Will AI Take My Job—or Change It?

For many people, the most personal question about artificial intelligence is not whether AI is conscious.

It is not whether AGI will arrive.

It is not even whether the latest model is more capable than the one released six months ago.

It is much simpler:

Is this going to take my job?

That fear deserves a serious answer.

I don't think the responsible answer is:

“Don't worry. AI is just a tool.”

Some jobs will be disrupted.

Some tasks that people are currently paid to perform will increasingly be automated.

Some positions may disappear.

Some new positions will appear.

And many existing jobs will survive while becoming substantially different.

The difficulty is that all of those things can be true at the same time.

So I think the better question is not merely:

“Will AI take my job?”

It is:

“What parts of my job can AI perform, what parts still require me, and what happens to the job when that balance changes?”

That is a much more useful place to begin.

Jobs Are Made of Tasks

We often talk about occupations as though each one were a single activity.

An accountant accounts.

A programmer programs.

A teacher teaches.

A system administrator administers systems.

A lawyer practices law.

But real jobs are collections of very different tasks.

A manager might:

• write reports;
• analyze numbers;
• schedule meetings;
• coach employees;
• resolve conflict;
• approve spending;
• make hiring decisions;
• interpret policy;
• communicate with customers;
• and take responsibility when something goes wrong.

AI may become very capable at some of those things while remaining poorly suited to others.

That distinction matters.

Current labor research increasingly examines AI in terms of the tasks inside occupations rather than treating an entire occupation as simply “automatable” or “not automatable.”

Anthropic's 2026 Economic Index, for example, finds AI usage appearing across meaningful portions of the task sets associated with many occupations. But its research also shows that task coverage differs from effective performance, and that the implications depend on which particular tasks AI can successfully perform and how important those tasks are to the occupation.

That means:

AI can change a job without eliminating the job.

And sometimes changing the job can be almost as significant as eliminating it.

AI May Remove Tasks Before It Removes Occupations

Suppose AI reduces the time required to produce a routine report from three hours to fifteen minutes.

What happens?

One possibility is that the organization needs fewer people producing reports.

That is real displacement risk.

But another possibility is that the same employee now produces more reports, handles more cases, performs deeper analysis, works with more customers, or takes on responsibilities that previously did not fit into the day.

Another possibility is that the job becomes more demanding because the routine work disappears and what remains requires more judgment.

Another is that the employer redesigns the position completely.

These outcomes are different.

And current evidence suggests we should expect a mixture of them.

Anthropic's March 2026 Economic Index found collaborative, augmentative AI use continuing alongside automation, while its broader research warns that movement of work into production API systems may indicate more substantial workplace transformation in some occupations.

That is why statements such as:

“AI will replace accountants.”

or:

“AI will never replace accountants.”

usually tell us less than we think.

A better question is:

Which accounting tasks are becoming automatable, and what happens to the accountant's role afterward?

Ask that question about your own profession.

Some Workers Really Will Be Displaced

There is another extreme I want to avoid.

Sometimes people who support AI are so determined not to sound alarmist that they minimize legitimate economic disruption.

I don't think that is responsible either.

If a company discovers that one worker using AI can perform work that previously required several workers, there can be real employment consequences.

If an entire category of routine work becomes dramatically cheaper to automate, some positions may no longer make economic sense to employers.

Workers have legitimate reasons to care about that.

Anthropic's April 2026 survey of more than 81,000 Claude users found that people in occupations with greater AI exposure reported greater concern about AI-driven job displacement. Workers reporting some of the largest AI-related speedups also expressed greater displacement concerns.

Those fears should not be dismissed as ignorance about technology.

People understand something important:

Productivity gains can benefit workers, companies, customers, or some combination of them. They do not automatically guarantee that every existing position survives.

But AI Exposure Does Not Automatically Mean Job Collapse

The reverse assumption is also dangerous.

A profession can be highly exposed to AI and still grow.

PwC's 2026 Global AI Jobs Barometer analyzed more than one billion job advertisements across six continents. Its findings do not describe a simple collapse of AI-exposed employment. Instead, PwC reports a two-track pattern in which some AI-exposed occupations are becoming more productive and increasingly demand human-intensive skills such as judgment and leadership.

That does not mean every AI-exposed worker is safe.

