Concept

A machine can make a task cheaper without deciding what happens to every worker’s income.

Issue

Learning a new tool is only part of a move; you also need a way into work that pays.

Solutions

Understand what a realistic move needs, then test one paid alternative.

Say a customer shows you a version of something you normally make, produced with a machine’s help. It’s good enough for what they need, and they want to pay less. You can still do the work. What’s changed is the price they’ll pay for that result.

Your next question is bigger than which tool to learn. Will there be enough paid work using the new way of doing it? Could you move into a different role? And could you manage the time and cost of getting there?

These questions pull a cheaper task apart from your ability to earn. They also give you something definite to look into before you sign up for a course or leave the work you have now.

This module is for working adults whose paid tasks could really be automated, and who have some way into other paid work or into retraining with support.

What you’ll get from this module

In this module

By the end of this module, you’ll know how to weigh up one possible move into paid work as a familiar task changes. You’ll learn to tell apart three things: whether people want the work, whether you can get into it and what it pays. You’ll choose a situation from your own work, and finish with a small test or a barrier you still need to deal with. We’ll use accounts from banking, telephone and newspaper history, a study of help for workers who lost their jobs, and a review of job-training experiments.

When a task gets cheaper, what happens to paid work?

Economic working paper: less work on each task can still leave more work overall.

James Bessen’s May 2016 paper pulls a task apart from the whole job it belongs to. Machines can take over part of a job while other work stays. What happens to jobs also depends on how much people want to buy.

His example from US banking describes growth in bank teller work, counted as full-time jobs, from the late 1990s and early 2000s into the early 2010s, even with cash machines. This measure turns all the hours worked into full-time jobs. It isn’t a count of people.

Bessen explains that cash machines made branches cheaper to run, so banks opened more of them. Less teller work in each branch could go along with more branches that needed staff.

Why it matters: That period of growth shows why a cheaper task doesn’t settle how much paid work will be left. You need to think about both the work saved and how much people want the work that’s left. The example doesn’t make that balance permanent, or show what happened to each teller’s income.

Read the economic paper on automation and occupations.

Historical employment study: whose recovery are we counting?

The total number of jobs tells you something different from what happens to the people already doing them. James Feigenbaum and Daniel P. Gross’s 2024 study looks at how US telephone exchanges were automated from 1920 to 1940.

They compare cities that switched to dial phones at different times. For women already working as operators, they also link census records and compare them with similar local women. Both parts of the study look at cities with at most 100,000 people in 1920.

For later groups of white, American-born women aged 16–25, the study finds no clear drop in jobs overall. Yet women who were already operators were more likely to end up in lower-paid jobs, or not working, ten years later. The findings depend on the historical comparisons, and on which women could be traced across the records.

Why it matters: So new openings for people starting out can go along with losses for people leaving the old work. A growing job market doesn’t tell you whether those particular workers got back to where they were.

Read the study of telephone automation and workers’ outcomes.

Government follow-up: finding work and replacing income are different outcomes.

Ethel L. Best’s 1933 government study gives a closer look at income after a change like this. In one city, it interviewed 78 of the 116 operators who lost their jobs when the exchange switched over. Most of those 116 had been temporary staff, or former operators who had come back.

Of 29 women working outside the telephone company who gave their pay at the last interview, 25 earned less than before. This follow-up covered a chosen group, about a year after the change, during the Great Depression.

Why it matters: Finding another job and earning as much as before were different results. That matters when a new role sounds promising: the fact that it exists is only the start of the question about income.

Read the government follow-up on displaced operators.

What does a cheaper task really change?

Workplace study: training existed, but access differed.

Roberta Hill’s 1984 paper looks at a different change, at The Star, a newspaper in Christchurch, New Zealand. Her study, from 1980 to 1983, used observation, interviews, conversations and documents to look at changes in how the paper was produced and how people worked.

Before the switch in May 1980, 26 linotype operators, who set lines of metal type, retrained to use computer terminals. No linotype operator or compositor (the workers who arranged type for printing) lost their job in the first stage of the change.

Other workers had a different path. Teletypesetters were another group who typed text to be set. Their numbers fell from ten in early 1980 to two by early 1983. Hill reports an apprenticeship for one of them; others left and weren’t replaced.

Long-standing printing tradesmen got first claim on retraining for the terminals. Getting into the new work depended on job agreements and bargaining between groups, as well as on learning the equipment.

