The Last Hour: Human Labour, Salaries and AI

There is a strange story in the Gospel of Matthew about work, wages and fairness.

A landowner needs workers for his vineyard. He goes out early in the morning and hires some. Then he goes out again at nine, again at noon, again at three, and finally at five in the afternoon.

Every time he finds people standing idle, he says essentially the same thing:

Come and work in my vineyard.

At the end of the day, he pays them all the same amount: one denarius.

The workers who arrived first are angry. They have worked through the heat of the entire day. The people who arrived at the last hour have worked for barely an hour, yet they receive the same wage.

The landowner's answer is uncomfortable:

You agreed to one denarius. You received one denarius.

And then comes the line that makes the story particularly interesting:

"I want to give the one who was hired last the same as I gave you."

We usually read this story as a lesson about God's grace.

But it can also make us think about something very contemporary:

What does it mean to be a worker when technology can increasingly do the work?

And what should an employer do when artificial intelligence makes one hour of human work potentially worth what previously required an entire day?

God keeps going back for workers

There is something peculiar about the landowner.

He already has workers.

Yet he keeps going back.

Nine o'clock.

Noon.

Three o'clock.

Five o'clock.

Why?

The parable doesn't say that the vineyard suddenly became more productive at five o'clock. It doesn't say that the late workers possessed some extraordinary skill that the morning workers lacked.

They were simply people who had not yet been hired.

When the landowner finds them, they explain why:

"Because no one has hired us."

And the landowner still says:

Go into my vineyard.

This is where I would slightly challenge the usual formulation of the theological argument.

God does not need human beings in the sense that a company needs employees.

God, by definition, is not a manager desperately trying to fill a vacancy.

The Christian understanding is closer to something more interesting:

God chooses to accomplish His work through human beings even though He does not have to.

The distinction matters.

A God who needs us would be dependent on us.

A God who does not need us but nevertheless invites us to participate gives human work a different dignity.

Our usefulness is not what creates our value.

The invitation comes first.

And that may be one of the deepest ideas in the parable.

God keeps going back to the marketplace because He does not write people off simply because they arrived late.

The person who discovers his vocation at twenty matters.

The person who discovers it at forty matters.

The person who changes careers at fifty matters.

The person who finally gets an opportunity at sixty matters.

Even the person who arrives at the metaphorical five o'clock hour matters.

Especially there.

The last hour is becoming our economic reality

For centuries, employment was built around a relatively simple equation:

human time → human output → salary.

If a company needed more output, it generally needed more people or more hours.

AI is beginning to break that equation.

A developer can use AI to generate and review code.

A designer can produce dozens of variations before lunch.

An accountant can automate parts of reconciliation.

A lawyer can search and summarise enormous quantities of documents.

A marketer can generate and test hundreds of pieces of content.

A small company can suddenly perform work that previously required an entire department.

This is real productivity potential, but it would be premature to conclude that human employment is therefore becoming obsolete.

The latest International Labour Organization research estimates that around one in four workers globally are in occupations with some degree of generative-AI exposure, while also finding that most jobs are more likely to be transformed than completely eliminated because human input remains necessary.

And the empirical evidence is still developing. A 2026 ILO review finds productivity gains from generative AI, but says those gains remain uneven and that large-scale job displacement has so far been limited. It identifies inequality, reduced opportunities for younger workers and changes in job quality and worker autonomy as important risks.

So perhaps the question is not:

"How do we stop AI from replacing people?"

A better question is:

"What do we do with the productivity created by AI?"

That is an economic question.

But it is also a moral one.

The mistake is thinking that salary should equal minutes of labour

The vineyard parable contains an uncomfortable economic idea.

The first workers think:

"More hours = more money."

The landowner says:

"You received exactly what we agreed."

The parable is not a treatise on modern compensation. We should not pretend that it establishes a universal rule that every worker should receive the same salary regardless of contribution.

But it does expose something important:

Human worth and economic productivity are not identical concepts.

Suppose AI makes a programmer twice as productive.

Should the programmer's salary immediately be cut in half because the same output now requires half as much human time?

Obviously not.

And if AI makes a company 30% more productive, why should the only possible destination of that 30% be shareholder returns?

There are several possible destinations:

• lower prices;
• higher profits;
• higher salaries;
• shorter working hours;
• more hiring;
• more investment;
• better products;
• more training;
• or some combination of these.

The technology does not decide the distribution.

People do.

That is why the AI debate is ultimately not only about technology.

