more with less, but a better designed workforce doing more of what actually matters.

The value of human work is shifting

When AI handles more of the routine work, the human layer becomes more important, not less.

What grows in value is judgment.

A new economic argument suggests AI won’t simply wipe out work, but will shift value toward the parts of the economy where human presence, provenance, care, taste, and relationship matter most.

Knowing what question to ask. Catching when an output is wrong. Understanding a customer’s real problem. Knowing when to ship and when to keep digging. Seeing the risk in a technically correct answer. Holding the context that lives inside a business because nobody ever wrote it down.

The risk isn’t just job loss. It’s hollow work.

There’s another risk that gets less attention.

If businesses use AI only to cut the visible task layer, they may leave behind hollow jobs: titles that still exist, but with less substance, less learning and less clear value.

The person is still there. The role is still there. But the work that gave the role its shape has been removed.

This matters most in the middle of organisations.

Senior people often have accumulated judgment. AI can now do much of the junior production work. But the middle layer, where people used to build capability through repeated exposure to real work, can start to disappear.

That creates a serious question for any business thinking beyond the next reporting cycle:

How will people develop judgment if the work that used to train judgment has been automated?

The answer can’t be to preserve busywork for its own sake. Nobody needs humans spending hours on work an agent can do well in minutes.

But businesses do need to replace accidental apprenticeship with deliberate development. They need to design new ways for people to learn, review, challenge, interpret, decide and take responsibility.

Otherwise they may become more efficient in the short term while quietly weakening their future capability.

What getting this right looks like

The future of work is a redesign story. The shape coming into focus is leaner, but not simply smaller.

Teams may become smaller. Jobs may become bigger. AI will handle more of the doing, while people handle more of the deciding, interpreting and owning.

For some roles, that shift will be energising. People will finally be freed from repetitive production and able to focus on the work that made them valuable in the first place.

For other roles, the transition will be harder. Some jobs were built heavily around the task layer. Some workers were rewarded for consistency, speed and reliability in work that AI can now do extremely well. Businesses will need to be honest about that too.

But the answer should be reasoned, not guessed.

The question is not just “where can we cut?”

It’s:

• Where is our actual intelligence sitting?
• Which tasks are consuming time without adding much value?
• Which decisions still need human responsibility?
• What work could our people do better if agents handled the production underneath?
• What kind of organisation do we become if we redesign work around that?

The headlines are calling this a layoff story. In some companies, it is.

But for the businesses paying closer attention, it’s also a redesign story.