It means exposure and elimination are not the same thing.

Technology can automate part of an occupation while increasing demand for another part.

It can make a skilled worker more productive.

It can allow less-experienced workers to perform tasks previously reserved for experts.

It can increase the importance of the expert who must verify the result.

It can eliminate a bottleneck and create more demand downstream.

Or it can make a position easier to consolidate.

Those possibilities have very different consequences.

Your Job May Become More Human, Not Less

One of the most interesting possibilities is that AI may make some jobs more dependent on distinctly human capabilities.

If AI performs more drafting, summarizing, formatting, searching, classification, and routine analysis, what remains?

Often the answer includes:

• judgment;
• prioritization;
• leadership;
• accountability;
• communication;
• negotiation;
• trust;
• domain expertise;
• understanding context;
• dealing with exceptions;
• and making consequential decisions.

The International Labour Organization's August 2026 work on AI and workplace skills describes growing importance for higher-order cognitive skills, socioemotional capabilities, adaptability, AI literacy, and human agency as workplaces adopt AI.

PwC likewise reports that highly AI-exposed entry-level roles are increasingly requesting skills traditionally associated with more senior work, including judgment and leadership.

That creates both an opportunity and a problem.

The opportunity is obvious:

AI can free people from lower-value repetitive work.

The problem is that employers may begin expecting higher-level performance much earlier.

The entry-level worker may no longer be paid simply to produce the first draft.

They may be expected to understand whether the AI-assisted draft is any good.

The Loss of Entry-Level Tasks Matters

That deserves more attention than it sometimes receives.

Many routine tasks are not only work.

They are how people learn.

Junior employees often become experts by doing simpler work repeatedly.

They review documents.

Write basic code.

Prepare first drafts.

Handle straightforward cases.

Troubleshoot common problems.

Perform initial research.

Over time, the accumulated experience teaches them what normal looks like and helps them recognize what is abnormal.

If AI performs all of the beginner work, we have to answer another question:

How does the beginner become an expert?

That is not an argument against automation.

It is an argument for deliberately designing learning into AI-assisted workplaces.

A junior employee who never exercises judgment will not magically acquire senior judgment because the job description eventually demands it.

Organizations need to think about this.

Workers should too.

Don't Compete With AI Only at What AI Does Best

If your professional value is defined entirely by a task that AI can increasingly perform faster and more cheaply, you have a vulnerability.

That does not mean you have no future.

It means your professional strategy needs to expand.

Suppose AI becomes extremely good at creating a basic first draft.

Trying to become the fastest human producer of generic first drafts may become a losing competition.

But knowing:

• what should be drafted;
• what information matters;
• whether the draft is correct;
• where it fails;
• how it fits the real situation;
• what risk it creates;
• and whether it should be approved

may become more valuable.

The goal is not to outrun AI at being AI.

The goal is to become better at the responsibilities that AI makes more important.

Learn to Use AI Without Making Yourself Replaceable

This can sound contradictory.

If I teach AI to help me do my work, am I training my replacement?

Sometimes that risk may exist.

But refusing to learn the technology is not a strong defense either.

If AI becomes part of your profession, the worker who cannot operate within an AI-assisted environment may become less competitive regardless of whether the underlying occupation survives.

The better strategy is to learn both:

how to use the system

and

how to remain responsible for the work.

That means learning to:

• give AI clear direction;
• provide appropriate context;
• identify what should not be disclosed;
• evaluate outputs;
• verify important claims;
• recognize failure;
• understand your domain deeply enough to challenge the machine;
• and make the final decision when the responsibility is yours.

Those are not merely prompting skills.

They are professional skills in an AI-assisted workplace.

Domain Knowledge Still Matters

There is a dangerous idea that can emerge when AI becomes good at explaining almost anything:

Why learn the subject if AI already knows it?

Because without knowledge of your own, you become dependent on the output.

You may not know when the answer is incomplete.

You may not recognize an impossible recommendation.

You may fail to notice that a technically valid answer does not fit your organization.

You may accept a plausible explanation of a problem that an experienced practitioner would immediately question.

AI can reduce the amount of information you need to memorize.

It does not eliminate the value of understanding your field.

In many roles, AI may actually make real expertise more important because experts become responsible for supervising a larger volume of machine-assisted work.

The AI can produce more.