It is about institutions.

The employer of the AI age should keep going back to the vineyard

This is where the parable becomes particularly provocative.

An employer in the AI era has an unprecedented opportunity.

Instead of saying:

"AI can do this job, therefore we don't need the worker."

the employer can ask:

"AI can do this part of the job. What can this person do now that they couldn't do before?"

That is a fundamentally different philosophy of automation.

Imagine a company where AI eliminates 40% of the repetitive work performed by an employee.

There are at least two ways to respond.

Model one: extraction

The company keeps the same workload and uses AI to eliminate positions.

Fewer people produce the same output.

The productivity gain flows primarily to the owners.

Model two: participation

The company keeps the people and changes the work.

The employee spends less time on repetitive tasks and more time on:

• judgement;
• customer relationships;
• creativity;
• mentoring;
• experimentation;
• quality;
• research;
• strategy;
• craftsmanship.

The company becomes more productive.

The worker becomes more capable.

And the productivity dividend is shared.

The second model is much closer to the spirit of the vineyard.

The landowner does not ask:

"How little labour can I extract from these people?"

He keeps finding people and inviting them into productive participation.

But what about the workers who started in the morning?

Their complaint deserves to be taken seriously.

This is where the parable is often misunderstood.

The morning workers aren't crazy.

They worked all day.

They bore the heat.

They had a contract for one denarius, and they received exactly one denarius. But psychologically they expected their greater sacrifice to produce a greater reward.

This is precisely what happens when AI changes productivity.

Imagine two employees.

One has spent ten years building expertise.

The other joins after AI has made the field dramatically easier.

The veteran may reasonably ask:

"Why should someone who arrived yesterday have access to the same opportunities that took me ten years to build?"

That is a legitimate question.

The answer should not be:

"Experience doesn't matter."

Nor should it be:

"The newcomer deserves nothing because they arrived late."

The better answer is to distinguish contractual fairness, contribution, opportunity and human dignity.

People should be compensated according to clear agreements and relevant contribution.

But access to meaningful work should not become a privilege reserved for people who happened to enter an industry before AI transformed it.

That distinction will become increasingly important.

AI could create a terrible paradox

There is a potential paradox at the heart of automation.

AI could make society dramatically more productive while simultaneously making it harder for people to enter the labour market.

This is especially important for younger workers.

Historically, a junior employee learned by doing relatively low-value tasks.

The junior programmer fixed small bugs.

The junior lawyer reviewed documents.

The junior analyst cleaned spreadsheets.

The junior designer made variations.

Those tasks were economically useful, but they were also training grounds.

If AI takes all of those tasks away, companies may discover something uncomfortable:

They have eliminated the very work through which the next generation used to become experienced.

The result could be a labour market with many highly productive senior workers and far fewer pathways for juniors to become senior.

The ILO has specifically identified erosion of employment opportunities for younger workers as one of the emerging risks of generative AI.

That is another reason employers should keep going back to the marketplace.

Not because every human must be employed at any price.

But because society needs pathways into competence.

Someone has to get their first opportunity.

So what should an AI-era employer actually do?

The principle can be surprisingly simple:

Automate tasks before you automate people.

Then redesign the jobs around what humans are uniquely positioned to contribute.

That means several practical commitments.

1. Share the productivity dividend

When AI substantially increases productivity, companies should have an explicit policy for deciding where the gains go.

For example:

AI productivity gain →

• some to customers;
• some to shareholders;
• some to employees;
• some to training;
• some to experimentation.

It does not have to be an equal split.

But it should be a conscious decision rather than an accidental transfer of wealth.

2. Keep hiring people at the beginning of their careers

Companies should treat junior positions partly as economic production and partly as human-capital formation.

An AI-enabled junior employee may not need to spend six months doing repetitive work.

Good.

Give them harder work sooner.

Pair them with experienced people.

Make them responsible for outcomes.

Use AI as their leverage rather than their replacement.

3. Reduce working hours where productivity allows it

If AI genuinely allows a team to produce the same value in four days instead of five, there is another possible dividend:

time.

The productivity revolution does not necessarily have to produce an economy in which humans work harder for machines.

It could produce an economy in which machines work more and humans get some of their time back.

4. Pay for responsibility, not keyboard activity

AI makes hours worked an increasingly poor proxy for value.

A person who makes an excellent decision in thirty minutes may create more value than someone who spends eight hours producing mediocre output.

