AI and Work

AI most affects tasks that involve reading, writing, summarizing, sorting information, answering routine questions, and writing computer code, while tasks that need hands-on physical skill, in-person care, complex judgment, or deep trust are affected much less, and nearly every job is a mix of both.

Think in tasks, not jobs

Every job is a bundle of tasks. A medical receptionist greets patients, answers phones, schedules appointments, handles insurance paperwork, calms worried people, and fixes billing mistakes. AI can help with scheduling and paperwork. It can't hold a frightened patient's hand.

Here's an illustration of how AI might touch the tasks in one kind of job:

An illustration, not a measurement. Real proportions vary widely by workplace.

Tasks AI affects most

• Writing and editing: drafts, marketing copy, reports, standard letters.
• Summarizing and searching: reading long documents, research, pulling out key points.
• Routine customer questions: chat and phone support for common issues.
• Data entry and processing: moving information between forms and systems.
• Computer programming: writing, checking, and fixing code, one of the most heavily affected areas.
• Translation and transcription.
• Basic analysis: spreadsheets, charts, standard financial reports.
• Graphic design and stock imagery (Lesson 4).

A 2026 study by Anthropic, measuring how people actually use AI at work, found the most exposed occupations included computer programmers, customer service representatives, and data entry workers.

Tasks AI affects least (for now)

• Hands-on physical work in varied settings: plumbing, electrical work, construction, auto repair, landscaping (Lesson 4, Section 5 explains why robots find this hard).
• In-person care: nursing, home health aides, childcare, physical therapy.
• Complex judgment with accountability: decisions where someone must take responsibility.
• Building trust and relationships: sales to longtime clients, counseling, teaching, managing people.
• Work in unpredictable situations: emergency response, skilled trades troubleshooting.

A surprise compared to past automation

Earlier waves of automation mostly affected factory and routine manual work. This wave is different: it reaches office and professional work, including jobs that required a college degree. Meanwhile, many skilled trades are among the least affected, at least until robots improve.

So far, most jobs are changing rather than vanishing, and research hasn't found widespread job losses from AI across the economy, but hiring of young workers into AI-exposed fields has dropped noticeably, and new kinds of jobs are emerging even as some old ones shrink.

Three things that happen to jobs

What the evidence shows so far (as of 2026)

Source: Stanford Digital Economy Lab, "Canaries in the Coal Mine" update, August 2026. Similar early signs appear in other research. These studies show patterns, not proof of cause, and other factors like interest rates also affect hiring.

The key finding: the effect so far looks less like mass layoffs and more like companies hiring fewer beginners in fields where AI can do entry-level tasks. The same research found a sharp difference: where AI automates tasks, young workers' employment fell; where AI assists workers, employment held steady or grew.

Why the entry-level squeeze matters

Entry-level jobs are how people learn a trade: the junior paralegal who reviews documents becomes the experienced paralegal who knows what to look for. If AI does the beginner tasks, where does the next generation get its experience? This is one of the most important open questions about AI and work, and one that affects families across generations.

Jobs being created

The World Economic Forum's 2025 survey of employers projected that between 2025 and 2030, technology and other trends would create about 170 million jobs worldwide and displace about 92 million, a net gain of 78 million, though very unevenly across countries and skill levels. Economist David Autor and colleagues found that about 60% of U.S. workers in 2018 were in job types that didn't exist in 1940. New work tends to appear in ways that are hard to predict in advance.

Two views of what comes next

Serious economists hold both views. The honest position is uncertainty, which is itself a reason to prepare.

As AI handles more routine information work, the skills that grow in value are the most human ones (good judgment, genuine care, original thinking, trustworthiness, and skilled physical work), plus the practical skill of using AI tools well.

Five human strengths

Plus: skills for working alongside AI

• Clear communication: the same skill that makes a good prompt (Lesson 5, Section 1) makes a good manager and colleague.
• Critical thinking and verification: knowing when to trust, when to check.
• Adaptability: learning new tools as they arrive.
• Knowing your field deeply: AI is most useful in the hands of someone who can tell good output from bad. Experience makes you a better AI user, not a redundant one.

Good news for experienced workers

Studies have found AI often helps less-experienced workers the most: in one widely cited study of customer support agents, an AI assistant raised productivity by about 14% on average and about 34% for newer, less-skilled workers. But the same research on employment shows experienced workers' jobs holding steady. Deep experience, judgment, and relationships are exactly what AI can't supply.

In a fast-changing economy, learning can't stop at graduation, and the good news is that there are more free and low-cost ways to learn than ever, including AI itself as a patient personal tutor, at any age.

