AI, Induced Demand and the Shape of Work

Work is getting faster. But it is not getting lighter.

Across organisations, the introduction of AI is increasing capacity in almost every direction. Writing, analysis, communication, production. Tasks that once took hours now take minutes, and things that once required coordination can now be done alone.

And yet, the experience of work is not simplifying. If anything, it is becoming more dense. More output, more communication, more expectation. The system is not easing. It is expanding.

There is a concept from urban design that helps explain this. Induced demand.

The idea is simple. When you increase capacity within a system, you often increase the use of that system, not because people suddenly behave irrationally, but because they adapt to what becomes easier. Build more roads, and more people drive. Reduce travel time, and people travel further. Remove friction, and activity increases to fill the space. The system reorganises itself around the new conditions.

But there is a lesser discussed version of this idea, one that does not start with adding capacity, but with changing it.

In the early 2000s, Seoul removed an elevated highway and restored the Cheonggyecheon Stream beneath it. Induced demand would suggest disruption, congestion, overflow.

Instead, something else happened. Traffic decreased. Pedestrian movement increased. Public transport usage rose. The space did not empty. It transformed. Not because demand disappeared, but because the system changed what it made easy.

This is where the concept becomes more useful. Induced demand is not about roads themselves. It is about behaviour, and more specifically, about what systems make possible, convenient, and normal.

If we translate the logic cleanly:

Roads become dominant workflows.

Congestion becomes friction, overload, and inefficiency.

Adding lanes becomes adding tools, automation, and capacity.

Removing lanes and building alternatives becomes redesigning how work actually flows.

Most organisations are currently doing the equivalent of road widening with AI.

Faster outputs, more content, more communication, and more throughput, already producing a familiar pattern.

The more capacity AI creates, the more work expands to fill it.

This is not a failure of the technology. It is a feature of the system. When you reduce the cost of producing something, you tend to get more of it. When you remove friction from communication, communication increases. When output becomes easier, expectations of output rise accordingly. The system recalibrates, not to less pressure, but to a new equilibrium, a pattern that will feel familiar if you’ve come across the idea of Jevons paradox.

There is a parallel here that is often described simply, garbage in, garbage out.

But the implication is broader than data quality or prompting. If the underlying system is poorly designed, the outcomes will scale accordingly.

AI does not just improve what exists. It amplifies it.

Which is why increasing capacity without redesign rarely solves the problem it is intended to address. It reinforces it.

The Cheonggyecheon project offers a different move. It did not attempt to optimise the existing system of traffic flow. It changed the system itself.

It removed a dominant pathway, introduced a more attractive alternative, and reshaped the environment in a way that made different behaviours easier to adopt. And behaviour followed.

This is the move that is still largely missing in how organisations approach AI. The focus remains on acceleration, with faster processes, more efficient workflows, and increased output layered onto the same underlying structures. But capacity alone does not determine outcomes. Design does.

Every system quietly rewards certain behaviours.

Speed over reflection, volume over judgement, responsiveness over clarity.

These are not neutral choices.

They are embedded into how work flows, how decisions are made, and what gets recognised.

AI does not change these dynamics by default. It scales them.

Which means the question is not simply how to use AI to do more. It is what kind of system it is being introduced into.

One that makes interruption, volume, and reactive work easy, or one that is deliberately structured to make something else easier instead.

Clarity.

Synthesis.

Decision-making.

Restraint.

The lesson from urban design was never just about traffic. It was about understanding that systems produce more of what they make easy.

When you remove a constraint without redesigning the system, the pressure does not disappear. It redistributes. When you redesign the system itself, you change the behaviour that creates the pressure in the first place.

AI has already begun to reshape work. What is less clear is whether it will simply scale existing patterns, or whether it will be used to design something different. Not just faster systems. Better ones.

If you’re thinking through how AI is reshaping work in your organisation, and where it is reinforcing rather than relieving pressure, this is exactly the kind of work I partner on. You can reach out via info@dialecticalconsulting.com.au or contact me via LinkedIn.