The study records a change. It doesn’t single out what training alone did. And the fall in staff numbers doesn’t tell us what work or income the people who left found later.

Why it matters: Training was on offer in this workplace, but the chance to use it differed between groups. Being willing to learn wasn’t enough. Workers also needed permission, and a real chance, to move into the new work.

Read the workplace study of newspaper retraining and access.

Working paper using differences in program approval: support can be part of the transition.

Benjamin G. Hyman’s February 2022 paper looks at US workers who lost their jobs because of trade, and who could get help from a program called Trade Adjustment Assistance. The program paid people to train and gave them income while they did.

Some of the officials who decided on applications were more likely to approve them than others. Hyman used those differences to work out what approval changed for workers, using job records for people aged 22–65. He didn’t pick by chance who would finish training.

After a first stretch of lost earnings while people trained, the paper estimates about $50,000 more in earnings, added up over ten years, in 2010 US dollars. The yearly gain fades away by year ten.

An added-up gain counts the earnings differences across all the years together. It can stay positive even after the yearly advantage has gone. The timing matters, because someone starting training has to get through the stretch before any benefit arrives.

Why it matters: This result is about getting a whole package of support, for workers who lost jobs because of trade. It can’t be read as what you’d get back from paying for a course yourself. It gives a reason to look at income support as well as the training on offer.

Read the working paper on Trade Adjustment Assistance.

Research review: training helps on average, with substantial differences between programs.

David Roodman and Maxim Massenkoff’s August 2026 report reviews 56 US job-training studies in which people were put into groups by chance. The programs mostly served adults and young people on low incomes, not workers chosen because machines had taken their jobs.

For programs that were mainly about training, the review estimates that the share of people in work rose by 1.7 percentage points in years 3–5. It estimates $791 more in yearly earnings before tax over that time, in 2025 dollars.

These are averages for each person offered the program, using different sets of estimates for jobs and for earnings. They aren’t the result you should expect for everyone who finishes a program.

Some programs tied to particular kinds of work did better, but copying successful programs in other places was hard. The review also asks how much of Hyman’s earlier results came from training, and how much from the income support that came with it.

Why it matters: Together, the findings give training a real place without making any course the automatic answer. The program, the work it leads to and the support you get while learning all need a close look.

Read the evidence review of worker retraining.

Ask the right things about a course

What this means for you

Start with something you get paid to produce, and a role or service you could realistically move into. Keep your present income going as far as you can while you look into it. The aim is to learn enough about one option to decide what to do next.

Look for a change in paid work, not just an impressive demonstration

Name the task a customer or employer pays you to do. Then look for signs that they keep accepting a cheaper result made with a machine’s help, at the quality they need. Count the time and cost of checking it and putting it right.

A demonstration alone doesn’t answer that question. Compare similar work: have fees, paid hours or demand changed while your work is still good enough? Think about other causes before you put a change down to automation.

Now think about demand. Are more customers creating enough paid work to make up for the time saved on each task? The banking example shows why these two sides go together.

You may find that one task is under pressure while other paid work stays. Or you may find that extra customers aren’t making up for the lost hours. Keep these two things apart, so you can look into the task that’s really changing.

Ask whether you can enter the destination

Start by looking into a move inside your workplace, or training with support. Ask about a clear role, what it takes to get in, how many places there are, whether you’re paid while you learn and how the move would work.

The newspaper case shows how much it matters whether you’re eligible. A course can teach you the equipment while the job itself stays out of your reach. If you don’t meet the conditions to get in, find out whether there’s a route that works, or support, before you put in more time.

For a role somewhere else, ask an employer or customer what result they’d pay for and what you’d have to show to qualify. Nearby work or a specialist corner of the market belongs on your list only when the demand and a way in both look likely.

A job advert is a lead to follow up. It doesn’t show that you can get the work, or earn enough from it. A clear answer about hiring or buying is worth more than a general promise about the skills of the future.

Treat the training arrangement as part of the decision

Look past what the course teaches. Ask how many people finish, whether they get into work they can actually get, what they earn and what support there is while you learn. Compare the answers with what you’d need to get in and the time you have.

The training studies are about particular programs and particular people. Their results give you reasons to look into things. They don’t give you an amount to add to your own income forecast.

Count the time away from paid work, the fees, travel or moving home, caring for others and any support on offer. A job that advertises higher pay may still be hard to reach if the move takes away income you need now.

If those costs put the route out of reach, look for a different arrangement or source of support. A barrier to getting in isn’t proof that you’re unwilling to learn.

