THE JOBS MOST AT RISK FROM ARTIFICIAL INTELLIGENCE

Which professions could shrink, which will survive, and why the next employment revolution may be about people using AI rather than AI simply replacing people

By AI TV INFO | Global Intelligence — Special Report

EXECUTIVE SUMMARY

Artificial intelligence is moving from an experimental technology into the everyday machinery of the global economy.

Chatbots answer customer questions. AI systems transcribe conversations, translate documents, generate advertising copy, analyse financial information, write software and review legal documents. In hospitals, AI can assist with medical-image interpretation. In offices, software agents can schedule meetings, process documents and perform increasingly sophisticated administrative tasks.

This transformation has triggered a fundamental question for workers, employers and governments:

Which jobs are most likely to disappear because of AI?

The answer is more complicated than a simple list of occupations.

AI rarely eliminates an entire profession overnight. More commonly, it automates particular tasks within a job. When enough tasks can be automated, however, companies may eventually require fewer employees to produce the same amount of work.

That creates a potentially profound shift in the labour market.

A company that once needed ten employees to perform a large volume of routine digital work may eventually accomplish much of that work with a smaller human team supported by AI systems.

The greatest exposure is concentrated in occupations involving repetitive, standardized, predictable and primarily digital tasks.

By contrast, occupations requiring physical adaptability, human trust, interpersonal judgment, responsibility, leadership or work in unpredictable environments are generally more resistant to full automation.

But there is another important lesson.

The future may not belong simply to workers whose jobs are “AI-proof.” It may belong to workers who know how to use AI better than their competitors.

The central employment question could therefore change from:

“Will AI take my job?”

to:

“Will someone using AI take the job that used to require me?”

PART I — THE AI AUTOMATION DIVIDE

What makes a job vulnerable?

AI systems are particularly effective when work follows a predictable pattern:

Digital input → repeatable process → predictable digital output.

Consider data entry.

A human employee might receive hundreds of invoices, read the information, identify the supplier, extract the amount, enter the data into a database and classify the transaction.

Modern AI systems can increasingly perform much of this workflow automatically.

The same pattern appears in customer support, transcription, translation, bookkeeping, document review and routine content production.

The more standardized the workflow, the easier it becomes to automate.

Three characteristics increase exposure

1. Digital-only workflows

Jobs are more exposed when both the information entering the system and the resulting work exist digitally.

Examples include:

• Documents
• Emails
• Spreadsheets
• Databases
• Customer messages
• Audio recordings
• Digital images
• Standardized forms
• Computer code

AI does not need a physical body to perform these tasks.

2. Low ambiguity

AI performs particularly well when there is a clear procedure.

If an employee can be given a detailed instruction such as:

“Read this document, identify these five fields and enter them into these database columns,”

the task is a strong candidate for automation.

The opposite is also true.

If the employee must enter an unpredictable environment, understand competing human interests and make a judgment with incomplete information, automation becomes considerably harder.

3. Limited human interaction

Some work depends heavily on trust, empathy, persuasion, physical presence or interpersonal relationships.

These characteristics make complete automation more difficult.

That does not mean AI cannot assist those workers.

It means the human remains a critical part of the service.

PART II — THE JOBS MOST AT RISK

1. DATA ENTRY AND TRANSCRIPTION

Automation risk: 🔴 VERY HIGH

Data entry is one of the clearest examples of work vulnerable to AI-driven automation.

Historically, organizations have employed large numbers of people to transfer information from documents, forms, invoices and other sources into databases.

AI can increasingly perform:

• Document ingestion
• Optical character recognition
• Information extraction
• Classification
• Database entry
• Form processing
• Invoice processing
• Audio transcription
• Meeting transcription

Intelligent Document Processing systems can turn unstructured documents into structured information with limited human intervention.

The consequence may not be the complete disappearance of every data-entry employee.

Instead, one employee may supervise an automated system that previously required several people.

The emerging role

The human worker may increasingly become:

Data Entry Clerk → AI Workflow Supervisor

That is a recurring pattern throughout the AI economy.

2. BASIC CUSTOMER SUPPORT

Automation risk: 🔴 VERY HIGH

Customer service is entering one of the most visible phases of AI adoption.

Traditional Tier-1 support frequently involves answering repetitive questions:

• “Where is my order?”
• “How do I reset my password?”
• “What are your opening hours?”
• “How do I return this product?”
• “What is the status of my account?”
• “How do I change my booking?”

These interactions are highly structured.

Conversational AI can operate continuously, respond instantly and handle enormous volumes of simultaneous interactions.

The economic attraction for companies is obvious.

