Labor Day has always been about more than having Monday off. At its core, it asks us to think about the relationship between people and work: who performs it, who benefits from it, and what dignity and power look like inside the systems we build.

In 2026, that conversation has a new participant.

AI is no longer limited to helping us write an email or summarize a document. Increasingly, AI systems can plan, use tools, coordinate tasks, and continue working after the human has stopped typing.

That changes the future-of-work question.

For years, we have asked: What jobs will AI replace?

I think that framing is becoming too narrow.

A more important question is: What should remain meaningfully human when the computer no longer needs to wait for us to operate it?

---

Workers Do Not Simply Want Maximum Automation

Recent research from Stanford's SALT Lab offers an important clue.

Researchers studying the future of work with AI agents surveyed 1,500 workers across 104 occupations and mapped hundreds of tasks against both AI capability and worker preferences. One of the most interesting findings was a mismatch between what technology may be capable of automating and what people actually want automated.

Across 844 tasks, workers preferred a higher level of human agency than experts considered technically necessary in 47.5% of cases. In 47 of the 104 occupations studied, the most common preference was an equal partnership between human and AI rather than full automation.

That distinction matters.

Capability answers: Can the machine do this?

Human-centered design has to ask another question: Should this be the relationship we create?

Those are not the same thing.

---

The Interface Is Becoming a Relationship With Delegated Work

My background is in software architecture, user experience, security, and increasingly the systems that govern AI action. For most of the history of human-computer interaction, the computer waited.

We clicked. It responded.
We typed. It processed.
We selected an action. The software performed it.

The interface was largely the place where human intent became machine instruction.

Agentic systems change that sequence.

A person can express an intent and the software may decide how to pursue it, which tools to use, what intermediate steps to take, and when to continue without asking again.

The interaction is no longer simply human → interface → command → computer.

It is increasingly human → intent → autonomous system → action → consequence.

That means HCI has to expand beyond usability.

A system can be easy to use and still leave the person with very little agency.

---

Human Agency Is Not the Same as Human Involvement

We often respond to AI risk by saying we will keep a "human in the loop."

But a human can be present and still be powerless.

If I am asked to approve an action I do not understand, my presence has not created meaningful control. If an agent performs twenty consequential steps and shows me the result only afterward, I may be involved without being able to influence the outcome. If automation makes every decision but leaves me responsible when something goes wrong, we have preserved accountability while removing agency.

Good human–AI interaction therefore cannot be measured by how often the interface asks a person to click Approve.

It should be measured by whether the person has meaningful choices.

• Can I decide what to delegate?
• Can I understand what the system intends to do?
• Can I see when its behavior departs from my intent?
• Can I interrupt it before an irreversible consequence?
• Can I take control back?
• And can I understand what happened afterward?

These are interface questions. They are also organizational, ethical, and increasingly economic questions.

---

The Goal Should Not Be to Keep Humans Busy

Preserving agency does not mean protecting every existing task from automation.

Some work should disappear.

Repetitive administrative work, unnecessary coordination, mechanical information processing, and tasks people find exhausting may be exactly where AI creates the most human value. In Stanford's research, the most commonly cited reason workers wanted automation was to free time for higher-value work.

The mistake would be assuming that every hour removed from human execution should also remove humans from judgment.

There is a difference between eliminating labor and eliminating agency.

A good system can do more work for us while giving us greater control over what matters.

That is the design opportunity.

---

A Different Labor Day Question

This Labor Day, I am less interested in predicting whether AI will create more jobs than it destroys.

We will spend years arguing over those numbers.

I am more interested in what kind of relationship we are designing between people and increasingly capable machines.

The future of work should not be defined by maximum automation.

It should be defined by intentional delegation.

Machines can execute more. Humans can spend less time operating software. But the systems we build should make it easier—not harder—for people to understand, direct, question, interrupt, and ultimately own the decisions that matter.

The next generation of human-computer interaction may not be about making computers easier to use.

It may be about making increasingly autonomous computers easier for humans to remain meaningfully in charge of.

That is a much harder design problem.

And it is one worth solving before the future of work simply happens to us.

---

Primary Sources

• Stanford SALT Lab — Future of Work with AI Agents: https://futureofwork.saltlab.stanford.edu/
• Microsoft 2026 Work Trend Index — Agents, human agency, and opportunity: https://www.microsoft.com/en-us/worklab/work-trend-index/agents-human-agency-and-the-opportunity-for-every-organization

---

Frequently Asked Questions

What is human agency in AI systems?

Human agency is the practical ability to make meaningful choices about what an AI system may do, understand its behavior, intervene before irreversible consequences occur, and retain ownership and judgment over consequential outcomes.

How should humans work with AI agents without losing agency?

Humans should work with AI agents through intentional delegation rather than uncritical automation. This requires systems designed with explicit execution boundaries, understandable rationale, runtime checkpoints for high-impact actions, and clear recovery paths rather than rubber-stamp approval dialogs.

What does human agency mean in the future of work?

In the future of work, human agency means ensuring that eliminating mechanical labor does not eliminate human judgment. Workers remain empowered to determine boundaries of delegation, question autonomous reasoning, and direct the purpose of the tools serving them.