It's never too late

Research on adult learning is encouraging: older adults can and do learn new skills well, especially when learning builds on experience and has a clear purpose. Lifelong learners often gain more than job skills: learning new things is widely associated with staying mentally sharp and socially connected.

Where to learn: free and low-cost options

AI as your tutor

Lesson 5 showed AI can be a patient teacher. For learning job skills, try:

Then use AI to explain concepts, quiz you, and practice, while checking key facts about certifications and requirements with official sources.

A practical learning strategy

1. Start from what you know. Build on your experience rather than starting from scratch. A nurse learning health IT or a mechanic learning electric vehicles brings enormous value.
2. Learn AI in your own field first. How are people in your line of work using it? That's often the fastest, most valuable learning.
3. Learn by doing. Small real projects teach more than watching videos.
4. Little and often. Twenty minutes a day beats a four-hour cram once a month.
5. Learn with others. Classes, study groups, and mentors keep you going.

For small businesses and self-employed people, AI can act like an affordable extra staff member for writing, marketing, bookkeeping help, customer questions, and planning, letting one person do more of what used to require a team, though it also brings new risks and new competition.

How many are using it?

Adoption is growing but uneven. U.S. Census Bureau surveys in spring 2026 found about 17–20% of businesses reported using AI to produce goods or services, with larger firms using it most (about 37% of firms with 250+ employees) and use among firms with fewer than 20 employees lower and not growing much. Many more owners use chatbots informally for everyday tasks, which such surveys may not fully capture.

Practical uses for a small operation

Opportunities for self-employed people

• Starting something is cheaper. One person can now create a website, logo, and marketing materials that once required hiring several professionals.
• Experience becomes a business. Retirees and career-changers with deep expertise can turn it into consulting, coaching, or teaching, with AI handling much of the administrative work.
• Reaching more customers through translation and online presence.

Risks to manage

• Accuracy: AI-written claims about your products, prices, or legal terms must be checked. You're responsible for what you publish.
• Customer data: don't paste customers' personal or payment information into consumer AI tools (Lesson 7, Section 4).
• Sounding like everyone else: if every business uses the same AI, generic content blends together. Your personal voice and local knowledge are your advantage.
• Honesty rules: fake AI-written reviews are illegal (Lesson 6, Section 3).
• New competition: freelancers in writing, design, translation, and similar fields face pressure from clients using AI instead. Many are adapting by offering expertise, judgment, and personal service that AI can't.

If AI makes work more productive, the economy can produce more with less effort, but whether that becomes higher wages, lower prices, shorter hours, bigger profits, or more wealth for a few depends on choices made by businesses, workers, and governments, and history shows it can go either way.

What productivity means

Productivity is how much gets produced for each hour of work. Rising productivity is the main reason living standards have risen over centuries. It's why a typical family today has things that were luxuries or impossible a century ago. So faster productivity growth from AI could, in principle, be very good news.

But history shows the gains aren't automatic

• After World War II, U.S. productivity and typical workers' pay rose together for about three decades.
• Since the 1970s, productivity has kept rising, but typical workers' pay has grown much more slowly. More of the gains went to profits and to top earners.
• In 2026, workers' share of U.S. national income fell to about 52.8%, the lowest since records began in 1947, while corporate profits reached a record share. Economists note these trends mostly predate AI, but they shape how AI's gains may be divided.

Where AI's gains could go

Augment or automate?

Some economists, including Nobel laureate Daron Acemoglu, argue that a key choice is whether AI is designed and used mainly to replace workers or to augment them, making people more capable. Augmentation tends to raise wages and create new tasks; pure automation tends to cut costs and concentrate gains. The Stanford findings in Section 2 echo this: where AI assisted workers, employment held up; where it automated, beginners lost out.

Ideas on the table

People across the political spectrum have proposed responses. Each has supporters and critics:

• Education and retraining investments, widely supported, though their track record is mixed.
• Wage insurance that tops up pay for workers who have to take lower-paying jobs after displacement.
• Tax changes, such as adjusting how labor and capital are taxed so replacing workers isn't artificially cheaper than keeping them.
• Worker voice: unions and contracts setting rules for AI use. In 2023, Hollywood writers won contract terms limiting how studios can use AI in scriptwriting.
• Profit-sharing and employee ownership, giving workers a stake in productivity gains.
• Shorter workweeks, taking some gains as time rather than money.
• Universal basic income: regular payments to everyone. Pilots have been tested; it remains debated on cost and effects.
• Letting markets adjust, with a lighter touch, on the view that competition will pass savings to consumers and create new jobs, and that heavy intervention could slow beneficial progress.

Both outcomes are possible. Which one we get depends partly on public understanding and public choices, which is one reason lessons like this exist.