AI does not need to sleep, take lunch breaks or answer one customer at a time.

But human agents are unlikely to disappear completely.

The difficult cases remain.

Customers with unusual problems, emotional complaints, complicated financial situations or situations requiring discretionary decisions may still require human intervention.

The likely transformation is therefore:

AI handles routine cases. Humans handle exceptions.

That could dramatically reduce the number of entry-level support positions.

3. TELEMARKETING AND SCRIPTED OUTBOUND SALES

Automation risk: 🔴 VERY HIGH

Telemarketing is another occupation vulnerable to conversational AI.

Voice-based AI systems can increasingly conduct standardized conversations, qualify leads and route promising prospects.

The basic workflow can be represented as:

Call → introduce offer → ask questions → classify response → continue or terminate → record result.

This is precisely the type of predictable workflow that automation systems can target.

The more scripted the sales interaction, the greater the potential for AI substitution.

However, complex sales remain different.

Selling enterprise software, negotiating a major business contract or developing a long-term relationship with an important client requires trust, strategic judgment and relationship management.

AI may become the salesperson’s tool rather than the salesperson’s replacement.

4. ROUTINE COPYWRITING AND PROOFREADING

Automation risk: 🔴 HIGH

Generative AI has fundamentally changed the economics of producing basic written content.

AI systems can generate:

• Product descriptions
• Advertising variations
• Email campaigns
• SEO articles
• Social-media copy
• Basic newsletters
• Summaries
• Standard corporate communications
• First drafts
• Surface-level proofreading

This creates pressure particularly on work where the client does not require original reporting, distinctive expertise or a unique creative voice.

A company that once commissioned hundreds of individually written product descriptions may now generate initial versions automatically.

What becomes more valuable?

The human writer may increasingly be responsible for:

• Editorial judgment
• Original reporting
• Brand strategy
• Fact-checking
• Investigative work
• Creative direction
• Interviewing
• Narrative development
• Accountability

The writing profession may therefore become smaller in some segments while becoming more specialized in others.

5. BOOKKEEPING AND ROUTINE ACCOUNTING

Automation risk: 🔴 HIGH

Accounting contains enormous amounts of structured information.

Receipts, invoices, transactions and financial records can often be categorized according to established rules.

AI-powered accounting systems can assist with:

• Receipt recognition
• Transaction categorization
• Reconciliation
• Invoice processing
• Expense classification
• Routine reporting
• Anomaly detection
• Draft tax documentation

This puts pressure particularly on repetitive junior accounting tasks.

But accounting itself is unlikely to disappear.

Complex tax strategy, auditing, financial judgment, regulatory interpretation and advising business owners involve considerably more than processing numbers.

The profession could therefore move toward a model in which:

AI processes the transactions.

Humans interpret the consequences.

6. ADMINISTRATIVE ASSISTANTS

Automation risk: 🔴 HIGH

Administrative work has traditionally involved a large collection of small tasks:

• Scheduling meetings
• Organizing calendars
• Writing routine emails
• Preparing documents
• Booking travel
• Taking notes
• Searching for information
• Managing files
• Creating reports

AI agents are increasingly capable of combining several of these functions.

Instead of simply answering a question, an AI assistant can potentially:

receive an instruction → search information → prepare a response → schedule an appointment → update a database.

This is significant because automation becomes much more powerful when AI moves beyond individual tasks and begins coordinating entire workflows.

The administrative profession may therefore shift from task execution toward workflow management and human coordination.

7. ROUTINE TRANSLATION

Automation risk: 🔴 HIGH

Machine translation has already transformed the translation industry.

AI can translate routine material including:

• Product documentation
• Websites
• Technical manuals
• Standard correspondence
• Internal documents
• Basic marketing material

The greatest pressure falls on translations where accuracy requirements are relatively straightforward and the material is highly standardized.

But language is not simply vocabulary substitution.

Legal translation, literary translation, diplomatic communication, cultural adaptation and highly specialized technical work can require significant human expertise.

The translator of the future may increasingly function as an:

AI translation editor, specialist and quality controller.

8. BASIC GRAPHIC DESIGN

Automation risk: 🟠 HIGH

Generative AI has introduced another major disruption: the ability to create images, layouts and visual concepts from natural-language instructions.

AI can rapidly produce:

• Advertising concepts
• Social-media graphics
• Product mockups
• Simple illustrations
• Presentation visuals
• Backgrounds
• Image variations
• Basic branding concepts

This creates particular pressure on low-complexity design work.

However, design is not merely image production.