Someone still has to know whether the production is trustworthy.

Human Judgment Must Be Real

This connects directly to the argument I make throughout The Necessary Evil series.

Human involvement is not meaningful merely because a person eventually clicks Approve.

If the person does not understand the work, cannot challenge the AI, cannot stop the process, and is expected to rubber-stamp the recommendation, then the organization has preserved the appearance of human authority while removing much of its substance.

That becomes especially important in AI-assisted work.

As systems become more productive, humans may be asked to oversee more outputs.

Ten reports.

A hundred cases.

A thousand automated decisions.

At some point, “human review” can become ceremonial unless organizations preserve enough time, expertise, authority, and evidence for the review to mean something.

The future of work should not be:

AI does everything and a human accepts liability.

Human authority must remain meaningful if human accountability remains meaningful.

Your Worth Is Not Your Productivity

There is also something deeper at stake.

One of the dangers of talking about AI and employment entirely in economic terms is that we can begin treating a person's worth as though it were determined by whether a machine can perform the same task more efficiently.

That is a mistake.

A person's dignity does not decline when technology becomes more capable.

If AI can produce a report faster than you can, that does not make the AI more valuable as a person.

It is not a person.

And it does not make you less human.

Economic systems still have to make difficult decisions about employment, productivity, wages, and organization.

But we should be careful not to turn those economic decisions into statements about human worth.

Efficiency and dignity are different questions.

Book 1 of The Necessary Evil exists in large part because I believe that distinction matters.

Ask What Becomes More Important

If you are worried about your career, try a different exercise.

List the major tasks in your job.

Then ask:

Which tasks could AI probably perform now?

Be realistic.

Which tasks could AI partially assist with?

These may become dramatically faster without disappearing.

Which tasks require physical presence, trust, relationship, judgment, accountability, or organization-specific context?

These may become more important.

Which tasks are currently protected only because AI is not reliable enough yet?

Do not assume current technical limits are permanent.

If the automatable work disappeared tomorrow, what value would I still provide?

That may be the most important career question of all.

Become the Person Who Can Govern the Work

Book 4 focuses heavily on orchestration: AI operating inside a governed workflow rather than as an uncontrolled substitute for human responsibility.

That suggests another professional opportunity.

As AI performs more execution, organizations need people who can decide:

• what work AI should receive;
• what context it needs;
• what tools it may use;
• what actions it may take;
• what evidence it must produce;
• what needs verification;
• where approval is required;
• when escalation is necessary;
• and who remains accountable.

In other words, one of the important jobs in an AI-enabled workplace is not simply doing the task.

It is governing how the task gets done.

You do not need the title “AI orchestrator” for that to become part of your work.

Managers will do it.

Engineers will do it.

Analysts will do it.

Teachers will do it.

Administrators will do it.

Healthcare professionals will do it.

Small-business owners will do it.

Anyone responsible for work that AI assists will increasingly need to understand where machine execution ends and human authority begins.

The Future Will Not Be Uniform

There will not be one answer to the job question.

Some workers will use AI and become substantially more productive.

Some occupations will grow.

Some roles will shrink.

Some entry-level pathways will change.

Some tasks will disappear.

Other tasks will become more valuable.

Entirely new kinds of work will emerge.

And some people will suffer genuine disruption during that transition.

Anyone promising that AI will either destroy work universally or create prosperity universally is claiming more certainty than the evidence justifies.

What we can say is that work is already changing.

The prudent response is neither panic nor denial.

It is preparation.

Practical Takeaway

Do not ask only:

“Can AI do my job?”

Break the question apart.

Ask:

1. Which parts of my job can AI already perform?
2. Which parts can it accelerate?
3. Which parts require real human judgment, trust, expertise, or accountability?
4. Which skills become more valuable when routine work is automated?
5. Am I learning to govern AI-assisted work rather than merely compete with it?

AI may take some jobs.

It may create others.

But for a very large number of people, the immediate experience will be something in between:

AI will change the job.

The worker who understands that change has a better chance of adapting to it.

Learn the technology.

Protect your expertise.

Strengthen your judgment.

Understand the responsibilities the machine cannot meaningfully own.

And remember that your value as a human being was never determined by whether you could perform a task faster than a machine.

AI can change the work.

It does not define the worth of the worker.