Use a small test to answer the missing question

Agree limits on time and money that you can afford before you make a sample of work or try a small job. Choose a test that answers an important question about the work you hope to move into.

Say you make a sample using a new way of working, and it meets the standard needed. That tells you something about what you can do. If no possible buyer has looked at it, whether anyone would pay is still open.

Ask an employer or customer about the result, and on what terms they’d pay for it. If the demand looks real but your sample falls short of the standard, find what you need to learn. If you can do it but nobody wants to buy it, look into a different offer or a different role.

A promising small test is a reason to look further. It doesn’t show that enough paid work will be there to replace what you earn now.

Choose the next move from the result

Put three questions together. Is there paid demand? Can you get into the work? Can you manage the move? For a move into paid work, what you can do, a way into work that pays and a move you can afford all matter.

If demand, a way in and the costs of the move look workable, take the search or the small job further, within what you can afford. Change the test when one clear gap is left. If you can’t get in or demand is weak, keep your income going while you look for another route or for support.

Come back to the decision when something important about the work changes. Changes in hiring, the rules for moving roles, paid hours, time to learn or support can change what’s possible. There’s no need to make up a deadline from someone else’s experience in history.

Where are you in thinking about a move?

How about your daily life?

Which paid task has become cheaper to produce, or is changing in a way your employer or customers already accept? Where are you thinking about a course, a move inside your workplace or a different service?

You might choose a regular report, a production task or a service you sell. Focus on the result someone pays for. Then work out what you need to understand about its future, or about another option.

Write down situations you could look into. One good idea is enough.

A situation in my workWhat I need to understand
________________________________________
________________________________________

Which situation would be most useful to look into now?

My chosen situation: ____________________

Choose a paid task that could really be automated, where you have some way into other work or into training with support. Take it into the exercise below. You can change your choice if the first route turns out to be closed to you.

Test one paid alternative

Start with your chosen situation. If you came straight here, choose a paid task that could really be automated, and another role or a training route with support that you can realistically look into.

Protect your present income as far as you can. Keep any test within limits on time and money that you can afford. If there’s no other option you could take, naming the support you need is a useful result.

  1. Describe the paid task and the evidence of change. Name the result that’s needed and the cheaper version on offer. Write down whether people really accept it at the quality needed, counting the cost of checking. Compare fees, hours and demand for similar work, and think about other causes.
    Task, observed change and what remains uncertain: ____________________
  2. Check an accessible destination. Look first at a role inside your workplace, or training with support. Ask what an employer or customer would pay for, what it takes to get in and what support there is for the move.
    Possible destination, demand evidence and entry route: ____________________
    If the way in is blocked, find another route or the support you’d need before you spend more on this option.
  3. Choose the smallest useful test. Set a limit on time and money that you can afford. Use a sample or a small, limited job to test what you can do. Use what an employer or customer tells you to test whether anyone would pay. A practice piece alone can’t answer both.
    Missing question, test and limits: ____________________
  4. Check whether the transition is feasible. Compare realistic pay with unpaid time spent learning, fees, travel or moving home, caring for others and any support. Mark any amounts or conditions you don’t know as missing information.
    Costs, support and unresolved conditions: ____________________
  5. Decide what the result supports. Look further into a promising option, deal with one clear gap, or keep your income going while you look elsewhere. If you’re missing an answer you need, name the question and who you can ask.
    One feasible next action, or support I need: ____________________
    A meaningful work change that would make me revisit this: ____________________

Ask Jing

Become a Jingius.

Test one move before you commit

Loading interactive tool…

Your next step

You’ve finished when you have one clear next step or a clearly named barrier, with the evidence behind that choice. You don’t need to commit to a new career to finish the exercise.

A note on the evidence

The cases come from different workplaces and different times. The telephone comparisons depend on old records. Best’s follow-up describes a chosen group during the Great Depression. Hill’s workplace study shows how the arrangements shaped who could get into the new work, without singling out what training did.

The training evidence is about programs with different people and different support. Roodman and Massenkoff’s report hasn’t yet been through peer review, and what it says about Hyman deals with his earlier draft. Keep each result tied to the people, the program and the outcome that were studied.

How we prepared this module

This module links accounts of how work changed in the past with research on training and support during a move. A made-up example about a work sample pulls apart what you can do from whether anyone will pay. The exercise helps you look into one possible move and write down the next step or barrier.

Sources