Professional designers also understand:

• Brand identity
• Audience psychology
• Visual systems
• Product strategy
• Art direction
• User experience
• Communication goals

AI may therefore eliminate some production tasks while increasing the importance of creative direction.

9. PARALEGAL AND DOCUMENT REVIEW

Automation risk: 🟠 HIGH

Law firms and corporate legal departments process enormous quantities of documents.

AI can assist with:

• Searching documents
• Identifying clauses
• Summarizing contracts
• Comparing versions
• Classifying evidence
• Finding relevant passages
• Extracting information
• Preparing preliminary research

This is particularly important in large-scale litigation and corporate transactions, where lawyers and paralegals may need to examine thousands or millions of documents.

The technology does not mean that lawyers suddenly become unnecessary.

Instead, the economics of legal work may change.

A task that previously required dozens of hours of human review could potentially be completed much faster with AI-assisted systems.

The human legal professional remains responsible for interpretation, strategy, judgment and accountability.

10. JUNIOR PROGRAMMING

Automation risk: 🟠 HIGH

Software development provides one of the most complicated examples of AI disruption.

AI coding systems can increasingly:

• Generate code
• Explain code
• Debug programs
• Write tests
• Convert code between languages
• Create prototypes
• Generate documentation
• Identify certain vulnerabilities
• Assist with software architecture

This creates particular pressure on routine junior-level programming work.

But software engineering is not simply writing lines of code.

A strong engineer must understand:

• What the customer actually needs
• System architecture
• Security
• Reliability
• Performance
• Integration
• Business constraints
• Technical tradeoffs

The likely result is not the disappearance of programming.

It is a dramatic increase in the amount of software a smaller number of engineers can produce.

That could make entry into the profession more difficult while making highly capable engineers more productive.

11. ROUTINE FINANCIAL ANALYSIS

Automation risk: 🟠 MEDIUM-HIGH

Financial professionals routinely process large amounts of structured information.

AI can help:

• Summarize financial reports
• Compare companies
• Identify trends
• Extract financial indicators
• Generate preliminary analysis
• Monitor market information
• Produce standardized reports

The more routine the analytical work, the greater the exposure.

But high-level investment decisions involve uncertainty, strategy, risk tolerance and accountability.

AI can analyze information.

It does not automatically determine what an institution should do.

That distinction is crucial.

12. RADIOLOGY AND IMAGE SCREENING

Automation risk: 🟠 MEDIUM

Medical imaging represents a different category.

AI can assist clinicians by identifying patterns in scans and highlighting potentially abnormal findings.

But medical diagnosis is not simply image recognition.

Doctors must consider:

• Patient history
• Symptoms
• Previous examinations
• Other tests
• Clinical context
• Treatment options
• Risk
• Ethical considerations

The likely future is therefore not simply:

AI replaces radiologists.

It is more likely:

Radiologists using AI outperform radiologists working without AI.

This distinction could become important across healthcare.

13. ROUTINE JOURNALISM

Automation risk: 🟠 MEDIUM-HIGH

Journalism is also undergoing transformation.

AI can produce structured stories from datasets, including certain:

• Earnings reports
• Sports results
• Weather updates
• Market summaries
• Corporate announcements
• Routine financial reporting

The technology is particularly effective when the story follows a predictable structure and the underlying data is already available.

But journalism at its strongest requires things AI cannot simply obtain from a database.

Reporters investigate.

They cultivate sources.

They ask uncomfortable questions.

They enter places where information is difficult to obtain.

They verify competing claims.

They make editorial judgments.

They can expose wrongdoing.

That means routine content may become increasingly automated while investigative, analytical and accountability journalism becomes more valuable.

PART III — THE JOBS THAT ARE HARDER TO REPLACE

AI’s limitations become clearer when work moves away from standardized digital environments.

SKILLED PHYSICAL TRADES

Examples include:

• Electricians
• Plumbers
• Mechanics
• Carpenters
• HVAC technicians
• Construction specialists

A language model can explain how to repair a pipe.

It cannot automatically walk into every 80-year-old building, identify the unexpected problem, manipulate tools in a cramped space and safely complete the repair.

The physical world is messy.

Buildings differ.

Equipment is damaged.

Parts are missing.

Instructions can be incomplete.

Humans remain remarkably adaptable in these environments.

HANDS-ON HEALTHCARE

Nurses, physical therapists, emergency medical professionals and many other healthcare workers combine knowledge with physical interaction and human relationships.

A nurse may have to:

• Assess a patient’s condition
• Respond to unexpected changes
• Move or position a patient
• Communicate with relatives
• Coordinate with clinicians
• Provide reassurance
• Make rapid judgments

AI can assist with information.