Compensation systems should therefore increasingly consider:

• responsibility;
• expertise;
• outcomes;
• scarcity of skills;
• customer value;
• risk;
• judgement;
• leadership;
• and institutional knowledge.

AI changes the economics of work.

It should therefore also change how we think about work.

5. Make retraining part of the employment contract

If an employer introduces AI that materially changes a job, training should not be treated merely as the employee's private responsibility.

The company benefits from the technology.

It should invest in helping its people adapt to it.

This is not charity.

It is organisational risk management.

And regulation?

The answer should not be to regulate AI out of existence.

Nor should it be to leave everything to the market.

The useful middle ground is to regulate the consequences and the power relationships, particularly where AI can materially affect someone's livelihood.

The European Union has already moved in this direction.

Under the EU AI Act, certain AI systems used for recruitment, worker management, allocation of tasks, promotion, termination, and worker performance evaluation are classified as high-risk.

The framework also requires meaningful human oversight for high-risk systems, including the ability for appropriately empowered humans to understand, monitor and override AI outputs.

For workplace deployment of high-risk AI, employers must also inform affected workers and their representatives.

These are useful principles.

But regulation should go further than simply asking whether an algorithm is accurate.

An AI system can be statistically accurate and still produce a bad labour-market outcome.

Consider an algorithm that perfectly identifies which employees can be eliminated.

The algorithm may be excellent.

The question is whether the decision it optimises is desirable.

That is not a machine-learning problem.

It is a governance problem.

A better regulatory framework

I would build AI labour regulation around five principles.

Transparency

Workers should know when AI materially affects:

• hiring;
• firing;
• promotion;
• compensation;
• scheduling;
• performance evaluation;
• workload;
• or access to opportunities.

Human accountability

There should always be a person with authority who is accountable for consequential employment decisions.

"The algorithm decided" should not be an acceptable answer.

Right to challenge

Workers should have a meaningful way to contest an AI-assisted decision.

Not a button that sends their complaint into a black hole.

A real review process.

Productivity sharing

Where AI produces substantial measurable productivity gains, companies should have incentives—or potentially sector-specific obligations—to invest part of those gains into workers, training, wages, reduced hours or new employment pathways.

The exact mechanism requires economic experimentation.

Protection of the labour-market entry point

Governments and companies should pay particular attention to apprenticeships, internships, junior roles and other entry-level pathways.

If AI destroys the first rung of the career ladder, we may create a generation that cannot accumulate the experience required for the jobs that remain.

The one-denarius principle

There is another beautiful detail in the Gospel story.

The employer pays everyone at the end of the day.

He starts with the people who arrived last.

Why?

Perhaps because the story wants us to look at them first.

The people who have been waiting all day are not invisible.

They are not worthless because nobody hired them.

They are not told:

"You should have arrived earlier."

They are invited into the vineyard.

And then they receive enough to live.

That does not mean that every worker should receive exactly the same salary.

It does mean that economic value cannot be the only measure by which we decide who deserves a place in society.

That distinction will become critical as AI becomes more capable.

If a machine can perform a task ten times faster than a human, the human being does not become ten times less valuable.

The task became cheaper.

The person did not become cheaper.

The last hour may be the most important hour

Perhaps this is the deeper connection between the Gospel and AI.

God keeps going to the marketplace.

Not because He lacks workers.

Because He chooses to invite people into the work.

And He keeps doing it until the last hour.

The AI revolution is going to tempt companies to do the opposite.

When technology becomes powerful enough, the temptation will be:

Stop going back to the marketplace.

We have enough workers.

We need fewer people.

Let the machine do it.

But there is another possible future.

A company could say:

The machine can do more, therefore more people can participate in meaningful work.

The senior engineer can work on architecture instead of boilerplate.

The junior engineer can work on real systems earlier.

The small business can afford capabilities previously available only to large corporations.

The older worker can remain productive without competing purely on speed.

The person changing careers at forty can use AI to overcome the experience gap.

The person who was standing in the marketplace at five o'clock can still enter the vineyard.

That is the future worth designing for.

Not a world without human labour.

A world where technology makes human labour more human.

The fundamental question of AI is therefore not:

"How many workers can we eliminate?"

It is:

"How many more people can participate in meaningful work because machines are helping us?"

The vineyard still needs workers.

The technology has simply changed what the work looks like.

And perhaps the most important lesson of the parable is that the person who arrives in the last hour is still worth inviting in.

Even when there is only an hour left.

#ai #labour #economics #future-of-work