The Productivity Paradox of AI: Why Work Is Increasing Before It Disappears

By Dr Luke Soon

Over the past year, an unusual paradox has begun to emerge inside organisations experimenting with artificial intelligence. Many leaders expected AI to dramatically reduce workloads. Instead, early deployments appear to be doing the opposite.

Employees report more emails, more messaging, more reviews and more coordination. Time spent on deep work has declined in many environments, while the volume of digital artefacts—documents, summaries, drafts and analysis—has increased dramatically.

At first glance, this appears to contradict the long-held belief that AI would usher in a productivity revolution.

In reality, what we are witnessing is a classic transition phase that has occurred in nearly every technological revolution. Before productivity improves, activity expands.

Understanding this dynamic is essential if we are to navigate the coming transformation of work.

The AI Productivity Paradox

The core paradox can be summarised simply:

AI lowers the cost of producing information, but raises the cost of coordinating it.

Large language models can generate documents, code, analysis, presentations and summaries in seconds. The barrier to creating content has effectively collapsed.

However, once content becomes cheap, something else becomes expensive: attention and coordination.

In many organisations today, AI is generating:

more reports more draft emails more documentation more meeting summaries more analysis

Yet all of this output still requires human oversight.

Someone must:

review the content validate accuracy correct hallucinations decide what matters integrate outputs into decisions

As a result, activity expands before productivity improves.

The Coordination Explosion

Economists have long recognised that coordination is the hidden cost of organisations.

For decades, companies invested heavily in technologies designed to reduce coordination friction:

email collaboration tools workflow platforms enterprise software

Artificial intelligence now sits on top of these systems.

But rather than immediately reducing coordination, it initially amplifies it.

Consider a simple example.

Before AI

An analyst writes a report.

After AI

An AI system produces five report drafts.

A human reviews them.

Another AI summarises the report.

A manager asks the AI for additional analysis.

The team reviews the results.

The total volume of work increases.

Not because the task became harder, but because the cost of generating information fell dramatically.

This is what economists refer to as the activity expansion effect.

When the cost of producing something drops close to zero, demand for it increases exponentially.

A Historical Pattern

The phenomenon we are seeing today is not unprecedented.

When electricity was first introduced into factories during the early twentieth century, productivity barely improved.

Factories simply replaced steam engines with electric motors, while keeping the same production layout.

The real productivity gains only appeared later when factories were redesigned around electricity.

Assembly lines were reorganised. Workflows were restructured. Entire production systems were rebuilt.

The lesson is simple:

Technology alone does not create productivity.

Organisational redesign does.

Artificial intelligence is currently in its early deployment phase, where organisations are adding AI tools to existing workflows rather than rebuilding workflows around AI.

The Next Phase: Task Rewiring

For AI to deliver meaningful productivity gains, organisations must fundamentally rethink the structure of work.

Research on the future of automation consistently highlights the same insight: the unit of transformation is not the job, but the task.

Most roles are composed of dozens of tasks, including:

analysis coordination communication decision-making documentation

Artificial intelligence can automate many individual tasks, but rarely the entire job.

The real transformation therefore occurs when organisations reconfigure how tasks are distributed between humans and machines.

In the coming decade, we are likely to see entire workflows redesigned around AI capabilities.

This transition will mark the true beginning of the AI productivity revolution.

The Rise of Autonomous Digital Agents

Another critical shift is the emergence of AI systems capable of planning and executing multi-step processes.

Today’s generative systems primarily assist humans by producing outputs.

The next generation of systems will increasingly:

plan tasks coordinate tools gather information execute workflows monitor outcomes

These systems will function as autonomous digital agents operating across organisational processes.

When this capability matures, the coordination burden that currently falls on humans may shift to machines.

Instead of humans coordinating AI outputs, AI systems will increasingly coordinate work itself.

The New Division of Labour

This shift implies a profound redefinition of human work.

Historically, most knowledge work involved:

gathering information analysing data producing artefacts coordinating teams

Artificial intelligence increasingly performs the first three functions.

The remaining domain where humans retain a critical advantage lies in:

judgement ethics context trust accountability strategic thinking

In other words, AI may automate tasks, but it simultaneously amplifies the importance of human responsibility.

The future of work may involve fewer routine tasks but far greater emphasis on oversight and decision-making.

The Transition Shock

There is another implication that governments and institutions have yet to fully confront.

The transition from augmentation to automation may not occur gradually.

Once AI systems begin coordinating workflows autonomously, the productivity impact could accelerate rapidly.

In many technological revolutions, productivity improvements arrive slowly for years and then suddenly surge.

If this pattern repeats, labour markets could experience a sharp adjustment once AI systems move from assistance to execution.

This possibility raises urgent questions about workforce adaptation.

Preparing for the AI Era of Work

The central challenge for organisations today is not simply adopting AI tools.

It is redesigning work itself.

Leaders should begin asking deeper questions:

Which tasks in our organisation exist purely because coordination was previously expensive? Which roles are primarily coordination roles? Which workflows could be executed autonomously by AI agents? Where do humans provide irreplaceable judgement?

The companies that succeed in the coming decade will not be those that deploy the most AI tools.

They will be the ones that rethink the structure of work entirely.

A Final Reflection

Artificial intelligence is often framed as a technology that will eliminate work.

That framing is too simplistic.

AI is not merely automating tasks.

It is transforming the economic architecture of organisations.

In the short term, this transformation may create more activity rather than less.

But over time, as workflows are redesigned and digital agents take on greater autonomy, the balance will shift.

The productivity revolution will not come from AI generating more content.

It will come from AI removing the hidden coordination layers that have shaped organisations for more than a century.

And when that happens, the nature of human work will change profoundly.

The question is not whether AI will reshape work.

It is whether our institutions, companies and societies are prepared for the speed at which that transformation may unfold.