It cannot automatically replace the entire human relationship and physical environment.

TEACHERS AND EDUCATORS

Education is likely to be transformed by AI rather than eliminated.

AI can generate exercises, explanations and personalized learning material.

But teaching involves much more than delivering information.

Teachers manage classrooms, motivate students, recognize emotional difficulties, establish relationships and make decisions about how individual students learn.

The classroom is a social environment.

That makes teaching considerably harder to automate completely.

THERAPISTS AND COUNSELORS

Mental-health and counseling professions depend heavily on trust, empathy, communication and human relationships.

AI can potentially provide conversational assistance and information.

But many people seek professional support precisely because they want another human being to understand their situation and accept responsibility for the therapeutic relationship.

This makes complete replacement substantially more difficult.

EMERGENCY RESPONDERS

Firefighters, paramedics, emergency technicians and related professions operate in unpredictable environments.

An emergency rarely follows a clean script.

Conditions change.

Information is incomplete.

People behave unpredictably.

Physical action is required.

These are precisely the characteristics that make complete automation difficult.

LEADERS, NEGOTIATORS AND ENTREPRENEURS

Leadership is another area where AI may become a powerful tool rather than a straightforward replacement.

Executives and entrepreneurs must make decisions when there is no obvious correct answer.

They must balance:

• Financial considerations
• Employee interests
• Customer expectations
• Regulation
• Reputation
• Competition
• Risk
• Long-term strategy

AI can provide analysis.

Someone still has to decide.

And someone must accept responsibility for that decision.

PART IV — THE REAL DIVIDE: TASKS, NOT JOB TITLES

One of the biggest mistakes in discussions about AI and employment is treating occupations as indivisible units.

A job is actually a collection of tasks.

Consider an accountant.

Some tasks may be highly automatable:

Processing receipts → automated

Categorizing transactions → automated

Generating standard reports → automated

Other tasks are less easily automated:

Advising a business owner → human-intensive

Interpreting unusual regulations → human-intensive

Making strategic financial recommendations → human-intensive

The profession does not necessarily disappear.

Its composition changes.

This distinction is crucial.

PART V — THE “AI-ASSISTED WORKER” ECONOMY

The emerging labour market may create a new hierarchy.

Level 1 — AI-exposed worker

Performs routine tasks that AI can increasingly perform.

Level 2 — AI-assisted worker

Uses AI to complete those tasks faster.

Level 3 — AI operator

Knows how to design, supervise and integrate AI workflows.

Level 4 — AI strategist

Uses AI as part of a larger business, technical or organizational strategy.

This suggests that learning AI may become less like learning a specialized software package and more like learning basic digital literacy.

PART VI — THE PLANIVITY PARADOX

AI can produce an uncomfortable economic paradox.

Suppose a company previously required:

10 workers × 100 units of output = 1,000 units

If AI makes each worker five times more productive, the company may theoretically need:

2 workers × 500 units = 1,000 units

But another possibility exists.

The company could keep all ten workers and produce:

10 workers × 500 units = 5,000 units

Which outcome occurs?

That depends on demand, competition, investment, regulation and management decisions.

If increased productivity creates enough additional demand, employment may survive or even grow.

If companies primarily use AI to reduce labour costs, employment can decline.

This is why technological unemployment cannot be determined by technology alone.

PART VII — THE JOBS MOST LIKELY TO CHANGE FIRST

The first occupations affected by AI are likely to share several characteristics:

High-risk characteristics

1. Repetition

The same task is performed hundreds or thousands of times.

2. Standardization

There is a relatively clear definition of a correct output.

3. Digital inputs

Information exists in documents, databases, messages, images or audio.

4. Predictability

Exceptions are relatively rare.

5. Low physical complexity

The work happens primarily on a computer.

6. Limited relationship requirements

The customer does not require deep human interaction.

7. Easy quality measurement

It is relatively straightforward to determine whether the output is acceptable.

When most of these conditions are present, AI automation becomes economically attractive.

PART VIII — THE HUMAN ADVANTAGE

What remains uniquely valuable?

Not necessarily memorizing information.

Not necessarily producing the first draft.

Not necessarily processing data faster than a computer.

The human advantage is increasingly concentrated in several areas.

EMPATHY AND HUMAN TRUST

People experiencing grief, conflict, fear or major life decisions often want genuine human connection.

JUDGMENT UNDER UNCERTAINTY

The hardest decisions often occur when information is incomplete.

ACCOUNTABILITY

Someone must ultimately be responsible for important decisions.

PHYSICAL ADAPTABILITY

The real world is not a standardized database.

LEADERSHIP

Organizations require people who can persuade, coordinate and motivate other people.

ORIGINAL PROBLEM FRAMING

AI is powerful at generating solutions.

Humans still frequently determine which problem deserves solving in the first place.

AI SUPERVISION

Paradoxically, AI itself creates new work.

Someone must:

• Check outputs
• Detect hallucinations
• Evaluate accuracy
• Design workflows
• Protect sensitive information
• Monitor automated systems
• Handle exceptions
• Make final decisions

PART IX — THE NEW VALUE OF FACT-CHECKING

As AI becomes more capable of producing convincing information, the ability to determine whether information is correct becomes more valuable.

This creates an unusual economic development.

When information becomes cheap, trustworthy judgment becomes expensive.

A person who merely produces text may face increasing competition.

A person who can determine:

What is true?

What matters?

What is missing?

What is misleading?

What should we do about it?

may become substantially more valuable.

This principle is especially important in journalism.

PART X — WHAT THIS MEANS FOR YOUNG PEOPLE CHOOSING CAREERS

Young people entering the workforce face a different environment from previous generations.

Choosing a profession solely because it currently has many vacancies may not be sufficient.

A better question is:

What combination of human capability and AI capability will be difficult to replace?

Strong career characteristics may include:

• Domain expertise
• Communication
• Critical thinking
• Leadership
• Physical skills
• Creativity
• Relationship building
• Ethical judgment
• Technical literacy
• AI fluency

The strongest strategy may be combination rather than specialization.

For example:

Doctor + AI

Lawyer + AI

Engineer + AI

Teacher + AI

Journalist + AI

Electrician + AI

Entrepreneur + AI

The advantage belongs to the person who understands both the profession and the technology.

PART XI — THE MOST IMPORTANT WARNING

There is a dangerous misconception that the only workers threatened by AI are people performing low-skilled jobs.

That is not correct.

Generative AI is capable of performing portions of highly educated professional work.

Programmers, lawyers, accountants, analysts, translators, designers and journalists are all seeing portions of their workflows become automatable.

The dividing line is therefore not simply:

high education vs. low education.

It is increasingly:

routine work vs. complex work.

And even that distinction is imperfect.

AI is becoming capable of increasingly complex tasks.

The real question is how much of a profession’s total workflow can be automated while maintaining acceptable accuracy, reliability, safety and accountability.

PART XII — A NEW WAY TO THINK ABOUT “AI-PROOF” JOBS

There may be no permanently AI-proof profession.

A job that appears difficult to automate today could become much easier to automate tomorrow if robotics, AI agents or new interfaces improve.

Therefore, the strongest form of career resilience may not be choosing a supposedly untouchable occupation.

It may be developing the ability to adapt repeatedly.

Workers who can learn new tools, understand technology and acquire complementary skills may have a significant advantage as the labour market evolves.

AI TV INFO’S VERDICT

The evidence points toward transformation rather than a single moment of mass job extinction.

The occupations under greatest pressure are those built around:

repetition + standardization + digital information + predictable outputs.

That puts data entry, routine transcription, basic customer support, telemarketing, routine bookkeeping, administrative processing, basic translation and standardized content production near the front of the automation queue.

But the story does not end there.

AI will also transform professions that survive.

Doctors will use AI.

Teachers will use AI.

Lawyers will use AI.

Engineers will use AI.

Journalists will use AI.

Accountants will use AI.

Designers will use AI.

The workplace of the future is therefore unlikely to be divided simply into humans versus machines.

It is more likely to be divided between:

people who know how to work with intelligent machines — and people whose work can be performed by them.

The most important employment revolution may therefore not be the replacement of humans by AI.

It may be the replacement of AI-inexperienced workers by AI-enabled workers.

THE AI TV INFO’s TAKEAWAY

🔴 Highest exposure

• Data entry
• Transcription
• Basic customer support
• Telemarketing
• Routine bookkeeping
• Administrative processing
• Basic translation
• Routine copywriting

🟠 Significant exposure

• Basic graphic design
• Paralegal document review
• Junior programming
• Routine financial analysis
• Routine journalism
• Medical image screening

🟢 Greater resilience

• Skilled trades
• Nursing and hands-on healthcare
• Emergency response
• Teaching
• Therapy
• Complex sales
• Leadership
• Entrepreneurship
• High-stakes professional judgment

THE BOTTOM LINE

AI is not simply coming for jobs. It is coming for tasks.

The jobs containing the largest concentration of automatable tasks are likely to shrink first.

The workers who combine human judgment, specialized expertise, adaptability and AI literacy may be best positioned for what comes next.